agora inbox for [email protected]  

help / color / mirror / Atom feed

[PATCH v1 1/4] autovac: save all relopts instead of just avopts
223+ messages / 2 participants
[nested] [flat]

* [PATCH v1 1/4] autovac: save all relopts instead of just avopts
@ 2025-06-23 19:07 Nathan Bossart <[email protected]>
  0 siblings, 0 replies; 223+ messages in thread

From: Nathan Bossart @ 2025-06-23 19:07 UTC (permalink / raw)

---
 src/backend/postmaster/autovacuum.c | 119 ++++++++++------------------
 1 file changed, 43 insertions(+), 76 deletions(-)

diff --git a/src/backend/postmaster/autovacuum.c b/src/backend/postmaster/autovacuum.c
index 451fb90a610..f86c9fed853 100644
--- a/src/backend/postmaster/autovacuum.c
+++ b/src/backend/postmaster/autovacuum.c
@@ -192,8 +192,8 @@ typedef struct av_relation
 	Oid			ar_toastrelid;	/* hash key - must be first */
 	Oid			ar_relid;
 	bool		ar_hasrelopts;
-	AutoVacOpts ar_reloptions;	/* copy of AutoVacOpts from the main table's
-								 * reloptions, or NULL if none */
+	StdRdOptions ar_reloptions; /* copy of main table's reloptions, or NULL if
+								 * none */
 } av_relation;
 
 /* struct to keep track of tables to vacuum and/or analyze, after rechecking */
@@ -333,11 +333,11 @@ static void FreeWorkerInfo(int code, Datum arg);
 static autovac_table *table_recheck_autovac(Oid relid, HTAB *table_toast_map,
 											TupleDesc pg_class_desc,
 											int effective_multixact_freeze_max_age);
-static void recheck_relation_needs_vacanalyze(Oid relid, AutoVacOpts *avopts,
+static void recheck_relation_needs_vacanalyze(Oid relid, StdRdOptions *relopts,
 											  Form_pg_class classForm,
 											  int effective_multixact_freeze_max_age,
 											  bool *dovacuum, bool *doanalyze, bool *wraparound);
-static void relation_needs_vacanalyze(Oid relid, AutoVacOpts *relopts,
+static void relation_needs_vacanalyze(Oid relid, StdRdOptions *relopts,
 									  Form_pg_class classForm,
 									  PgStat_StatTabEntry *tabentry,
 									  int effective_multixact_freeze_max_age,
@@ -345,8 +345,6 @@ static void relation_needs_vacanalyze(Oid relid, AutoVacOpts *relopts,
 
 static void autovacuum_do_vac_analyze(autovac_table *tab,
 									  BufferAccessStrategy bstrategy);
-static AutoVacOpts *extract_autovac_opts(HeapTuple tup,
-										 TupleDesc pg_class_desc);
 static void perform_work_item(AutoVacuumWorkItem *workitem);
 static void autovac_report_activity(autovac_table *tab);
 static void autovac_report_workitem(AutoVacuumWorkItem *workitem,
@@ -1995,7 +1993,7 @@ do_autovacuum(void)
 	{
 		Form_pg_class classForm = (Form_pg_class) GETSTRUCT(tuple);
 		PgStat_StatTabEntry *tabentry;
-		AutoVacOpts *relopts;
+		StdRdOptions *relopts;
 		Oid			relid;
 		bool		dovacuum;
 		bool		doanalyze;
@@ -2033,7 +2031,7 @@ do_autovacuum(void)
 		}
 
 		/* Fetch reloptions and the pgstat entry for this table */
-		relopts = extract_autovac_opts(tuple, pg_class_desc);
+		relopts = (StdRdOptions *) extractRelOptions(tuple, pg_class_desc, NULL);
 		tabentry = pgstat_fetch_stat_tabentry_ext(classForm->relisshared,
 												  relid);
 
@@ -2069,7 +2067,7 @@ do_autovacuum(void)
 				{
 					hentry->ar_hasrelopts = true;
 					memcpy(&hentry->ar_reloptions, relopts,
-						   sizeof(AutoVacOpts));
+						   sizeof(StdRdOptions));
 				}
 			}
 		}
@@ -2095,7 +2093,7 @@ do_autovacuum(void)
 		Form_pg_class classForm = (Form_pg_class) GETSTRUCT(tuple);
 		PgStat_StatTabEntry *tabentry;
 		Oid			relid;
-		AutoVacOpts *relopts;
+		StdRdOptions *relopts;
 		bool		free_relopts = false;
 		bool		dovacuum;
 		bool		doanalyze;
@@ -2113,7 +2111,7 @@ do_autovacuum(void)
 		 * fetch reloptions -- if this toast table does not have them, try the
 		 * main rel
 		 */
-		relopts = extract_autovac_opts(tuple, pg_class_desc);
+		relopts = (StdRdOptions *) extractRelOptions(tuple, pg_class_desc, NULL);
 		if (relopts)
 			free_relopts = true;
 		else
@@ -2701,39 +2699,6 @@ deleted2:
 		pfree(cur_relname);
 }
 
-/*
- * extract_autovac_opts
- *
- * Given a relation's pg_class tuple, return a palloc'd copy of the
- * AutoVacOpts portion of reloptions, if set; otherwise, return NULL.
- *
- * Note: callers do not have a relation lock on the table at this point,
- * so the table could have been dropped, and its catalog rows gone, after
- * we acquired the pg_class row.  If pg_class had a TOAST table, this would
- * be a risk; fortunately, it doesn't.
- */
-static AutoVacOpts *
-extract_autovac_opts(HeapTuple tup, TupleDesc pg_class_desc)
-{
-	bytea	   *relopts;
-	AutoVacOpts *av;
-
-	Assert(((Form_pg_class) GETSTRUCT(tup))->relkind == RELKIND_RELATION ||
-		   ((Form_pg_class) GETSTRUCT(tup))->relkind == RELKIND_MATVIEW ||
-		   ((Form_pg_class) GETSTRUCT(tup))->relkind == RELKIND_TOASTVALUE);
-
-	relopts = extractRelOptions(tup, pg_class_desc, NULL);
-	if (relopts == NULL)
-		return NULL;
-
-	av = palloc(sizeof(AutoVacOpts));
-	memcpy(av, &(((StdRdOptions *) relopts)->autovacuum), sizeof(AutoVacOpts));
-	pfree(relopts);
-
-	return av;
-}
-
-
 /*
  * table_recheck_autovac
  *
@@ -2753,8 +2718,8 @@ table_recheck_autovac(Oid relid, HTAB *table_toast_map,
 	bool		doanalyze;
 	autovac_table *tab = NULL;
 	bool		wraparound;
-	AutoVacOpts *avopts;
-	bool		free_avopts = false;
+	StdRdOptions *relopts;
+	bool		free_relopts = false;
 
 	/* fetch the relation's relcache entry */
 	classTup = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(relid));
@@ -2766,9 +2731,9 @@ table_recheck_autovac(Oid relid, HTAB *table_toast_map,
 	 * Get the applicable reloptions.  If it is a TOAST table, try to get the
 	 * main table reloptions if the toast table itself doesn't have.
 	 */
-	avopts = extract_autovac_opts(classTup, pg_class_desc);
-	if (avopts)
-		free_avopts = true;
+	relopts = (StdRdOptions *) extractRelOptions(classTup, pg_class_desc, NULL);
+	if (relopts)
+		free_relopts = true;
 	else if (classForm->relkind == RELKIND_TOASTVALUE &&
 			 table_toast_map != NULL)
 	{
@@ -2777,10 +2742,10 @@ table_recheck_autovac(Oid relid, HTAB *table_toast_map,
 
 		hentry = hash_search(table_toast_map, &relid, HASH_FIND, &found);
 		if (found && hentry->ar_hasrelopts)
-			avopts = &hentry->ar_reloptions;
+			relopts = &hentry->ar_reloptions;
 	}
 
-	recheck_relation_needs_vacanalyze(relid, avopts, classForm,
+	recheck_relation_needs_vacanalyze(relid, relopts, classForm,
 									  effective_multixact_freeze_max_age,
 									  &dovacuum, &doanalyze, &wraparound);
 
@@ -2792,6 +2757,7 @@ table_recheck_autovac(Oid relid, HTAB *table_toast_map,
 		int			multixact_freeze_min_age;
 		int			multixact_freeze_table_age;
 		int			log_min_duration;
+		AutoVacOpts *avopts = (relopts ? &relopts->autovacuum : NULL);
 
 		/*
 		 * Calculate the vacuum cost parameters and the freeze ages.  If there
@@ -2879,8 +2845,8 @@ table_recheck_autovac(Oid relid, HTAB *table_toast_map,
 						 avopts->vacuum_cost_delay >= 0));
 	}
 
-	if (free_avopts)
-		pfree(avopts);
+	if (free_relopts)
+		pfree(relopts);
 	heap_freetuple(classTup);
 	return tab;
 }
@@ -2895,7 +2861,7 @@ table_recheck_autovac(Oid relid, HTAB *table_toast_map,
  */
 static void
 recheck_relation_needs_vacanalyze(Oid relid,
-								  AutoVacOpts *avopts,
+								  StdRdOptions *relopts,
 								  Form_pg_class classForm,
 								  int effective_multixact_freeze_max_age,
 								  bool *dovacuum,
@@ -2908,7 +2874,7 @@ recheck_relation_needs_vacanalyze(Oid relid,
 	tabentry = pgstat_fetch_stat_tabentry_ext(classForm->relisshared,
 											  relid);
 
-	relation_needs_vacanalyze(relid, avopts, classForm, tabentry,
+	relation_needs_vacanalyze(relid, relopts, classForm, tabentry,
 							  effective_multixact_freeze_max_age,
 							  dovacuum, doanalyze, wraparound);
 
@@ -2928,7 +2894,7 @@ recheck_relation_needs_vacanalyze(Oid relid,
  * "dovacuum" and "doanalyze", respectively.  Also return whether the vacuum is
  * being forced because of Xid or multixact wraparound.
  *
- * relopts is a pointer to the AutoVacOpts options (either for itself in the
+ * relopts is a pointer to the StdRdOptions options (either for itself in the
  * case of a plain table, or for either itself or its parent table in the case
  * of a TOAST table), NULL if none; tabentry is the pgstats entry, which can be
  * NULL.
@@ -2962,7 +2928,7 @@ recheck_relation_needs_vacanalyze(Oid relid,
  */
 static void
 relation_needs_vacanalyze(Oid relid,
-						  AutoVacOpts *relopts,
+						  StdRdOptions *relopts,
 						  Form_pg_class classForm,
 						  PgStat_StatTabEntry *tabentry,
 						  int effective_multixact_freeze_max_age,
@@ -2973,6 +2939,7 @@ relation_needs_vacanalyze(Oid relid,
 {
 	bool		force_vacuum;
 	bool		av_enabled;
+	AutoVacOpts *avopts = (relopts ? &relopts->autovacuum : NULL);
 
 	/* constants from reloptions or GUC variables */
 	int			vac_base_thresh,
@@ -3010,45 +2977,45 @@ relation_needs_vacanalyze(Oid relid,
 	 */
 
 	/* -1 in autovac setting means use plain vacuum_scale_factor */
-	vac_scale_factor = (relopts && relopts->vacuum_scale_factor >= 0)
-		? relopts->vacuum_scale_factor
+	vac_scale_factor = (avopts && avopts->vacuum_scale_factor >= 0)
+		? avopts->vacuum_scale_factor
 		: autovacuum_vac_scale;
 
-	vac_base_thresh = (relopts && relopts->vacuum_threshold >= 0)
-		? relopts->vacuum_threshold
+	vac_base_thresh = (avopts && avopts->vacuum_threshold >= 0)
+		? avopts->vacuum_threshold
 		: autovacuum_vac_thresh;
 
 	/* -1 is used to disable max threshold */
-	vac_max_thresh = (relopts && relopts->vacuum_max_threshold >= -1)
-		? relopts->vacuum_max_threshold
+	vac_max_thresh = (avopts && avopts->vacuum_max_threshold >= -1)
+		? avopts->vacuum_max_threshold
 		: autovacuum_vac_max_thresh;
 
-	vac_ins_scale_factor = (relopts && relopts->vacuum_ins_scale_factor >= 0)
-		? relopts->vacuum_ins_scale_factor
+	vac_ins_scale_factor = (avopts && avopts->vacuum_ins_scale_factor >= 0)
+		? avopts->vacuum_ins_scale_factor
 		: autovacuum_vac_ins_scale;
 
 	/* -1 is used to disable insert vacuums */
-	vac_ins_base_thresh = (relopts && relopts->vacuum_ins_threshold >= -1)
-		? relopts->vacuum_ins_threshold
+	vac_ins_base_thresh = (avopts && avopts->vacuum_ins_threshold >= -1)
+		? avopts->vacuum_ins_threshold
 		: autovacuum_vac_ins_thresh;
 
-	anl_scale_factor = (relopts && relopts->analyze_scale_factor >= 0)
-		? relopts->analyze_scale_factor
+	anl_scale_factor = (avopts && avopts->analyze_scale_factor >= 0)
+		? avopts->analyze_scale_factor
 		: autovacuum_anl_scale;
 
-	anl_base_thresh = (relopts && relopts->analyze_threshold >= 0)
-		? relopts->analyze_threshold
+	anl_base_thresh = (avopts && avopts->analyze_threshold >= 0)
+		? avopts->analyze_threshold
 		: autovacuum_anl_thresh;
 
-	freeze_max_age = (relopts && relopts->freeze_max_age >= 0)
-		? Min(relopts->freeze_max_age, autovacuum_freeze_max_age)
+	freeze_max_age = (avopts && avopts->freeze_max_age >= 0)
+		? Min(avopts->freeze_max_age, autovacuum_freeze_max_age)
 		: autovacuum_freeze_max_age;
 
-	multixact_freeze_max_age = (relopts && relopts->multixact_freeze_max_age >= 0)
-		? Min(relopts->multixact_freeze_max_age, effective_multixact_freeze_max_age)
+	multixact_freeze_max_age = (avopts && avopts->multixact_freeze_max_age >= 0)
+		? Min(avopts->multixact_freeze_max_age, effective_multixact_freeze_max_age)
 		: effective_multixact_freeze_max_age;
 
-	av_enabled = (relopts ? relopts->enabled : true);
+	av_enabled = (avopts ? avopts->enabled : true);
 
 	/* Force vacuum if table is at risk of wraparound */
 	xidForceLimit = recentXid - freeze_max_age;
-- 
2.39.5 (Apple Git-154)


--QWBYbd5Ar5k8NvSv
Content-Type: text/plain; charset=us-ascii
Content-Disposition: attachment;
	filename=v1-0002-autovac-resolve-relopts-before-vacuuming.patch



^ permalink  raw  reply  [nested|flat] 223+ messages in thread

* [PATCH 2/2] Publish list of tables being repacked in shared memory
@ 2026-04-07 20:29 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 223+ messages in thread

From: Álvaro Herrera @ 2026-04-07 20:29 UTC (permalink / raw)

Use it in autovacuum to skip processing tables that are being repacked.
This is mostly to avoid repeated attempts to process such tables, which
would fail due to the special deadlock checker behavior for repack.

Author: Álvaro Herrera <[email protected]>
Discussion: https://postgr.es/m/[email protected]
---
 src/backend/commands/repack.c                 | 195 ++++++++++++++++--
 src/backend/postmaster/autovacuum.c           |  20 ++
 .../utils/activity/wait_event_names.txt       |   1 +
 src/include/commands/repack.h                 |   2 +
 src/include/storage/lwlocklist.h              |   2 +-
 src/include/storage/subsystemlist.h           |   1 +
 src/tools/pgindent/typedefs.list              |   3 +
 7 files changed, 210 insertions(+), 14 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index a5f5df77291..ee7072dce6a 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -63,9 +63,11 @@
 #include "optimizer/optimizer.h"
 #include "pgstat.h"
 #include "storage/bufmgr.h"
+#include "storage/ipc.h"
 #include "storage/lmgr.h"
 #include "storage/predicate.h"
 #include "storage/proc.h"
+#include "storage/subsystems.h"
 #include "utils/acl.h"
 #include "utils/fmgroids.h"
 #include "utils/guc.h"
@@ -79,6 +81,32 @@
 #include "utils/syscache.h"
 #include "utils/wait_event_types.h"
 
+
+/* Shared memory layout for REPACK */
+typedef struct RepackWorkerInfo
+{
+	bool		ri_in_use;
+	pid_t		ri_backendpid;
+	Oid			ri_dbid;
+	Oid			ri_relid;
+	Oid			ri_toastrelid;
+} RepackWorkerInfo;
+
+typedef struct
+{
+	bool		re_useless;
+	RepackWorkerInfo re_workerinfo[FLEXIBLE_ARRAY_MEMBER];
+} RepackShmemStruct;
+
+static RepackShmemStruct *RepackShmem;
+
+typedef struct RepackCleanupContext
+{
+	bool		concurrent;
+	int			workerindex;
+} RepackCleanupContext;
+
+
 /*
  * This struct is used to pass around the information on tables to be
  * clustered. We need this so we can make a list of them when invoked without
@@ -90,6 +118,7 @@ typedef struct
 	Oid			indexOid;
 } RelToCluster;
 
+
 /*
  * The first file exported by the decoding worker must contain a snapshot, the
  * following ones contain the data changes.
@@ -166,6 +195,10 @@ static List *get_tables_to_repack_partitioned(RepackCommand cmd,
 											  MemoryContext permcxt);
 static bool repack_is_permitted_for_relation(RepackCommand cmd,
 											 Oid relid, Oid userid);
+static void RepackCleanup(RepackCleanupContext *context);
+static void RepackCleanupCb(int code, Datum arg);
+static void RepackShmemRequest(void *arg);
+static void RepackShmemInit(void *arg);
 
 static void apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt);
 static void apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
@@ -210,6 +243,11 @@ static void ProcessRepackMessage(StringInfo msg);
 static const char *RepackCommandAsString(RepackCommand cmd);
 
 
+const ShmemCallbacks RepackShmemCallbacks = {
+	.request_fn = RepackShmemRequest,
+	.init_fn = RepackShmemInit,
+};
+
 /*
  * The repack code allows for processing multiple tables at once. Because
  * of this, we cannot just run everything on a single transaction, or we
@@ -514,6 +552,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 	Oid			tableOid = RelationGetRelid(OldHeap);
 	Relation	index;
 	LOCKMODE	lmode;
+	RepackCleanupContext context;
 	Oid			save_userid;
 	int			save_sec_context;
 	int			save_nestlevel;
@@ -660,24 +699,43 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 		TransferPredicateLocksToHeapRelation(OldHeap);
 
 	/* rebuild_relation does all the dirty work */
-	PG_TRY();
-	{
-		rebuild_relation(OldHeap, index, verbose, ident_idx);
-	}
-	PG_FINALLY();
+	context.concurrent = concurrent;
+
+	PG_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
 	{
 		if (concurrent)
 		{
-			/*
-			 * Since during normal operation the worker was already asked to
-			 * exit, stopping it explicitly is especially important on ERROR.
-			 * However it still seems a good practice to make sure that the
-			 * worker never survives the REPACK command.
-			 */
-			stop_repack_decoding_worker();
+			bool		freefound = false;
+
+			LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+			for (int i = 0; i < max_repack_replication_slots; i++)
+			{
+				RepackWorkerInfo *worker;
+
+				if (RepackShmem->re_workerinfo[i].ri_in_use)
+					continue;
+
+				freefound = true;
+				worker = &RepackShmem->re_workerinfo[i];
+				context.workerindex = i;
+
+				worker->ri_in_use = true;
+				worker->ri_backendpid = MyProcPid;
+				worker->ri_dbid = MyDatabaseId;
+				worker->ri_relid = RelationGetRelid(OldHeap);
+				worker->ri_toastrelid = OldHeap->rd_rel->reltoastrelid;
+				break;
+			}
+			if (!freefound)
+				elog(ERROR, "could not find free repack entry");
+			LWLockRelease(RepackLock);
 		}
+
+		rebuild_relation(OldHeap, index, verbose, ident_idx);
 	}
-	PG_END_TRY();
+	PG_END_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
+
+	RepackCleanup(&context);
 
 	/* rebuild_relation closes OldHeap, and index if valid */
 
@@ -691,6 +749,117 @@ out:
 	pgstat_progress_end_command();
 }
 
+/*
+ * Return whether any backend is running concurrent REPACK on the given table
+ * (which could be a toast table).
+ */
+bool
+is_table_under_repack(Oid databaseId, Oid relid)
+{
+	bool		retval = false;
+
+	LWLockAcquire(RepackLock, LW_SHARED);
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		RepackWorkerInfo *rworker;
+
+		if (!RepackShmem->re_workerinfo[i].ri_in_use)
+			continue;
+
+		rworker = &RepackShmem->re_workerinfo[i];
+		if (rworker->ri_dbid == MyDatabaseId &&
+			(rworker->ri_relid == relid ||
+			 rworker->ri_toastrelid == relid))
+			retval = true;
+	}
+	LWLockRelease(RepackLock);
+
+	return retval;
+}
+
+/*
+ * Remove ourselves from the workerinfo array.
+ */
+static void
+RepackCleanup(RepackCleanupContext *context)
+{
+	if (context->concurrent)
+	{
+		RepackWorkerInfo *worker;
+
+		/*
+		 * The worker would normally terminate on its own when the work is
+		 * done, but make sure we signal it just in case.
+		 */
+		stop_repack_decoding_worker();
+
+		/*
+		 * also, make sure we stop advertising the relation we were repacking,
+		 * so that autovacuum reverts to handling it normally.
+		 */
+		LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+
+		worker = &RepackShmem->re_workerinfo[context->workerindex];
+		Assert(worker->ri_backendpid == MyProcPid);
+		worker->ri_in_use = false;
+		worker->ri_backendpid = 0;
+		worker->ri_dbid = InvalidOid;
+		worker->ri_relid = InvalidOid;
+		worker->ri_toastrelid = InvalidOid;
+		LWLockRelease(RepackLock);
+	}
+}
+
+/*
+ * RepackCleanup wrapped as an on_shmem_exit callback function
+ */
+static void
+RepackCleanupCb(int code, Datum arg)
+{
+	RepackCleanup((RepackCleanupContext *) DatumGetPointer(arg));
+}
+
+/*
+ * RepackShmemRequest
+ *		Register shared memory space needed for repack
+ */
+static void
+RepackShmemRequest(void *arg)
+{
+	Size		size;
+
+	/*
+	 * Need the fixed struct and the array of RepackWorkerInfo.
+	 */
+	size = sizeof(RepackShmemStruct);
+	size = MAXALIGN(size);
+	size = add_size(size, mul_size(max_repack_replication_slots,
+								   sizeof(RepackWorkerInfo)));
+
+	ShmemRequestStruct(.name = "Repack Data",
+					   .size = size,
+					   .ptr = (void **) &RepackShmem,
+		);
+}
+
+static void
+RepackShmemInit(void *arg)
+{
+	RepackWorkerInfo *reinfo;
+
+	reinfo = (RepackWorkerInfo *) ((char *) RepackShmem +
+								   MAXALIGN(sizeof(RepackShmemStruct)));
+
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		reinfo[i].ri_in_use = false;
+		reinfo[i].ri_backendpid = 0;
+		reinfo[i].ri_dbid = InvalidOid;
+		reinfo[i].ri_relid = InvalidOid;
+		reinfo[i].ri_toastrelid = InvalidOid;
+	}
+}
+
 /*
  * Check if the table (and its index) still meets the requirements of
  * cluster_rel().
diff --git a/src/backend/postmaster/autovacuum.c b/src/backend/postmaster/autovacuum.c
index bd626a16363..080c64ea3c8 100644
--- a/src/backend/postmaster/autovacuum.c
+++ b/src/backend/postmaster/autovacuum.c
@@ -78,6 +78,7 @@
 #include "catalog/namespace.h"
 #include "catalog/pg_database.h"
 #include "catalog/pg_namespace.h"
+#include "commands/repack.h"
 #include "commands/vacuum.h"
 #include "common/int.h"
 #include "funcapi.h"
@@ -2422,6 +2423,25 @@ do_autovacuum(void)
 			}
 		}
 		LWLockRelease(AutovacuumLock);
+
+		/*
+		 * Similarly, if the table is being processed by concurrent repack,
+		 * skip it (but make a note of that).  We wouldn't be able to acquire
+		 * its lock anyway.
+		 */
+		if (!skipit)
+		{
+			MemoryContextSwitchTo(PortalContext);
+
+			skipit = is_table_under_repack(MyDatabaseId, relid);
+			if (skipit)
+				ereport(LOG,
+						errmsg("skipping table \"%s.%s.%s\" because it's being repacked in concurrent mode",
+							   get_database_name(MyDatabaseId),
+							   get_namespace_name(get_rel_namespace(relid)),
+							   get_rel_name(relid)));
+		}
+
 		if (skipit)
 		{
 			LWLockRelease(AutovacuumScheduleLock);
diff --git a/src/backend/utils/activity/wait_event_names.txt b/src/backend/utils/activity/wait_event_names.txt
index 7bda5298558..e206304f204 100644
--- a/src/backend/utils/activity/wait_event_names.txt
+++ b/src/backend/utils/activity/wait_event_names.txt
@@ -332,6 +332,7 @@ SInvalWrite	"Waiting to add a message to the shared catalog invalidation queue."
 WALBufMapping	"Waiting to replace a page in WAL buffers."
 WALWrite	"Waiting for WAL buffers to be written to disk."
 ControlFile	"Waiting to read or update the <filename>pg_control</filename> file or create a new WAL file."
+Repack	"Waiting to read or update tables in process by concurrent repack."
 MultiXactGen	"Waiting to read or update shared multixact state."
 RelCacheInit	"Waiting to read or update a <filename>pg_internal.init</filename> relation cache initialization file."
 CheckpointerComm	"Waiting to manage fsync requests."
diff --git a/src/include/commands/repack.h b/src/include/commands/repack.h
index fd16e74b179..be7d38b5fae 100644
--- a/src/include/commands/repack.h
+++ b/src/include/commands/repack.h
@@ -42,6 +42,8 @@ extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel);
 
 extern void cluster_rel(RepackCommand command, Relation OldHeap, Oid indexOid,
 						ClusterParams *params, bool isTopLevel);
+extern bool is_table_under_repack(Oid databaseId, Oid relid);
+
 extern void check_index_is_clusterable(Relation OldHeap, Oid indexOid,
 									   LOCKMODE lockmode);
 extern void mark_index_clustered(Relation rel, Oid indexOid, bool is_internal);
diff --git a/src/include/storage/lwlocklist.h b/src/include/storage/lwlocklist.h
index af8553bcb6c..3f08f4a15d4 100644
--- a/src/include/storage/lwlocklist.h
+++ b/src/include/storage/lwlocklist.h
@@ -41,7 +41,7 @@ PG_LWLOCK(6, SInvalWrite)
 PG_LWLOCK(7, WALBufMapping)
 PG_LWLOCK(8, WALWrite)
 PG_LWLOCK(9, ControlFile)
-/* 10 was CheckpointLock */
+PG_LWLOCK(10, Repack)
 /* 11 was XactSLRULock */
 /* 12 was SubtransSLRULock */
 PG_LWLOCK(13, MultiXactGen)
diff --git a/src/include/storage/subsystemlist.h b/src/include/storage/subsystemlist.h
index 9ad619080be..4e683b8b0a8 100644
--- a/src/include/storage/subsystemlist.h
+++ b/src/include/storage/subsystemlist.h
@@ -72,6 +72,7 @@ PG_SHMEM_SUBSYSTEM(WalSummarizerShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(PgArchShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(ApplyLauncherShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(SlotSyncShmemCallbacks)
+PG_SHMEM_SUBSYSTEM(RepackShmemCallbacks)
 
 /* other modules that need some shared memory space */
 PG_SHMEM_SUBSYSTEM(BTreeShmemCallbacks)
diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list
index 637c669a146..d019e03aaf1 100644
--- a/src/tools/pgindent/typedefs.list
+++ b/src/tools/pgindent/typedefs.list
@@ -2639,9 +2639,12 @@ ReorderBufferTupleCidEnt
 ReorderBufferTupleCidKey
 ReorderBufferUpdateProgressTxnCB
 ReorderTuple
+RepackCleanupContext
 RepackCommand
 RepackDecodingState
+RepackShmemStruct
 RepackStmt
+RepackWorkerInfo
 ReparameterizeForeignPathByChild_function
 ReplOriginId
 ReplOriginXactState
-- 
2.47.3


--kdrcpfmkbkc4lqhu--





^ permalink  raw  reply  [nested|flat] 223+ messages in thread

* [PATCH 2/2] Publish list of tables being repacked in shared memory
@ 2026-04-07 20:29 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 223+ messages in thread

From: Álvaro Herrera @ 2026-04-07 20:29 UTC (permalink / raw)

Use it in autovacuum to skip processing tables that are being repacked.
This is mostly to avoid repeated attempts to process such tables, which
would fail due to the special deadlock checker behavior for repack.

Author: Álvaro Herrera <[email protected]>
Discussion: https://postgr.es/m/[email protected]
---
 src/backend/commands/repack.c                 | 195 ++++++++++++++++--
 src/backend/postmaster/autovacuum.c           |  20 ++
 .../utils/activity/wait_event_names.txt       |   1 +
 src/include/commands/repack.h                 |   2 +
 src/include/storage/lwlocklist.h              |   2 +-
 src/include/storage/subsystemlist.h           |   1 +
 src/tools/pgindent/typedefs.list              |   3 +
 7 files changed, 210 insertions(+), 14 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index a5f5df77291..ee7072dce6a 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -63,9 +63,11 @@
 #include "optimizer/optimizer.h"
 #include "pgstat.h"
 #include "storage/bufmgr.h"
+#include "storage/ipc.h"
 #include "storage/lmgr.h"
 #include "storage/predicate.h"
 #include "storage/proc.h"
+#include "storage/subsystems.h"
 #include "utils/acl.h"
 #include "utils/fmgroids.h"
 #include "utils/guc.h"
@@ -79,6 +81,32 @@
 #include "utils/syscache.h"
 #include "utils/wait_event_types.h"
 
+
+/* Shared memory layout for REPACK */
+typedef struct RepackWorkerInfo
+{
+	bool		ri_in_use;
+	pid_t		ri_backendpid;
+	Oid			ri_dbid;
+	Oid			ri_relid;
+	Oid			ri_toastrelid;
+} RepackWorkerInfo;
+
+typedef struct
+{
+	bool		re_useless;
+	RepackWorkerInfo re_workerinfo[FLEXIBLE_ARRAY_MEMBER];
+} RepackShmemStruct;
+
+static RepackShmemStruct *RepackShmem;
+
+typedef struct RepackCleanupContext
+{
+	bool		concurrent;
+	int			workerindex;
+} RepackCleanupContext;
+
+
 /*
  * This struct is used to pass around the information on tables to be
  * clustered. We need this so we can make a list of them when invoked without
@@ -90,6 +118,7 @@ typedef struct
 	Oid			indexOid;
 } RelToCluster;
 
+
 /*
  * The first file exported by the decoding worker must contain a snapshot, the
  * following ones contain the data changes.
@@ -166,6 +195,10 @@ static List *get_tables_to_repack_partitioned(RepackCommand cmd,
 											  MemoryContext permcxt);
 static bool repack_is_permitted_for_relation(RepackCommand cmd,
 											 Oid relid, Oid userid);
+static void RepackCleanup(RepackCleanupContext *context);
+static void RepackCleanupCb(int code, Datum arg);
+static void RepackShmemRequest(void *arg);
+static void RepackShmemInit(void *arg);
 
 static void apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt);
 static void apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
@@ -210,6 +243,11 @@ static void ProcessRepackMessage(StringInfo msg);
 static const char *RepackCommandAsString(RepackCommand cmd);
 
 
+const ShmemCallbacks RepackShmemCallbacks = {
+	.request_fn = RepackShmemRequest,
+	.init_fn = RepackShmemInit,
+};
+
 /*
  * The repack code allows for processing multiple tables at once. Because
  * of this, we cannot just run everything on a single transaction, or we
@@ -514,6 +552,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 	Oid			tableOid = RelationGetRelid(OldHeap);
 	Relation	index;
 	LOCKMODE	lmode;
+	RepackCleanupContext context;
 	Oid			save_userid;
 	int			save_sec_context;
 	int			save_nestlevel;
@@ -660,24 +699,43 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 		TransferPredicateLocksToHeapRelation(OldHeap);
 
 	/* rebuild_relation does all the dirty work */
-	PG_TRY();
-	{
-		rebuild_relation(OldHeap, index, verbose, ident_idx);
-	}
-	PG_FINALLY();
+	context.concurrent = concurrent;
+
+	PG_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
 	{
 		if (concurrent)
 		{
-			/*
-			 * Since during normal operation the worker was already asked to
-			 * exit, stopping it explicitly is especially important on ERROR.
-			 * However it still seems a good practice to make sure that the
-			 * worker never survives the REPACK command.
-			 */
-			stop_repack_decoding_worker();
+			bool		freefound = false;
+
+			LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+			for (int i = 0; i < max_repack_replication_slots; i++)
+			{
+				RepackWorkerInfo *worker;
+
+				if (RepackShmem->re_workerinfo[i].ri_in_use)
+					continue;
+
+				freefound = true;
+				worker = &RepackShmem->re_workerinfo[i];
+				context.workerindex = i;
+
+				worker->ri_in_use = true;
+				worker->ri_backendpid = MyProcPid;
+				worker->ri_dbid = MyDatabaseId;
+				worker->ri_relid = RelationGetRelid(OldHeap);
+				worker->ri_toastrelid = OldHeap->rd_rel->reltoastrelid;
+				break;
+			}
+			if (!freefound)
+				elog(ERROR, "could not find free repack entry");
+			LWLockRelease(RepackLock);
 		}
+
+		rebuild_relation(OldHeap, index, verbose, ident_idx);
 	}
-	PG_END_TRY();
+	PG_END_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
+
+	RepackCleanup(&context);
 
 	/* rebuild_relation closes OldHeap, and index if valid */
 
@@ -691,6 +749,117 @@ out:
 	pgstat_progress_end_command();
 }
 
+/*
+ * Return whether any backend is running concurrent REPACK on the given table
+ * (which could be a toast table).
+ */
+bool
+is_table_under_repack(Oid databaseId, Oid relid)
+{
+	bool		retval = false;
+
+	LWLockAcquire(RepackLock, LW_SHARED);
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		RepackWorkerInfo *rworker;
+
+		if (!RepackShmem->re_workerinfo[i].ri_in_use)
+			continue;
+
+		rworker = &RepackShmem->re_workerinfo[i];
+		if (rworker->ri_dbid == MyDatabaseId &&
+			(rworker->ri_relid == relid ||
+			 rworker->ri_toastrelid == relid))
+			retval = true;
+	}
+	LWLockRelease(RepackLock);
+
+	return retval;
+}
+
+/*
+ * Remove ourselves from the workerinfo array.
+ */
+static void
+RepackCleanup(RepackCleanupContext *context)
+{
+	if (context->concurrent)
+	{
+		RepackWorkerInfo *worker;
+
+		/*
+		 * The worker would normally terminate on its own when the work is
+		 * done, but make sure we signal it just in case.
+		 */
+		stop_repack_decoding_worker();
+
+		/*
+		 * also, make sure we stop advertising the relation we were repacking,
+		 * so that autovacuum reverts to handling it normally.
+		 */
+		LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+
+		worker = &RepackShmem->re_workerinfo[context->workerindex];
+		Assert(worker->ri_backendpid == MyProcPid);
+		worker->ri_in_use = false;
+		worker->ri_backendpid = 0;
+		worker->ri_dbid = InvalidOid;
+		worker->ri_relid = InvalidOid;
+		worker->ri_toastrelid = InvalidOid;
+		LWLockRelease(RepackLock);
+	}
+}
+
+/*
+ * RepackCleanup wrapped as an on_shmem_exit callback function
+ */
+static void
+RepackCleanupCb(int code, Datum arg)
+{
+	RepackCleanup((RepackCleanupContext *) DatumGetPointer(arg));
+}
+
+/*
+ * RepackShmemRequest
+ *		Register shared memory space needed for repack
+ */
+static void
+RepackShmemRequest(void *arg)
+{
+	Size		size;
+
+	/*
+	 * Need the fixed struct and the array of RepackWorkerInfo.
+	 */
+	size = sizeof(RepackShmemStruct);
+	size = MAXALIGN(size);
+	size = add_size(size, mul_size(max_repack_replication_slots,
+								   sizeof(RepackWorkerInfo)));
+
+	ShmemRequestStruct(.name = "Repack Data",
+					   .size = size,
+					   .ptr = (void **) &RepackShmem,
+		);
+}
+
+static void
+RepackShmemInit(void *arg)
+{
+	RepackWorkerInfo *reinfo;
+
+	reinfo = (RepackWorkerInfo *) ((char *) RepackShmem +
+								   MAXALIGN(sizeof(RepackShmemStruct)));
+
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		reinfo[i].ri_in_use = false;
+		reinfo[i].ri_backendpid = 0;
+		reinfo[i].ri_dbid = InvalidOid;
+		reinfo[i].ri_relid = InvalidOid;
+		reinfo[i].ri_toastrelid = InvalidOid;
+	}
+}
+
 /*
  * Check if the table (and its index) still meets the requirements of
  * cluster_rel().
diff --git a/src/backend/postmaster/autovacuum.c b/src/backend/postmaster/autovacuum.c
index bd626a16363..080c64ea3c8 100644
--- a/src/backend/postmaster/autovacuum.c
+++ b/src/backend/postmaster/autovacuum.c
@@ -78,6 +78,7 @@
 #include "catalog/namespace.h"
 #include "catalog/pg_database.h"
 #include "catalog/pg_namespace.h"
+#include "commands/repack.h"
 #include "commands/vacuum.h"
 #include "common/int.h"
 #include "funcapi.h"
@@ -2422,6 +2423,25 @@ do_autovacuum(void)
 			}
 		}
 		LWLockRelease(AutovacuumLock);
+
+		/*
+		 * Similarly, if the table is being processed by concurrent repack,
+		 * skip it (but make a note of that).  We wouldn't be able to acquire
+		 * its lock anyway.
+		 */
+		if (!skipit)
+		{
+			MemoryContextSwitchTo(PortalContext);
+
+			skipit = is_table_under_repack(MyDatabaseId, relid);
+			if (skipit)
+				ereport(LOG,
+						errmsg("skipping table \"%s.%s.%s\" because it's being repacked in concurrent mode",
+							   get_database_name(MyDatabaseId),
+							   get_namespace_name(get_rel_namespace(relid)),
+							   get_rel_name(relid)));
+		}
+
 		if (skipit)
 		{
 			LWLockRelease(AutovacuumScheduleLock);
diff --git a/src/backend/utils/activity/wait_event_names.txt b/src/backend/utils/activity/wait_event_names.txt
index 7bda5298558..e206304f204 100644
--- a/src/backend/utils/activity/wait_event_names.txt
+++ b/src/backend/utils/activity/wait_event_names.txt
@@ -332,6 +332,7 @@ SInvalWrite	"Waiting to add a message to the shared catalog invalidation queue."
 WALBufMapping	"Waiting to replace a page in WAL buffers."
 WALWrite	"Waiting for WAL buffers to be written to disk."
 ControlFile	"Waiting to read or update the <filename>pg_control</filename> file or create a new WAL file."
+Repack	"Waiting to read or update tables in process by concurrent repack."
 MultiXactGen	"Waiting to read or update shared multixact state."
 RelCacheInit	"Waiting to read or update a <filename>pg_internal.init</filename> relation cache initialization file."
 CheckpointerComm	"Waiting to manage fsync requests."
diff --git a/src/include/commands/repack.h b/src/include/commands/repack.h
index fd16e74b179..be7d38b5fae 100644
--- a/src/include/commands/repack.h
+++ b/src/include/commands/repack.h
@@ -42,6 +42,8 @@ extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel);
 
 extern void cluster_rel(RepackCommand command, Relation OldHeap, Oid indexOid,
 						ClusterParams *params, bool isTopLevel);
+extern bool is_table_under_repack(Oid databaseId, Oid relid);
+
 extern void check_index_is_clusterable(Relation OldHeap, Oid indexOid,
 									   LOCKMODE lockmode);
 extern void mark_index_clustered(Relation rel, Oid indexOid, bool is_internal);
diff --git a/src/include/storage/lwlocklist.h b/src/include/storage/lwlocklist.h
index af8553bcb6c..3f08f4a15d4 100644
--- a/src/include/storage/lwlocklist.h
+++ b/src/include/storage/lwlocklist.h
@@ -41,7 +41,7 @@ PG_LWLOCK(6, SInvalWrite)
 PG_LWLOCK(7, WALBufMapping)
 PG_LWLOCK(8, WALWrite)
 PG_LWLOCK(9, ControlFile)
-/* 10 was CheckpointLock */
+PG_LWLOCK(10, Repack)
 /* 11 was XactSLRULock */
 /* 12 was SubtransSLRULock */
 PG_LWLOCK(13, MultiXactGen)
diff --git a/src/include/storage/subsystemlist.h b/src/include/storage/subsystemlist.h
index 9ad619080be..4e683b8b0a8 100644
--- a/src/include/storage/subsystemlist.h
+++ b/src/include/storage/subsystemlist.h
@@ -72,6 +72,7 @@ PG_SHMEM_SUBSYSTEM(WalSummarizerShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(PgArchShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(ApplyLauncherShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(SlotSyncShmemCallbacks)
+PG_SHMEM_SUBSYSTEM(RepackShmemCallbacks)
 
 /* other modules that need some shared memory space */
 PG_SHMEM_SUBSYSTEM(BTreeShmemCallbacks)
diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list
index 637c669a146..d019e03aaf1 100644
--- a/src/tools/pgindent/typedefs.list
+++ b/src/tools/pgindent/typedefs.list
@@ -2639,9 +2639,12 @@ ReorderBufferTupleCidEnt
 ReorderBufferTupleCidKey
 ReorderBufferUpdateProgressTxnCB
 ReorderTuple
+RepackCleanupContext
 RepackCommand
 RepackDecodingState
+RepackShmemStruct
 RepackStmt
+RepackWorkerInfo
 ReparameterizeForeignPathByChild_function
 ReplOriginId
 ReplOriginXactState
-- 
2.47.3


--kdrcpfmkbkc4lqhu--





^ permalink  raw  reply  [nested|flat] 223+ messages in thread

* [PATCH 2/2] Publish list of tables being repacked in shared memory
@ 2026-04-07 20:29 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 223+ messages in thread

From: Álvaro Herrera @ 2026-04-07 20:29 UTC (permalink / raw)

Use it in autovacuum to skip processing tables that are being repacked.
This is mostly to avoid repeated attempts to process such tables, which
would fail due to the special deadlock checker behavior for repack.

Author: Álvaro Herrera <[email protected]>
Discussion: https://postgr.es/m/[email protected]
---
 src/backend/commands/repack.c                 | 195 ++++++++++++++++--
 src/backend/postmaster/autovacuum.c           |  20 ++
 .../utils/activity/wait_event_names.txt       |   1 +
 src/include/commands/repack.h                 |   2 +
 src/include/storage/lwlocklist.h              |   2 +-
 src/include/storage/subsystemlist.h           |   1 +
 src/tools/pgindent/typedefs.list              |   3 +
 7 files changed, 210 insertions(+), 14 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index a5f5df77291..ee7072dce6a 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -63,9 +63,11 @@
 #include "optimizer/optimizer.h"
 #include "pgstat.h"
 #include "storage/bufmgr.h"
+#include "storage/ipc.h"
 #include "storage/lmgr.h"
 #include "storage/predicate.h"
 #include "storage/proc.h"
+#include "storage/subsystems.h"
 #include "utils/acl.h"
 #include "utils/fmgroids.h"
 #include "utils/guc.h"
@@ -79,6 +81,32 @@
 #include "utils/syscache.h"
 #include "utils/wait_event_types.h"
 
+
+/* Shared memory layout for REPACK */
+typedef struct RepackWorkerInfo
+{
+	bool		ri_in_use;
+	pid_t		ri_backendpid;
+	Oid			ri_dbid;
+	Oid			ri_relid;
+	Oid			ri_toastrelid;
+} RepackWorkerInfo;
+
+typedef struct
+{
+	bool		re_useless;
+	RepackWorkerInfo re_workerinfo[FLEXIBLE_ARRAY_MEMBER];
+} RepackShmemStruct;
+
+static RepackShmemStruct *RepackShmem;
+
+typedef struct RepackCleanupContext
+{
+	bool		concurrent;
+	int			workerindex;
+} RepackCleanupContext;
+
+
 /*
  * This struct is used to pass around the information on tables to be
  * clustered. We need this so we can make a list of them when invoked without
@@ -90,6 +118,7 @@ typedef struct
 	Oid			indexOid;
 } RelToCluster;
 
+
 /*
  * The first file exported by the decoding worker must contain a snapshot, the
  * following ones contain the data changes.
@@ -166,6 +195,10 @@ static List *get_tables_to_repack_partitioned(RepackCommand cmd,
 											  MemoryContext permcxt);
 static bool repack_is_permitted_for_relation(RepackCommand cmd,
 											 Oid relid, Oid userid);
+static void RepackCleanup(RepackCleanupContext *context);
+static void RepackCleanupCb(int code, Datum arg);
+static void RepackShmemRequest(void *arg);
+static void RepackShmemInit(void *arg);
 
 static void apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt);
 static void apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
@@ -210,6 +243,11 @@ static void ProcessRepackMessage(StringInfo msg);
 static const char *RepackCommandAsString(RepackCommand cmd);
 
 
+const ShmemCallbacks RepackShmemCallbacks = {
+	.request_fn = RepackShmemRequest,
+	.init_fn = RepackShmemInit,
+};
+
 /*
  * The repack code allows for processing multiple tables at once. Because
  * of this, we cannot just run everything on a single transaction, or we
@@ -514,6 +552,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 	Oid			tableOid = RelationGetRelid(OldHeap);
 	Relation	index;
 	LOCKMODE	lmode;
+	RepackCleanupContext context;
 	Oid			save_userid;
 	int			save_sec_context;
 	int			save_nestlevel;
@@ -660,24 +699,43 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 		TransferPredicateLocksToHeapRelation(OldHeap);
 
 	/* rebuild_relation does all the dirty work */
-	PG_TRY();
-	{
-		rebuild_relation(OldHeap, index, verbose, ident_idx);
-	}
-	PG_FINALLY();
+	context.concurrent = concurrent;
+
+	PG_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
 	{
 		if (concurrent)
 		{
-			/*
-			 * Since during normal operation the worker was already asked to
-			 * exit, stopping it explicitly is especially important on ERROR.
-			 * However it still seems a good practice to make sure that the
-			 * worker never survives the REPACK command.
-			 */
-			stop_repack_decoding_worker();
+			bool		freefound = false;
+
+			LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+			for (int i = 0; i < max_repack_replication_slots; i++)
+			{
+				RepackWorkerInfo *worker;
+
+				if (RepackShmem->re_workerinfo[i].ri_in_use)
+					continue;
+
+				freefound = true;
+				worker = &RepackShmem->re_workerinfo[i];
+				context.workerindex = i;
+
+				worker->ri_in_use = true;
+				worker->ri_backendpid = MyProcPid;
+				worker->ri_dbid = MyDatabaseId;
+				worker->ri_relid = RelationGetRelid(OldHeap);
+				worker->ri_toastrelid = OldHeap->rd_rel->reltoastrelid;
+				break;
+			}
+			if (!freefound)
+				elog(ERROR, "could not find free repack entry");
+			LWLockRelease(RepackLock);
 		}
+
+		rebuild_relation(OldHeap, index, verbose, ident_idx);
 	}
-	PG_END_TRY();
+	PG_END_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
+
+	RepackCleanup(&context);
 
 	/* rebuild_relation closes OldHeap, and index if valid */
 
@@ -691,6 +749,117 @@ out:
 	pgstat_progress_end_command();
 }
 
+/*
+ * Return whether any backend is running concurrent REPACK on the given table
+ * (which could be a toast table).
+ */
+bool
+is_table_under_repack(Oid databaseId, Oid relid)
+{
+	bool		retval = false;
+
+	LWLockAcquire(RepackLock, LW_SHARED);
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		RepackWorkerInfo *rworker;
+
+		if (!RepackShmem->re_workerinfo[i].ri_in_use)
+			continue;
+
+		rworker = &RepackShmem->re_workerinfo[i];
+		if (rworker->ri_dbid == MyDatabaseId &&
+			(rworker->ri_relid == relid ||
+			 rworker->ri_toastrelid == relid))
+			retval = true;
+	}
+	LWLockRelease(RepackLock);
+
+	return retval;
+}
+
+/*
+ * Remove ourselves from the workerinfo array.
+ */
+static void
+RepackCleanup(RepackCleanupContext *context)
+{
+	if (context->concurrent)
+	{
+		RepackWorkerInfo *worker;
+
+		/*
+		 * The worker would normally terminate on its own when the work is
+		 * done, but make sure we signal it just in case.
+		 */
+		stop_repack_decoding_worker();
+
+		/*
+		 * also, make sure we stop advertising the relation we were repacking,
+		 * so that autovacuum reverts to handling it normally.
+		 */
+		LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+
+		worker = &RepackShmem->re_workerinfo[context->workerindex];
+		Assert(worker->ri_backendpid == MyProcPid);
+		worker->ri_in_use = false;
+		worker->ri_backendpid = 0;
+		worker->ri_dbid = InvalidOid;
+		worker->ri_relid = InvalidOid;
+		worker->ri_toastrelid = InvalidOid;
+		LWLockRelease(RepackLock);
+	}
+}
+
+/*
+ * RepackCleanup wrapped as an on_shmem_exit callback function
+ */
+static void
+RepackCleanupCb(int code, Datum arg)
+{
+	RepackCleanup((RepackCleanupContext *) DatumGetPointer(arg));
+}
+
+/*
+ * RepackShmemRequest
+ *		Register shared memory space needed for repack
+ */
+static void
+RepackShmemRequest(void *arg)
+{
+	Size		size;
+
+	/*
+	 * Need the fixed struct and the array of RepackWorkerInfo.
+	 */
+	size = sizeof(RepackShmemStruct);
+	size = MAXALIGN(size);
+	size = add_size(size, mul_size(max_repack_replication_slots,
+								   sizeof(RepackWorkerInfo)));
+
+	ShmemRequestStruct(.name = "Repack Data",
+					   .size = size,
+					   .ptr = (void **) &RepackShmem,
+		);
+}
+
+static void
+RepackShmemInit(void *arg)
+{
+	RepackWorkerInfo *reinfo;
+
+	reinfo = (RepackWorkerInfo *) ((char *) RepackShmem +
+								   MAXALIGN(sizeof(RepackShmemStruct)));
+
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		reinfo[i].ri_in_use = false;
+		reinfo[i].ri_backendpid = 0;
+		reinfo[i].ri_dbid = InvalidOid;
+		reinfo[i].ri_relid = InvalidOid;
+		reinfo[i].ri_toastrelid = InvalidOid;
+	}
+}
+
 /*
  * Check if the table (and its index) still meets the requirements of
  * cluster_rel().
diff --git a/src/backend/postmaster/autovacuum.c b/src/backend/postmaster/autovacuum.c
index bd626a16363..080c64ea3c8 100644
--- a/src/backend/postmaster/autovacuum.c
+++ b/src/backend/postmaster/autovacuum.c
@@ -78,6 +78,7 @@
 #include "catalog/namespace.h"
 #include "catalog/pg_database.h"
 #include "catalog/pg_namespace.h"
+#include "commands/repack.h"
 #include "commands/vacuum.h"
 #include "common/int.h"
 #include "funcapi.h"
@@ -2422,6 +2423,25 @@ do_autovacuum(void)
 			}
 		}
 		LWLockRelease(AutovacuumLock);
+
+		/*
+		 * Similarly, if the table is being processed by concurrent repack,
+		 * skip it (but make a note of that).  We wouldn't be able to acquire
+		 * its lock anyway.
+		 */
+		if (!skipit)
+		{
+			MemoryContextSwitchTo(PortalContext);
+
+			skipit = is_table_under_repack(MyDatabaseId, relid);
+			if (skipit)
+				ereport(LOG,
+						errmsg("skipping table \"%s.%s.%s\" because it's being repacked in concurrent mode",
+							   get_database_name(MyDatabaseId),
+							   get_namespace_name(get_rel_namespace(relid)),
+							   get_rel_name(relid)));
+		}
+
 		if (skipit)
 		{
 			LWLockRelease(AutovacuumScheduleLock);
diff --git a/src/backend/utils/activity/wait_event_names.txt b/src/backend/utils/activity/wait_event_names.txt
index 7bda5298558..e206304f204 100644
--- a/src/backend/utils/activity/wait_event_names.txt
+++ b/src/backend/utils/activity/wait_event_names.txt
@@ -332,6 +332,7 @@ SInvalWrite	"Waiting to add a message to the shared catalog invalidation queue."
 WALBufMapping	"Waiting to replace a page in WAL buffers."
 WALWrite	"Waiting for WAL buffers to be written to disk."
 ControlFile	"Waiting to read or update the <filename>pg_control</filename> file or create a new WAL file."
+Repack	"Waiting to read or update tables in process by concurrent repack."
 MultiXactGen	"Waiting to read or update shared multixact state."
 RelCacheInit	"Waiting to read or update a <filename>pg_internal.init</filename> relation cache initialization file."
 CheckpointerComm	"Waiting to manage fsync requests."
diff --git a/src/include/commands/repack.h b/src/include/commands/repack.h
index fd16e74b179..be7d38b5fae 100644
--- a/src/include/commands/repack.h
+++ b/src/include/commands/repack.h
@@ -42,6 +42,8 @@ extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel);
 
 extern void cluster_rel(RepackCommand command, Relation OldHeap, Oid indexOid,
 						ClusterParams *params, bool isTopLevel);
+extern bool is_table_under_repack(Oid databaseId, Oid relid);
+
 extern void check_index_is_clusterable(Relation OldHeap, Oid indexOid,
 									   LOCKMODE lockmode);
 extern void mark_index_clustered(Relation rel, Oid indexOid, bool is_internal);
diff --git a/src/include/storage/lwlocklist.h b/src/include/storage/lwlocklist.h
index af8553bcb6c..3f08f4a15d4 100644
--- a/src/include/storage/lwlocklist.h
+++ b/src/include/storage/lwlocklist.h
@@ -41,7 +41,7 @@ PG_LWLOCK(6, SInvalWrite)
 PG_LWLOCK(7, WALBufMapping)
 PG_LWLOCK(8, WALWrite)
 PG_LWLOCK(9, ControlFile)
-/* 10 was CheckpointLock */
+PG_LWLOCK(10, Repack)
 /* 11 was XactSLRULock */
 /* 12 was SubtransSLRULock */
 PG_LWLOCK(13, MultiXactGen)
diff --git a/src/include/storage/subsystemlist.h b/src/include/storage/subsystemlist.h
index 9ad619080be..4e683b8b0a8 100644
--- a/src/include/storage/subsystemlist.h
+++ b/src/include/storage/subsystemlist.h
@@ -72,6 +72,7 @@ PG_SHMEM_SUBSYSTEM(WalSummarizerShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(PgArchShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(ApplyLauncherShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(SlotSyncShmemCallbacks)
+PG_SHMEM_SUBSYSTEM(RepackShmemCallbacks)
 
 /* other modules that need some shared memory space */
 PG_SHMEM_SUBSYSTEM(BTreeShmemCallbacks)
diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list
index 637c669a146..d019e03aaf1 100644
--- a/src/tools/pgindent/typedefs.list
+++ b/src/tools/pgindent/typedefs.list
@@ -2639,9 +2639,12 @@ ReorderBufferTupleCidEnt
 ReorderBufferTupleCidKey
 ReorderBufferUpdateProgressTxnCB
 ReorderTuple
+RepackCleanupContext
 RepackCommand
 RepackDecodingState
+RepackShmemStruct
 RepackStmt
+RepackWorkerInfo
 ReparameterizeForeignPathByChild_function
 ReplOriginId
 ReplOriginXactState
-- 
2.47.3


--kdrcpfmkbkc4lqhu--





^ permalink  raw  reply  [nested|flat] 223+ messages in thread

* [PATCH 2/2] Publish list of tables being repacked in shared memory
@ 2026-04-07 20:29 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 223+ messages in thread

From: Álvaro Herrera @ 2026-04-07 20:29 UTC (permalink / raw)

Use it in autovacuum to skip processing tables that are being repacked.
This is mostly to avoid repeated attempts to process such tables, which
would fail due to the special deadlock checker behavior for repack.

Author: Álvaro Herrera <[email protected]>
Discussion: https://postgr.es/m/[email protected]
---
 src/backend/commands/repack.c                 | 195 ++++++++++++++++--
 src/backend/postmaster/autovacuum.c           |  20 ++
 .../utils/activity/wait_event_names.txt       |   1 +
 src/include/commands/repack.h                 |   2 +
 src/include/storage/lwlocklist.h              |   2 +-
 src/include/storage/subsystemlist.h           |   1 +
 src/tools/pgindent/typedefs.list              |   3 +
 7 files changed, 210 insertions(+), 14 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index a5f5df77291..ee7072dce6a 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -63,9 +63,11 @@
 #include "optimizer/optimizer.h"
 #include "pgstat.h"
 #include "storage/bufmgr.h"
+#include "storage/ipc.h"
 #include "storage/lmgr.h"
 #include "storage/predicate.h"
 #include "storage/proc.h"
+#include "storage/subsystems.h"
 #include "utils/acl.h"
 #include "utils/fmgroids.h"
 #include "utils/guc.h"
@@ -79,6 +81,32 @@
 #include "utils/syscache.h"
 #include "utils/wait_event_types.h"
 
+
+/* Shared memory layout for REPACK */
+typedef struct RepackWorkerInfo
+{
+	bool		ri_in_use;
+	pid_t		ri_backendpid;
+	Oid			ri_dbid;
+	Oid			ri_relid;
+	Oid			ri_toastrelid;
+} RepackWorkerInfo;
+
+typedef struct
+{
+	bool		re_useless;
+	RepackWorkerInfo re_workerinfo[FLEXIBLE_ARRAY_MEMBER];
+} RepackShmemStruct;
+
+static RepackShmemStruct *RepackShmem;
+
+typedef struct RepackCleanupContext
+{
+	bool		concurrent;
+	int			workerindex;
+} RepackCleanupContext;
+
+
 /*
  * This struct is used to pass around the information on tables to be
  * clustered. We need this so we can make a list of them when invoked without
@@ -90,6 +118,7 @@ typedef struct
 	Oid			indexOid;
 } RelToCluster;
 
+
 /*
  * The first file exported by the decoding worker must contain a snapshot, the
  * following ones contain the data changes.
@@ -166,6 +195,10 @@ static List *get_tables_to_repack_partitioned(RepackCommand cmd,
 											  MemoryContext permcxt);
 static bool repack_is_permitted_for_relation(RepackCommand cmd,
 											 Oid relid, Oid userid);
+static void RepackCleanup(RepackCleanupContext *context);
+static void RepackCleanupCb(int code, Datum arg);
+static void RepackShmemRequest(void *arg);
+static void RepackShmemInit(void *arg);
 
 static void apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt);
 static void apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
@@ -210,6 +243,11 @@ static void ProcessRepackMessage(StringInfo msg);
 static const char *RepackCommandAsString(RepackCommand cmd);
 
 
+const ShmemCallbacks RepackShmemCallbacks = {
+	.request_fn = RepackShmemRequest,
+	.init_fn = RepackShmemInit,
+};
+
 /*
  * The repack code allows for processing multiple tables at once. Because
  * of this, we cannot just run everything on a single transaction, or we
@@ -514,6 +552,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 	Oid			tableOid = RelationGetRelid(OldHeap);
 	Relation	index;
 	LOCKMODE	lmode;
+	RepackCleanupContext context;
 	Oid			save_userid;
 	int			save_sec_context;
 	int			save_nestlevel;
@@ -660,24 +699,43 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 		TransferPredicateLocksToHeapRelation(OldHeap);
 
 	/* rebuild_relation does all the dirty work */
-	PG_TRY();
-	{
-		rebuild_relation(OldHeap, index, verbose, ident_idx);
-	}
-	PG_FINALLY();
+	context.concurrent = concurrent;
+
+	PG_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
 	{
 		if (concurrent)
 		{
-			/*
-			 * Since during normal operation the worker was already asked to
-			 * exit, stopping it explicitly is especially important on ERROR.
-			 * However it still seems a good practice to make sure that the
-			 * worker never survives the REPACK command.
-			 */
-			stop_repack_decoding_worker();
+			bool		freefound = false;
+
+			LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+			for (int i = 0; i < max_repack_replication_slots; i++)
+			{
+				RepackWorkerInfo *worker;
+
+				if (RepackShmem->re_workerinfo[i].ri_in_use)
+					continue;
+
+				freefound = true;
+				worker = &RepackShmem->re_workerinfo[i];
+				context.workerindex = i;
+
+				worker->ri_in_use = true;
+				worker->ri_backendpid = MyProcPid;
+				worker->ri_dbid = MyDatabaseId;
+				worker->ri_relid = RelationGetRelid(OldHeap);
+				worker->ri_toastrelid = OldHeap->rd_rel->reltoastrelid;
+				break;
+			}
+			if (!freefound)
+				elog(ERROR, "could not find free repack entry");
+			LWLockRelease(RepackLock);
 		}
+
+		rebuild_relation(OldHeap, index, verbose, ident_idx);
 	}
-	PG_END_TRY();
+	PG_END_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
+
+	RepackCleanup(&context);
 
 	/* rebuild_relation closes OldHeap, and index if valid */
 
@@ -691,6 +749,117 @@ out:
 	pgstat_progress_end_command();
 }
 
+/*
+ * Return whether any backend is running concurrent REPACK on the given table
+ * (which could be a toast table).
+ */
+bool
+is_table_under_repack(Oid databaseId, Oid relid)
+{
+	bool		retval = false;
+
+	LWLockAcquire(RepackLock, LW_SHARED);
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		RepackWorkerInfo *rworker;
+
+		if (!RepackShmem->re_workerinfo[i].ri_in_use)
+			continue;
+
+		rworker = &RepackShmem->re_workerinfo[i];
+		if (rworker->ri_dbid == MyDatabaseId &&
+			(rworker->ri_relid == relid ||
+			 rworker->ri_toastrelid == relid))
+			retval = true;
+	}
+	LWLockRelease(RepackLock);
+
+	return retval;
+}
+
+/*
+ * Remove ourselves from the workerinfo array.
+ */
+static void
+RepackCleanup(RepackCleanupContext *context)
+{
+	if (context->concurrent)
+	{
+		RepackWorkerInfo *worker;
+
+		/*
+		 * The worker would normally terminate on its own when the work is
+		 * done, but make sure we signal it just in case.
+		 */
+		stop_repack_decoding_worker();
+
+		/*
+		 * also, make sure we stop advertising the relation we were repacking,
+		 * so that autovacuum reverts to handling it normally.
+		 */
+		LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+
+		worker = &RepackShmem->re_workerinfo[context->workerindex];
+		Assert(worker->ri_backendpid == MyProcPid);
+		worker->ri_in_use = false;
+		worker->ri_backendpid = 0;
+		worker->ri_dbid = InvalidOid;
+		worker->ri_relid = InvalidOid;
+		worker->ri_toastrelid = InvalidOid;
+		LWLockRelease(RepackLock);
+	}
+}
+
+/*
+ * RepackCleanup wrapped as an on_shmem_exit callback function
+ */
+static void
+RepackCleanupCb(int code, Datum arg)
+{
+	RepackCleanup((RepackCleanupContext *) DatumGetPointer(arg));
+}
+
+/*
+ * RepackShmemRequest
+ *		Register shared memory space needed for repack
+ */
+static void
+RepackShmemRequest(void *arg)
+{
+	Size		size;
+
+	/*
+	 * Need the fixed struct and the array of RepackWorkerInfo.
+	 */
+	size = sizeof(RepackShmemStruct);
+	size = MAXALIGN(size);
+	size = add_size(size, mul_size(max_repack_replication_slots,
+								   sizeof(RepackWorkerInfo)));
+
+	ShmemRequestStruct(.name = "Repack Data",
+					   .size = size,
+					   .ptr = (void **) &RepackShmem,
+		);
+}
+
+static void
+RepackShmemInit(void *arg)
+{
+	RepackWorkerInfo *reinfo;
+
+	reinfo = (RepackWorkerInfo *) ((char *) RepackShmem +
+								   MAXALIGN(sizeof(RepackShmemStruct)));
+
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		reinfo[i].ri_in_use = false;
+		reinfo[i].ri_backendpid = 0;
+		reinfo[i].ri_dbid = InvalidOid;
+		reinfo[i].ri_relid = InvalidOid;
+		reinfo[i].ri_toastrelid = InvalidOid;
+	}
+}
+
 /*
  * Check if the table (and its index) still meets the requirements of
  * cluster_rel().
diff --git a/src/backend/postmaster/autovacuum.c b/src/backend/postmaster/autovacuum.c
index bd626a16363..080c64ea3c8 100644
--- a/src/backend/postmaster/autovacuum.c
+++ b/src/backend/postmaster/autovacuum.c
@@ -78,6 +78,7 @@
 #include "catalog/namespace.h"
 #include "catalog/pg_database.h"
 #include "catalog/pg_namespace.h"
+#include "commands/repack.h"
 #include "commands/vacuum.h"
 #include "common/int.h"
 #include "funcapi.h"
@@ -2422,6 +2423,25 @@ do_autovacuum(void)
 			}
 		}
 		LWLockRelease(AutovacuumLock);
+
+		/*
+		 * Similarly, if the table is being processed by concurrent repack,
+		 * skip it (but make a note of that).  We wouldn't be able to acquire
+		 * its lock anyway.
+		 */
+		if (!skipit)
+		{
+			MemoryContextSwitchTo(PortalContext);
+
+			skipit = is_table_under_repack(MyDatabaseId, relid);
+			if (skipit)
+				ereport(LOG,
+						errmsg("skipping table \"%s.%s.%s\" because it's being repacked in concurrent mode",
+							   get_database_name(MyDatabaseId),
+							   get_namespace_name(get_rel_namespace(relid)),
+							   get_rel_name(relid)));
+		}
+
 		if (skipit)
 		{
 			LWLockRelease(AutovacuumScheduleLock);
diff --git a/src/backend/utils/activity/wait_event_names.txt b/src/backend/utils/activity/wait_event_names.txt
index 7bda5298558..e206304f204 100644
--- a/src/backend/utils/activity/wait_event_names.txt
+++ b/src/backend/utils/activity/wait_event_names.txt
@@ -332,6 +332,7 @@ SInvalWrite	"Waiting to add a message to the shared catalog invalidation queue."
 WALBufMapping	"Waiting to replace a page in WAL buffers."
 WALWrite	"Waiting for WAL buffers to be written to disk."
 ControlFile	"Waiting to read or update the <filename>pg_control</filename> file or create a new WAL file."
+Repack	"Waiting to read or update tables in process by concurrent repack."
 MultiXactGen	"Waiting to read or update shared multixact state."
 RelCacheInit	"Waiting to read or update a <filename>pg_internal.init</filename> relation cache initialization file."
 CheckpointerComm	"Waiting to manage fsync requests."
diff --git a/src/include/commands/repack.h b/src/include/commands/repack.h
index fd16e74b179..be7d38b5fae 100644
--- a/src/include/commands/repack.h
+++ b/src/include/commands/repack.h
@@ -42,6 +42,8 @@ extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel);
 
 extern void cluster_rel(RepackCommand command, Relation OldHeap, Oid indexOid,
 						ClusterParams *params, bool isTopLevel);
+extern bool is_table_under_repack(Oid databaseId, Oid relid);
+
 extern void check_index_is_clusterable(Relation OldHeap, Oid indexOid,
 									   LOCKMODE lockmode);
 extern void mark_index_clustered(Relation rel, Oid indexOid, bool is_internal);
diff --git a/src/include/storage/lwlocklist.h b/src/include/storage/lwlocklist.h
index af8553bcb6c..3f08f4a15d4 100644
--- a/src/include/storage/lwlocklist.h
+++ b/src/include/storage/lwlocklist.h
@@ -41,7 +41,7 @@ PG_LWLOCK(6, SInvalWrite)
 PG_LWLOCK(7, WALBufMapping)
 PG_LWLOCK(8, WALWrite)
 PG_LWLOCK(9, ControlFile)
-/* 10 was CheckpointLock */
+PG_LWLOCK(10, Repack)
 /* 11 was XactSLRULock */
 /* 12 was SubtransSLRULock */
 PG_LWLOCK(13, MultiXactGen)
diff --git a/src/include/storage/subsystemlist.h b/src/include/storage/subsystemlist.h
index 9ad619080be..4e683b8b0a8 100644
--- a/src/include/storage/subsystemlist.h
+++ b/src/include/storage/subsystemlist.h
@@ -72,6 +72,7 @@ PG_SHMEM_SUBSYSTEM(WalSummarizerShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(PgArchShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(ApplyLauncherShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(SlotSyncShmemCallbacks)
+PG_SHMEM_SUBSYSTEM(RepackShmemCallbacks)
 
 /* other modules that need some shared memory space */
 PG_SHMEM_SUBSYSTEM(BTreeShmemCallbacks)
diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list
index 637c669a146..d019e03aaf1 100644
--- a/src/tools/pgindent/typedefs.list
+++ b/src/tools/pgindent/typedefs.list
@@ -2639,9 +2639,12 @@ ReorderBufferTupleCidEnt
 ReorderBufferTupleCidKey
 ReorderBufferUpdateProgressTxnCB
 ReorderTuple
+RepackCleanupContext
 RepackCommand
 RepackDecodingState
+RepackShmemStruct
 RepackStmt
+RepackWorkerInfo
 ReparameterizeForeignPathByChild_function
 ReplOriginId
 ReplOriginXactState
-- 
2.47.3


--kdrcpfmkbkc4lqhu--





^ permalink  raw  reply  [nested|flat] 223+ messages in thread

* [PATCH 2/2] Publish list of tables being repacked in shared memory
@ 2026-04-07 20:29 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 223+ messages in thread

From: Álvaro Herrera @ 2026-04-07 20:29 UTC (permalink / raw)

Use it in autovacuum to skip processing tables that are being repacked.
This is mostly to avoid repeated attempts to process such tables, which
would fail due to the special deadlock checker behavior for repack.

Author: Álvaro Herrera <[email protected]>
Discussion: https://postgr.es/m/[email protected]
---
 src/backend/commands/repack.c                 | 195 ++++++++++++++++--
 src/backend/postmaster/autovacuum.c           |  20 ++
 .../utils/activity/wait_event_names.txt       |   1 +
 src/include/commands/repack.h                 |   2 +
 src/include/storage/lwlocklist.h              |   2 +-
 src/include/storage/subsystemlist.h           |   1 +
 src/tools/pgindent/typedefs.list              |   3 +
 7 files changed, 210 insertions(+), 14 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index a5f5df77291..ee7072dce6a 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -63,9 +63,11 @@
 #include "optimizer/optimizer.h"
 #include "pgstat.h"
 #include "storage/bufmgr.h"
+#include "storage/ipc.h"
 #include "storage/lmgr.h"
 #include "storage/predicate.h"
 #include "storage/proc.h"
+#include "storage/subsystems.h"
 #include "utils/acl.h"
 #include "utils/fmgroids.h"
 #include "utils/guc.h"
@@ -79,6 +81,32 @@
 #include "utils/syscache.h"
 #include "utils/wait_event_types.h"
 
+
+/* Shared memory layout for REPACK */
+typedef struct RepackWorkerInfo
+{
+	bool		ri_in_use;
+	pid_t		ri_backendpid;
+	Oid			ri_dbid;
+	Oid			ri_relid;
+	Oid			ri_toastrelid;
+} RepackWorkerInfo;
+
+typedef struct
+{
+	bool		re_useless;
+	RepackWorkerInfo re_workerinfo[FLEXIBLE_ARRAY_MEMBER];
+} RepackShmemStruct;
+
+static RepackShmemStruct *RepackShmem;
+
+typedef struct RepackCleanupContext
+{
+	bool		concurrent;
+	int			workerindex;
+} RepackCleanupContext;
+
+
 /*
  * This struct is used to pass around the information on tables to be
  * clustered. We need this so we can make a list of them when invoked without
@@ -90,6 +118,7 @@ typedef struct
 	Oid			indexOid;
 } RelToCluster;
 
+
 /*
  * The first file exported by the decoding worker must contain a snapshot, the
  * following ones contain the data changes.
@@ -166,6 +195,10 @@ static List *get_tables_to_repack_partitioned(RepackCommand cmd,
 											  MemoryContext permcxt);
 static bool repack_is_permitted_for_relation(RepackCommand cmd,
 											 Oid relid, Oid userid);
+static void RepackCleanup(RepackCleanupContext *context);
+static void RepackCleanupCb(int code, Datum arg);
+static void RepackShmemRequest(void *arg);
+static void RepackShmemInit(void *arg);
 
 static void apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt);
 static void apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
@@ -210,6 +243,11 @@ static void ProcessRepackMessage(StringInfo msg);
 static const char *RepackCommandAsString(RepackCommand cmd);
 
 
+const ShmemCallbacks RepackShmemCallbacks = {
+	.request_fn = RepackShmemRequest,
+	.init_fn = RepackShmemInit,
+};
+
 /*
  * The repack code allows for processing multiple tables at once. Because
  * of this, we cannot just run everything on a single transaction, or we
@@ -514,6 +552,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 	Oid			tableOid = RelationGetRelid(OldHeap);
 	Relation	index;
 	LOCKMODE	lmode;
+	RepackCleanupContext context;
 	Oid			save_userid;
 	int			save_sec_context;
 	int			save_nestlevel;
@@ -660,24 +699,43 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 		TransferPredicateLocksToHeapRelation(OldHeap);
 
 	/* rebuild_relation does all the dirty work */
-	PG_TRY();
-	{
-		rebuild_relation(OldHeap, index, verbose, ident_idx);
-	}
-	PG_FINALLY();
+	context.concurrent = concurrent;
+
+	PG_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
 	{
 		if (concurrent)
 		{
-			/*
-			 * Since during normal operation the worker was already asked to
-			 * exit, stopping it explicitly is especially important on ERROR.
-			 * However it still seems a good practice to make sure that the
-			 * worker never survives the REPACK command.
-			 */
-			stop_repack_decoding_worker();
+			bool		freefound = false;
+
+			LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+			for (int i = 0; i < max_repack_replication_slots; i++)
+			{
+				RepackWorkerInfo *worker;
+
+				if (RepackShmem->re_workerinfo[i].ri_in_use)
+					continue;
+
+				freefound = true;
+				worker = &RepackShmem->re_workerinfo[i];
+				context.workerindex = i;
+
+				worker->ri_in_use = true;
+				worker->ri_backendpid = MyProcPid;
+				worker->ri_dbid = MyDatabaseId;
+				worker->ri_relid = RelationGetRelid(OldHeap);
+				worker->ri_toastrelid = OldHeap->rd_rel->reltoastrelid;
+				break;
+			}
+			if (!freefound)
+				elog(ERROR, "could not find free repack entry");
+			LWLockRelease(RepackLock);
 		}
+
+		rebuild_relation(OldHeap, index, verbose, ident_idx);
 	}
-	PG_END_TRY();
+	PG_END_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
+
+	RepackCleanup(&context);
 
 	/* rebuild_relation closes OldHeap, and index if valid */
 
@@ -691,6 +749,117 @@ out:
 	pgstat_progress_end_command();
 }
 
+/*
+ * Return whether any backend is running concurrent REPACK on the given table
+ * (which could be a toast table).
+ */
+bool
+is_table_under_repack(Oid databaseId, Oid relid)
+{
+	bool		retval = false;
+
+	LWLockAcquire(RepackLock, LW_SHARED);
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		RepackWorkerInfo *rworker;
+
+		if (!RepackShmem->re_workerinfo[i].ri_in_use)
+			continue;
+
+		rworker = &RepackShmem->re_workerinfo[i];
+		if (rworker->ri_dbid == MyDatabaseId &&
+			(rworker->ri_relid == relid ||
+			 rworker->ri_toastrelid == relid))
+			retval = true;
+	}
+	LWLockRelease(RepackLock);
+
+	return retval;
+}
+
+/*
+ * Remove ourselves from the workerinfo array.
+ */
+static void
+RepackCleanup(RepackCleanupContext *context)
+{
+	if (context->concurrent)
+	{
+		RepackWorkerInfo *worker;
+
+		/*
+		 * The worker would normally terminate on its own when the work is
+		 * done, but make sure we signal it just in case.
+		 */
+		stop_repack_decoding_worker();
+
+		/*
+		 * also, make sure we stop advertising the relation we were repacking,
+		 * so that autovacuum reverts to handling it normally.
+		 */
+		LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+
+		worker = &RepackShmem->re_workerinfo[context->workerindex];
+		Assert(worker->ri_backendpid == MyProcPid);
+		worker->ri_in_use = false;
+		worker->ri_backendpid = 0;
+		worker->ri_dbid = InvalidOid;
+		worker->ri_relid = InvalidOid;
+		worker->ri_toastrelid = InvalidOid;
+		LWLockRelease(RepackLock);
+	}
+}
+
+/*
+ * RepackCleanup wrapped as an on_shmem_exit callback function
+ */
+static void
+RepackCleanupCb(int code, Datum arg)
+{
+	RepackCleanup((RepackCleanupContext *) DatumGetPointer(arg));
+}
+
+/*
+ * RepackShmemRequest
+ *		Register shared memory space needed for repack
+ */
+static void
+RepackShmemRequest(void *arg)
+{
+	Size		size;
+
+	/*
+	 * Need the fixed struct and the array of RepackWorkerInfo.
+	 */
+	size = sizeof(RepackShmemStruct);
+	size = MAXALIGN(size);
+	size = add_size(size, mul_size(max_repack_replication_slots,
+								   sizeof(RepackWorkerInfo)));
+
+	ShmemRequestStruct(.name = "Repack Data",
+					   .size = size,
+					   .ptr = (void **) &RepackShmem,
+		);
+}
+
+static void
+RepackShmemInit(void *arg)
+{
+	RepackWorkerInfo *reinfo;
+
+	reinfo = (RepackWorkerInfo *) ((char *) RepackShmem +
+								   MAXALIGN(sizeof(RepackShmemStruct)));
+
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		reinfo[i].ri_in_use = false;
+		reinfo[i].ri_backendpid = 0;
+		reinfo[i].ri_dbid = InvalidOid;
+		reinfo[i].ri_relid = InvalidOid;
+		reinfo[i].ri_toastrelid = InvalidOid;
+	}
+}
+
 /*
  * Check if the table (and its index) still meets the requirements of
  * cluster_rel().
diff --git a/src/backend/postmaster/autovacuum.c b/src/backend/postmaster/autovacuum.c
index bd626a16363..080c64ea3c8 100644
--- a/src/backend/postmaster/autovacuum.c
+++ b/src/backend/postmaster/autovacuum.c
@@ -78,6 +78,7 @@
 #include "catalog/namespace.h"
 #include "catalog/pg_database.h"
 #include "catalog/pg_namespace.h"
+#include "commands/repack.h"
 #include "commands/vacuum.h"
 #include "common/int.h"
 #include "funcapi.h"
@@ -2422,6 +2423,25 @@ do_autovacuum(void)
 			}
 		}
 		LWLockRelease(AutovacuumLock);
+
+		/*
+		 * Similarly, if the table is being processed by concurrent repack,
+		 * skip it (but make a note of that).  We wouldn't be able to acquire
+		 * its lock anyway.
+		 */
+		if (!skipit)
+		{
+			MemoryContextSwitchTo(PortalContext);
+
+			skipit = is_table_under_repack(MyDatabaseId, relid);
+			if (skipit)
+				ereport(LOG,
+						errmsg("skipping table \"%s.%s.%s\" because it's being repacked in concurrent mode",
+							   get_database_name(MyDatabaseId),
+							   get_namespace_name(get_rel_namespace(relid)),
+							   get_rel_name(relid)));
+		}
+
 		if (skipit)
 		{
 			LWLockRelease(AutovacuumScheduleLock);
diff --git a/src/backend/utils/activity/wait_event_names.txt b/src/backend/utils/activity/wait_event_names.txt
index 7bda5298558..e206304f204 100644
--- a/src/backend/utils/activity/wait_event_names.txt
+++ b/src/backend/utils/activity/wait_event_names.txt
@@ -332,6 +332,7 @@ SInvalWrite	"Waiting to add a message to the shared catalog invalidation queue."
 WALBufMapping	"Waiting to replace a page in WAL buffers."
 WALWrite	"Waiting for WAL buffers to be written to disk."
 ControlFile	"Waiting to read or update the <filename>pg_control</filename> file or create a new WAL file."
+Repack	"Waiting to read or update tables in process by concurrent repack."
 MultiXactGen	"Waiting to read or update shared multixact state."
 RelCacheInit	"Waiting to read or update a <filename>pg_internal.init</filename> relation cache initialization file."
 CheckpointerComm	"Waiting to manage fsync requests."
diff --git a/src/include/commands/repack.h b/src/include/commands/repack.h
index fd16e74b179..be7d38b5fae 100644
--- a/src/include/commands/repack.h
+++ b/src/include/commands/repack.h
@@ -42,6 +42,8 @@ extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel);
 
 extern void cluster_rel(RepackCommand command, Relation OldHeap, Oid indexOid,
 						ClusterParams *params, bool isTopLevel);
+extern bool is_table_under_repack(Oid databaseId, Oid relid);
+
 extern void check_index_is_clusterable(Relation OldHeap, Oid indexOid,
 									   LOCKMODE lockmode);
 extern void mark_index_clustered(Relation rel, Oid indexOid, bool is_internal);
diff --git a/src/include/storage/lwlocklist.h b/src/include/storage/lwlocklist.h
index af8553bcb6c..3f08f4a15d4 100644
--- a/src/include/storage/lwlocklist.h
+++ b/src/include/storage/lwlocklist.h
@@ -41,7 +41,7 @@ PG_LWLOCK(6, SInvalWrite)
 PG_LWLOCK(7, WALBufMapping)
 PG_LWLOCK(8, WALWrite)
 PG_LWLOCK(9, ControlFile)
-/* 10 was CheckpointLock */
+PG_LWLOCK(10, Repack)
 /* 11 was XactSLRULock */
 /* 12 was SubtransSLRULock */
 PG_LWLOCK(13, MultiXactGen)
diff --git a/src/include/storage/subsystemlist.h b/src/include/storage/subsystemlist.h
index 9ad619080be..4e683b8b0a8 100644
--- a/src/include/storage/subsystemlist.h
+++ b/src/include/storage/subsystemlist.h
@@ -72,6 +72,7 @@ PG_SHMEM_SUBSYSTEM(WalSummarizerShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(PgArchShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(ApplyLauncherShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(SlotSyncShmemCallbacks)
+PG_SHMEM_SUBSYSTEM(RepackShmemCallbacks)
 
 /* other modules that need some shared memory space */
 PG_SHMEM_SUBSYSTEM(BTreeShmemCallbacks)
diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list
index 637c669a146..d019e03aaf1 100644
--- a/src/tools/pgindent/typedefs.list
+++ b/src/tools/pgindent/typedefs.list
@@ -2639,9 +2639,12 @@ ReorderBufferTupleCidEnt
 ReorderBufferTupleCidKey
 ReorderBufferUpdateProgressTxnCB
 ReorderTuple
+RepackCleanupContext
 RepackCommand
 RepackDecodingState
+RepackShmemStruct
 RepackStmt
+RepackWorkerInfo
 ReparameterizeForeignPathByChild_function
 ReplOriginId
 ReplOriginXactState
-- 
2.47.3


--kdrcpfmkbkc4lqhu--





^ permalink  raw  reply  [nested|flat] 223+ messages in thread

* [PATCH 2/2] Publish list of tables being repacked in shared memory
@ 2026-04-07 20:29 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 223+ messages in thread

From: Álvaro Herrera @ 2026-04-07 20:29 UTC (permalink / raw)

Use it in autovacuum to skip processing tables that are being repacked.
This is mostly to avoid repeated attempts to process such tables, which
would fail due to the special deadlock checker behavior for repack.

Author: Álvaro Herrera <[email protected]>
Discussion: https://postgr.es/m/[email protected]
---
 src/backend/commands/repack.c                 | 195 ++++++++++++++++--
 src/backend/postmaster/autovacuum.c           |  20 ++
 .../utils/activity/wait_event_names.txt       |   1 +
 src/include/commands/repack.h                 |   2 +
 src/include/storage/lwlocklist.h              |   2 +-
 src/include/storage/subsystemlist.h           |   1 +
 src/tools/pgindent/typedefs.list              |   3 +
 7 files changed, 210 insertions(+), 14 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index a5f5df77291..ee7072dce6a 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -63,9 +63,11 @@
 #include "optimizer/optimizer.h"
 #include "pgstat.h"
 #include "storage/bufmgr.h"
+#include "storage/ipc.h"
 #include "storage/lmgr.h"
 #include "storage/predicate.h"
 #include "storage/proc.h"
+#include "storage/subsystems.h"
 #include "utils/acl.h"
 #include "utils/fmgroids.h"
 #include "utils/guc.h"
@@ -79,6 +81,32 @@
 #include "utils/syscache.h"
 #include "utils/wait_event_types.h"
 
+
+/* Shared memory layout for REPACK */
+typedef struct RepackWorkerInfo
+{
+	bool		ri_in_use;
+	pid_t		ri_backendpid;
+	Oid			ri_dbid;
+	Oid			ri_relid;
+	Oid			ri_toastrelid;
+} RepackWorkerInfo;
+
+typedef struct
+{
+	bool		re_useless;
+	RepackWorkerInfo re_workerinfo[FLEXIBLE_ARRAY_MEMBER];
+} RepackShmemStruct;
+
+static RepackShmemStruct *RepackShmem;
+
+typedef struct RepackCleanupContext
+{
+	bool		concurrent;
+	int			workerindex;
+} RepackCleanupContext;
+
+
 /*
  * This struct is used to pass around the information on tables to be
  * clustered. We need this so we can make a list of them when invoked without
@@ -90,6 +118,7 @@ typedef struct
 	Oid			indexOid;
 } RelToCluster;
 
+
 /*
  * The first file exported by the decoding worker must contain a snapshot, the
  * following ones contain the data changes.
@@ -166,6 +195,10 @@ static List *get_tables_to_repack_partitioned(RepackCommand cmd,
 											  MemoryContext permcxt);
 static bool repack_is_permitted_for_relation(RepackCommand cmd,
 											 Oid relid, Oid userid);
+static void RepackCleanup(RepackCleanupContext *context);
+static void RepackCleanupCb(int code, Datum arg);
+static void RepackShmemRequest(void *arg);
+static void RepackShmemInit(void *arg);
 
 static void apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt);
 static void apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
@@ -210,6 +243,11 @@ static void ProcessRepackMessage(StringInfo msg);
 static const char *RepackCommandAsString(RepackCommand cmd);
 
 
+const ShmemCallbacks RepackShmemCallbacks = {
+	.request_fn = RepackShmemRequest,
+	.init_fn = RepackShmemInit,
+};
+
 /*
  * The repack code allows for processing multiple tables at once. Because
  * of this, we cannot just run everything on a single transaction, or we
@@ -514,6 +552,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 	Oid			tableOid = RelationGetRelid(OldHeap);
 	Relation	index;
 	LOCKMODE	lmode;
+	RepackCleanupContext context;
 	Oid			save_userid;
 	int			save_sec_context;
 	int			save_nestlevel;
@@ -660,24 +699,43 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 		TransferPredicateLocksToHeapRelation(OldHeap);
 
 	/* rebuild_relation does all the dirty work */
-	PG_TRY();
-	{
-		rebuild_relation(OldHeap, index, verbose, ident_idx);
-	}
-	PG_FINALLY();
+	context.concurrent = concurrent;
+
+	PG_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
 	{
 		if (concurrent)
 		{
-			/*
-			 * Since during normal operation the worker was already asked to
-			 * exit, stopping it explicitly is especially important on ERROR.
-			 * However it still seems a good practice to make sure that the
-			 * worker never survives the REPACK command.
-			 */
-			stop_repack_decoding_worker();
+			bool		freefound = false;
+
+			LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+			for (int i = 0; i < max_repack_replication_slots; i++)
+			{
+				RepackWorkerInfo *worker;
+
+				if (RepackShmem->re_workerinfo[i].ri_in_use)
+					continue;
+
+				freefound = true;
+				worker = &RepackShmem->re_workerinfo[i];
+				context.workerindex = i;
+
+				worker->ri_in_use = true;
+				worker->ri_backendpid = MyProcPid;
+				worker->ri_dbid = MyDatabaseId;
+				worker->ri_relid = RelationGetRelid(OldHeap);
+				worker->ri_toastrelid = OldHeap->rd_rel->reltoastrelid;
+				break;
+			}
+			if (!freefound)
+				elog(ERROR, "could not find free repack entry");
+			LWLockRelease(RepackLock);
 		}
+
+		rebuild_relation(OldHeap, index, verbose, ident_idx);
 	}
-	PG_END_TRY();
+	PG_END_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
+
+	RepackCleanup(&context);
 
 	/* rebuild_relation closes OldHeap, and index if valid */
 
@@ -691,6 +749,117 @@ out:
 	pgstat_progress_end_command();
 }
 
+/*
+ * Return whether any backend is running concurrent REPACK on the given table
+ * (which could be a toast table).
+ */
+bool
+is_table_under_repack(Oid databaseId, Oid relid)
+{
+	bool		retval = false;
+
+	LWLockAcquire(RepackLock, LW_SHARED);
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		RepackWorkerInfo *rworker;
+
+		if (!RepackShmem->re_workerinfo[i].ri_in_use)
+			continue;
+
+		rworker = &RepackShmem->re_workerinfo[i];
+		if (rworker->ri_dbid == MyDatabaseId &&
+			(rworker->ri_relid == relid ||
+			 rworker->ri_toastrelid == relid))
+			retval = true;
+	}
+	LWLockRelease(RepackLock);
+
+	return retval;
+}
+
+/*
+ * Remove ourselves from the workerinfo array.
+ */
+static void
+RepackCleanup(RepackCleanupContext *context)
+{
+	if (context->concurrent)
+	{
+		RepackWorkerInfo *worker;
+
+		/*
+		 * The worker would normally terminate on its own when the work is
+		 * done, but make sure we signal it just in case.
+		 */
+		stop_repack_decoding_worker();
+
+		/*
+		 * also, make sure we stop advertising the relation we were repacking,
+		 * so that autovacuum reverts to handling it normally.
+		 */
+		LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+
+		worker = &RepackShmem->re_workerinfo[context->workerindex];
+		Assert(worker->ri_backendpid == MyProcPid);
+		worker->ri_in_use = false;
+		worker->ri_backendpid = 0;
+		worker->ri_dbid = InvalidOid;
+		worker->ri_relid = InvalidOid;
+		worker->ri_toastrelid = InvalidOid;
+		LWLockRelease(RepackLock);
+	}
+}
+
+/*
+ * RepackCleanup wrapped as an on_shmem_exit callback function
+ */
+static void
+RepackCleanupCb(int code, Datum arg)
+{
+	RepackCleanup((RepackCleanupContext *) DatumGetPointer(arg));
+}
+
+/*
+ * RepackShmemRequest
+ *		Register shared memory space needed for repack
+ */
+static void
+RepackShmemRequest(void *arg)
+{
+	Size		size;
+
+	/*
+	 * Need the fixed struct and the array of RepackWorkerInfo.
+	 */
+	size = sizeof(RepackShmemStruct);
+	size = MAXALIGN(size);
+	size = add_size(size, mul_size(max_repack_replication_slots,
+								   sizeof(RepackWorkerInfo)));
+
+	ShmemRequestStruct(.name = "Repack Data",
+					   .size = size,
+					   .ptr = (void **) &RepackShmem,
+		);
+}
+
+static void
+RepackShmemInit(void *arg)
+{
+	RepackWorkerInfo *reinfo;
+
+	reinfo = (RepackWorkerInfo *) ((char *) RepackShmem +
+								   MAXALIGN(sizeof(RepackShmemStruct)));
+
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		reinfo[i].ri_in_use = false;
+		reinfo[i].ri_backendpid = 0;
+		reinfo[i].ri_dbid = InvalidOid;
+		reinfo[i].ri_relid = InvalidOid;
+		reinfo[i].ri_toastrelid = InvalidOid;
+	}
+}
+
 /*
  * Check if the table (and its index) still meets the requirements of
  * cluster_rel().
diff --git a/src/backend/postmaster/autovacuum.c b/src/backend/postmaster/autovacuum.c
index bd626a16363..080c64ea3c8 100644
--- a/src/backend/postmaster/autovacuum.c
+++ b/src/backend/postmaster/autovacuum.c
@@ -78,6 +78,7 @@
 #include "catalog/namespace.h"
 #include "catalog/pg_database.h"
 #include "catalog/pg_namespace.h"
+#include "commands/repack.h"
 #include "commands/vacuum.h"
 #include "common/int.h"
 #include "funcapi.h"
@@ -2422,6 +2423,25 @@ do_autovacuum(void)
 			}
 		}
 		LWLockRelease(AutovacuumLock);
+
+		/*
+		 * Similarly, if the table is being processed by concurrent repack,
+		 * skip it (but make a note of that).  We wouldn't be able to acquire
+		 * its lock anyway.
+		 */
+		if (!skipit)
+		{
+			MemoryContextSwitchTo(PortalContext);
+
+			skipit = is_table_under_repack(MyDatabaseId, relid);
+			if (skipit)
+				ereport(LOG,
+						errmsg("skipping table \"%s.%s.%s\" because it's being repacked in concurrent mode",
+							   get_database_name(MyDatabaseId),
+							   get_namespace_name(get_rel_namespace(relid)),
+							   get_rel_name(relid)));
+		}
+
 		if (skipit)
 		{
 			LWLockRelease(AutovacuumScheduleLock);
diff --git a/src/backend/utils/activity/wait_event_names.txt b/src/backend/utils/activity/wait_event_names.txt
index 7bda5298558..e206304f204 100644
--- a/src/backend/utils/activity/wait_event_names.txt
+++ b/src/backend/utils/activity/wait_event_names.txt
@@ -332,6 +332,7 @@ SInvalWrite	"Waiting to add a message to the shared catalog invalidation queue."
 WALBufMapping	"Waiting to replace a page in WAL buffers."
 WALWrite	"Waiting for WAL buffers to be written to disk."
 ControlFile	"Waiting to read or update the <filename>pg_control</filename> file or create a new WAL file."
+Repack	"Waiting to read or update tables in process by concurrent repack."
 MultiXactGen	"Waiting to read or update shared multixact state."
 RelCacheInit	"Waiting to read or update a <filename>pg_internal.init</filename> relation cache initialization file."
 CheckpointerComm	"Waiting to manage fsync requests."
diff --git a/src/include/commands/repack.h b/src/include/commands/repack.h
index fd16e74b179..be7d38b5fae 100644
--- a/src/include/commands/repack.h
+++ b/src/include/commands/repack.h
@@ -42,6 +42,8 @@ extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel);
 
 extern void cluster_rel(RepackCommand command, Relation OldHeap, Oid indexOid,
 						ClusterParams *params, bool isTopLevel);
+extern bool is_table_under_repack(Oid databaseId, Oid relid);
+
 extern void check_index_is_clusterable(Relation OldHeap, Oid indexOid,
 									   LOCKMODE lockmode);
 extern void mark_index_clustered(Relation rel, Oid indexOid, bool is_internal);
diff --git a/src/include/storage/lwlocklist.h b/src/include/storage/lwlocklist.h
index af8553bcb6c..3f08f4a15d4 100644
--- a/src/include/storage/lwlocklist.h
+++ b/src/include/storage/lwlocklist.h
@@ -41,7 +41,7 @@ PG_LWLOCK(6, SInvalWrite)
 PG_LWLOCK(7, WALBufMapping)
 PG_LWLOCK(8, WALWrite)
 PG_LWLOCK(9, ControlFile)
-/* 10 was CheckpointLock */
+PG_LWLOCK(10, Repack)
 /* 11 was XactSLRULock */
 /* 12 was SubtransSLRULock */
 PG_LWLOCK(13, MultiXactGen)
diff --git a/src/include/storage/subsystemlist.h b/src/include/storage/subsystemlist.h
index 9ad619080be..4e683b8b0a8 100644
--- a/src/include/storage/subsystemlist.h
+++ b/src/include/storage/subsystemlist.h
@@ -72,6 +72,7 @@ PG_SHMEM_SUBSYSTEM(WalSummarizerShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(PgArchShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(ApplyLauncherShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(SlotSyncShmemCallbacks)
+PG_SHMEM_SUBSYSTEM(RepackShmemCallbacks)
 
 /* other modules that need some shared memory space */
 PG_SHMEM_SUBSYSTEM(BTreeShmemCallbacks)
diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list
index 637c669a146..d019e03aaf1 100644
--- a/src/tools/pgindent/typedefs.list
+++ b/src/tools/pgindent/typedefs.list
@@ -2639,9 +2639,12 @@ ReorderBufferTupleCidEnt
 ReorderBufferTupleCidKey
 ReorderBufferUpdateProgressTxnCB
 ReorderTuple
+RepackCleanupContext
 RepackCommand
 RepackDecodingState
+RepackShmemStruct
 RepackStmt
+RepackWorkerInfo
 ReparameterizeForeignPathByChild_function
 ReplOriginId
 ReplOriginXactState
-- 
2.47.3


--kdrcpfmkbkc4lqhu--





^ permalink  raw  reply  [nested|flat] 223+ messages in thread

* [PATCH 2/2] Publish list of tables being repacked in shared memory
@ 2026-04-07 20:29 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 223+ messages in thread

From: Álvaro Herrera @ 2026-04-07 20:29 UTC (permalink / raw)

Use it in autovacuum to skip processing tables that are being repacked.
This is mostly to avoid repeated attempts to process such tables, which
would fail due to the special deadlock checker behavior for repack.

Author: Álvaro Herrera <[email protected]>
Discussion: https://postgr.es/m/[email protected]
---
 src/backend/commands/repack.c                 | 195 ++++++++++++++++--
 src/backend/postmaster/autovacuum.c           |  20 ++
 .../utils/activity/wait_event_names.txt       |   1 +
 src/include/commands/repack.h                 |   2 +
 src/include/storage/lwlocklist.h              |   2 +-
 src/include/storage/subsystemlist.h           |   1 +
 src/tools/pgindent/typedefs.list              |   3 +
 7 files changed, 210 insertions(+), 14 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index a5f5df77291..ee7072dce6a 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -63,9 +63,11 @@
 #include "optimizer/optimizer.h"
 #include "pgstat.h"
 #include "storage/bufmgr.h"
+#include "storage/ipc.h"
 #include "storage/lmgr.h"
 #include "storage/predicate.h"
 #include "storage/proc.h"
+#include "storage/subsystems.h"
 #include "utils/acl.h"
 #include "utils/fmgroids.h"
 #include "utils/guc.h"
@@ -79,6 +81,32 @@
 #include "utils/syscache.h"
 #include "utils/wait_event_types.h"
 
+
+/* Shared memory layout for REPACK */
+typedef struct RepackWorkerInfo
+{
+	bool		ri_in_use;
+	pid_t		ri_backendpid;
+	Oid			ri_dbid;
+	Oid			ri_relid;
+	Oid			ri_toastrelid;
+} RepackWorkerInfo;
+
+typedef struct
+{
+	bool		re_useless;
+	RepackWorkerInfo re_workerinfo[FLEXIBLE_ARRAY_MEMBER];
+} RepackShmemStruct;
+
+static RepackShmemStruct *RepackShmem;
+
+typedef struct RepackCleanupContext
+{
+	bool		concurrent;
+	int			workerindex;
+} RepackCleanupContext;
+
+
 /*
  * This struct is used to pass around the information on tables to be
  * clustered. We need this so we can make a list of them when invoked without
@@ -90,6 +118,7 @@ typedef struct
 	Oid			indexOid;
 } RelToCluster;
 
+
 /*
  * The first file exported by the decoding worker must contain a snapshot, the
  * following ones contain the data changes.
@@ -166,6 +195,10 @@ static List *get_tables_to_repack_partitioned(RepackCommand cmd,
 											  MemoryContext permcxt);
 static bool repack_is_permitted_for_relation(RepackCommand cmd,
 											 Oid relid, Oid userid);
+static void RepackCleanup(RepackCleanupContext *context);
+static void RepackCleanupCb(int code, Datum arg);
+static void RepackShmemRequest(void *arg);
+static void RepackShmemInit(void *arg);
 
 static void apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt);
 static void apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
@@ -210,6 +243,11 @@ static void ProcessRepackMessage(StringInfo msg);
 static const char *RepackCommandAsString(RepackCommand cmd);
 
 
+const ShmemCallbacks RepackShmemCallbacks = {
+	.request_fn = RepackShmemRequest,
+	.init_fn = RepackShmemInit,
+};
+
 /*
  * The repack code allows for processing multiple tables at once. Because
  * of this, we cannot just run everything on a single transaction, or we
@@ -514,6 +552,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 	Oid			tableOid = RelationGetRelid(OldHeap);
 	Relation	index;
 	LOCKMODE	lmode;
+	RepackCleanupContext context;
 	Oid			save_userid;
 	int			save_sec_context;
 	int			save_nestlevel;
@@ -660,24 +699,43 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 		TransferPredicateLocksToHeapRelation(OldHeap);
 
 	/* rebuild_relation does all the dirty work */
-	PG_TRY();
-	{
-		rebuild_relation(OldHeap, index, verbose, ident_idx);
-	}
-	PG_FINALLY();
+	context.concurrent = concurrent;
+
+	PG_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
 	{
 		if (concurrent)
 		{
-			/*
-			 * Since during normal operation the worker was already asked to
-			 * exit, stopping it explicitly is especially important on ERROR.
-			 * However it still seems a good practice to make sure that the
-			 * worker never survives the REPACK command.
-			 */
-			stop_repack_decoding_worker();
+			bool		freefound = false;
+
+			LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+			for (int i = 0; i < max_repack_replication_slots; i++)
+			{
+				RepackWorkerInfo *worker;
+
+				if (RepackShmem->re_workerinfo[i].ri_in_use)
+					continue;
+
+				freefound = true;
+				worker = &RepackShmem->re_workerinfo[i];
+				context.workerindex = i;
+
+				worker->ri_in_use = true;
+				worker->ri_backendpid = MyProcPid;
+				worker->ri_dbid = MyDatabaseId;
+				worker->ri_relid = RelationGetRelid(OldHeap);
+				worker->ri_toastrelid = OldHeap->rd_rel->reltoastrelid;
+				break;
+			}
+			if (!freefound)
+				elog(ERROR, "could not find free repack entry");
+			LWLockRelease(RepackLock);
 		}
+
+		rebuild_relation(OldHeap, index, verbose, ident_idx);
 	}
-	PG_END_TRY();
+	PG_END_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
+
+	RepackCleanup(&context);
 
 	/* rebuild_relation closes OldHeap, and index if valid */
 
@@ -691,6 +749,117 @@ out:
 	pgstat_progress_end_command();
 }
 
+/*
+ * Return whether any backend is running concurrent REPACK on the given table
+ * (which could be a toast table).
+ */
+bool
+is_table_under_repack(Oid databaseId, Oid relid)
+{
+	bool		retval = false;
+
+	LWLockAcquire(RepackLock, LW_SHARED);
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		RepackWorkerInfo *rworker;
+
+		if (!RepackShmem->re_workerinfo[i].ri_in_use)
+			continue;
+
+		rworker = &RepackShmem->re_workerinfo[i];
+		if (rworker->ri_dbid == MyDatabaseId &&
+			(rworker->ri_relid == relid ||
+			 rworker->ri_toastrelid == relid))
+			retval = true;
+	}
+	LWLockRelease(RepackLock);
+
+	return retval;
+}
+
+/*
+ * Remove ourselves from the workerinfo array.
+ */
+static void
+RepackCleanup(RepackCleanupContext *context)
+{
+	if (context->concurrent)
+	{
+		RepackWorkerInfo *worker;
+
+		/*
+		 * The worker would normally terminate on its own when the work is
+		 * done, but make sure we signal it just in case.
+		 */
+		stop_repack_decoding_worker();
+
+		/*
+		 * also, make sure we stop advertising the relation we were repacking,
+		 * so that autovacuum reverts to handling it normally.
+		 */
+		LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+
+		worker = &RepackShmem->re_workerinfo[context->workerindex];
+		Assert(worker->ri_backendpid == MyProcPid);
+		worker->ri_in_use = false;
+		worker->ri_backendpid = 0;
+		worker->ri_dbid = InvalidOid;
+		worker->ri_relid = InvalidOid;
+		worker->ri_toastrelid = InvalidOid;
+		LWLockRelease(RepackLock);
+	}
+}
+
+/*
+ * RepackCleanup wrapped as an on_shmem_exit callback function
+ */
+static void
+RepackCleanupCb(int code, Datum arg)
+{
+	RepackCleanup((RepackCleanupContext *) DatumGetPointer(arg));
+}
+
+/*
+ * RepackShmemRequest
+ *		Register shared memory space needed for repack
+ */
+static void
+RepackShmemRequest(void *arg)
+{
+	Size		size;
+
+	/*
+	 * Need the fixed struct and the array of RepackWorkerInfo.
+	 */
+	size = sizeof(RepackShmemStruct);
+	size = MAXALIGN(size);
+	size = add_size(size, mul_size(max_repack_replication_slots,
+								   sizeof(RepackWorkerInfo)));
+
+	ShmemRequestStruct(.name = "Repack Data",
+					   .size = size,
+					   .ptr = (void **) &RepackShmem,
+		);
+}
+
+static void
+RepackShmemInit(void *arg)
+{
+	RepackWorkerInfo *reinfo;
+
+	reinfo = (RepackWorkerInfo *) ((char *) RepackShmem +
+								   MAXALIGN(sizeof(RepackShmemStruct)));
+
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		reinfo[i].ri_in_use = false;
+		reinfo[i].ri_backendpid = 0;
+		reinfo[i].ri_dbid = InvalidOid;
+		reinfo[i].ri_relid = InvalidOid;
+		reinfo[i].ri_toastrelid = InvalidOid;
+	}
+}
+
 /*
  * Check if the table (and its index) still meets the requirements of
  * cluster_rel().
diff --git a/src/backend/postmaster/autovacuum.c b/src/backend/postmaster/autovacuum.c
index bd626a16363..080c64ea3c8 100644
--- a/src/backend/postmaster/autovacuum.c
+++ b/src/backend/postmaster/autovacuum.c
@@ -78,6 +78,7 @@
 #include "catalog/namespace.h"
 #include "catalog/pg_database.h"
 #include "catalog/pg_namespace.h"
+#include "commands/repack.h"
 #include "commands/vacuum.h"
 #include "common/int.h"
 #include "funcapi.h"
@@ -2422,6 +2423,25 @@ do_autovacuum(void)
 			}
 		}
 		LWLockRelease(AutovacuumLock);
+
+		/*
+		 * Similarly, if the table is being processed by concurrent repack,
+		 * skip it (but make a note of that).  We wouldn't be able to acquire
+		 * its lock anyway.
+		 */
+		if (!skipit)
+		{
+			MemoryContextSwitchTo(PortalContext);
+
+			skipit = is_table_under_repack(MyDatabaseId, relid);
+			if (skipit)
+				ereport(LOG,
+						errmsg("skipping table \"%s.%s.%s\" because it's being repacked in concurrent mode",
+							   get_database_name(MyDatabaseId),
+							   get_namespace_name(get_rel_namespace(relid)),
+							   get_rel_name(relid)));
+		}
+
 		if (skipit)
 		{
 			LWLockRelease(AutovacuumScheduleLock);
diff --git a/src/backend/utils/activity/wait_event_names.txt b/src/backend/utils/activity/wait_event_names.txt
index 7bda5298558..e206304f204 100644
--- a/src/backend/utils/activity/wait_event_names.txt
+++ b/src/backend/utils/activity/wait_event_names.txt
@@ -332,6 +332,7 @@ SInvalWrite	"Waiting to add a message to the shared catalog invalidation queue."
 WALBufMapping	"Waiting to replace a page in WAL buffers."
 WALWrite	"Waiting for WAL buffers to be written to disk."
 ControlFile	"Waiting to read or update the <filename>pg_control</filename> file or create a new WAL file."
+Repack	"Waiting to read or update tables in process by concurrent repack."
 MultiXactGen	"Waiting to read or update shared multixact state."
 RelCacheInit	"Waiting to read or update a <filename>pg_internal.init</filename> relation cache initialization file."
 CheckpointerComm	"Waiting to manage fsync requests."
diff --git a/src/include/commands/repack.h b/src/include/commands/repack.h
index fd16e74b179..be7d38b5fae 100644
--- a/src/include/commands/repack.h
+++ b/src/include/commands/repack.h
@@ -42,6 +42,8 @@ extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel);
 
 extern void cluster_rel(RepackCommand command, Relation OldHeap, Oid indexOid,
 						ClusterParams *params, bool isTopLevel);
+extern bool is_table_under_repack(Oid databaseId, Oid relid);
+
 extern void check_index_is_clusterable(Relation OldHeap, Oid indexOid,
 									   LOCKMODE lockmode);
 extern void mark_index_clustered(Relation rel, Oid indexOid, bool is_internal);
diff --git a/src/include/storage/lwlocklist.h b/src/include/storage/lwlocklist.h
index af8553bcb6c..3f08f4a15d4 100644
--- a/src/include/storage/lwlocklist.h
+++ b/src/include/storage/lwlocklist.h
@@ -41,7 +41,7 @@ PG_LWLOCK(6, SInvalWrite)
 PG_LWLOCK(7, WALBufMapping)
 PG_LWLOCK(8, WALWrite)
 PG_LWLOCK(9, ControlFile)
-/* 10 was CheckpointLock */
+PG_LWLOCK(10, Repack)
 /* 11 was XactSLRULock */
 /* 12 was SubtransSLRULock */
 PG_LWLOCK(13, MultiXactGen)
diff --git a/src/include/storage/subsystemlist.h b/src/include/storage/subsystemlist.h
index 9ad619080be..4e683b8b0a8 100644
--- a/src/include/storage/subsystemlist.h
+++ b/src/include/storage/subsystemlist.h
@@ -72,6 +72,7 @@ PG_SHMEM_SUBSYSTEM(WalSummarizerShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(PgArchShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(ApplyLauncherShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(SlotSyncShmemCallbacks)
+PG_SHMEM_SUBSYSTEM(RepackShmemCallbacks)
 
 /* other modules that need some shared memory space */
 PG_SHMEM_SUBSYSTEM(BTreeShmemCallbacks)
diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list
index 637c669a146..d019e03aaf1 100644
--- a/src/tools/pgindent/typedefs.list
+++ b/src/tools/pgindent/typedefs.list
@@ -2639,9 +2639,12 @@ ReorderBufferTupleCidEnt
 ReorderBufferTupleCidKey
 ReorderBufferUpdateProgressTxnCB
 ReorderTuple
+RepackCleanupContext
 RepackCommand
 RepackDecodingState
+RepackShmemStruct
 RepackStmt
+RepackWorkerInfo
 ReparameterizeForeignPathByChild_function
 ReplOriginId
 ReplOriginXactState
-- 
2.47.3


--kdrcpfmkbkc4lqhu--





^ permalink  raw  reply  [nested|flat] 223+ messages in thread

* [PATCH 2/2] Publish list of tables being repacked in shared memory
@ 2026-04-07 20:29 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 223+ messages in thread

From: Álvaro Herrera @ 2026-04-07 20:29 UTC (permalink / raw)

Use it in autovacuum to skip processing tables that are being repacked.
This is mostly to avoid repeated attempts to process such tables, which
would fail due to the special deadlock checker behavior for repack.

Author: Álvaro Herrera <[email protected]>
Discussion: https://postgr.es/m/[email protected]
---
 src/backend/commands/repack.c                 | 195 ++++++++++++++++--
 src/backend/postmaster/autovacuum.c           |  20 ++
 .../utils/activity/wait_event_names.txt       |   1 +
 src/include/commands/repack.h                 |   2 +
 src/include/storage/lwlocklist.h              |   2 +-
 src/include/storage/subsystemlist.h           |   1 +
 src/tools/pgindent/typedefs.list              |   3 +
 7 files changed, 210 insertions(+), 14 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index a5f5df77291..ee7072dce6a 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -63,9 +63,11 @@
 #include "optimizer/optimizer.h"
 #include "pgstat.h"
 #include "storage/bufmgr.h"
+#include "storage/ipc.h"
 #include "storage/lmgr.h"
 #include "storage/predicate.h"
 #include "storage/proc.h"
+#include "storage/subsystems.h"
 #include "utils/acl.h"
 #include "utils/fmgroids.h"
 #include "utils/guc.h"
@@ -79,6 +81,32 @@
 #include "utils/syscache.h"
 #include "utils/wait_event_types.h"
 
+
+/* Shared memory layout for REPACK */
+typedef struct RepackWorkerInfo
+{
+	bool		ri_in_use;
+	pid_t		ri_backendpid;
+	Oid			ri_dbid;
+	Oid			ri_relid;
+	Oid			ri_toastrelid;
+} RepackWorkerInfo;
+
+typedef struct
+{
+	bool		re_useless;
+	RepackWorkerInfo re_workerinfo[FLEXIBLE_ARRAY_MEMBER];
+} RepackShmemStruct;
+
+static RepackShmemStruct *RepackShmem;
+
+typedef struct RepackCleanupContext
+{
+	bool		concurrent;
+	int			workerindex;
+} RepackCleanupContext;
+
+
 /*
  * This struct is used to pass around the information on tables to be
  * clustered. We need this so we can make a list of them when invoked without
@@ -90,6 +118,7 @@ typedef struct
 	Oid			indexOid;
 } RelToCluster;
 
+
 /*
  * The first file exported by the decoding worker must contain a snapshot, the
  * following ones contain the data changes.
@@ -166,6 +195,10 @@ static List *get_tables_to_repack_partitioned(RepackCommand cmd,
 											  MemoryContext permcxt);
 static bool repack_is_permitted_for_relation(RepackCommand cmd,
 											 Oid relid, Oid userid);
+static void RepackCleanup(RepackCleanupContext *context);
+static void RepackCleanupCb(int code, Datum arg);
+static void RepackShmemRequest(void *arg);
+static void RepackShmemInit(void *arg);
 
 static void apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt);
 static void apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
@@ -210,6 +243,11 @@ static void ProcessRepackMessage(StringInfo msg);
 static const char *RepackCommandAsString(RepackCommand cmd);
 
 
+const ShmemCallbacks RepackShmemCallbacks = {
+	.request_fn = RepackShmemRequest,
+	.init_fn = RepackShmemInit,
+};
+
 /*
  * The repack code allows for processing multiple tables at once. Because
  * of this, we cannot just run everything on a single transaction, or we
@@ -514,6 +552,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 	Oid			tableOid = RelationGetRelid(OldHeap);
 	Relation	index;
 	LOCKMODE	lmode;
+	RepackCleanupContext context;
 	Oid			save_userid;
 	int			save_sec_context;
 	int			save_nestlevel;
@@ -660,24 +699,43 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 		TransferPredicateLocksToHeapRelation(OldHeap);
 
 	/* rebuild_relation does all the dirty work */
-	PG_TRY();
-	{
-		rebuild_relation(OldHeap, index, verbose, ident_idx);
-	}
-	PG_FINALLY();
+	context.concurrent = concurrent;
+
+	PG_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
 	{
 		if (concurrent)
 		{
-			/*
-			 * Since during normal operation the worker was already asked to
-			 * exit, stopping it explicitly is especially important on ERROR.
-			 * However it still seems a good practice to make sure that the
-			 * worker never survives the REPACK command.
-			 */
-			stop_repack_decoding_worker();
+			bool		freefound = false;
+
+			LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+			for (int i = 0; i < max_repack_replication_slots; i++)
+			{
+				RepackWorkerInfo *worker;
+
+				if (RepackShmem->re_workerinfo[i].ri_in_use)
+					continue;
+
+				freefound = true;
+				worker = &RepackShmem->re_workerinfo[i];
+				context.workerindex = i;
+
+				worker->ri_in_use = true;
+				worker->ri_backendpid = MyProcPid;
+				worker->ri_dbid = MyDatabaseId;
+				worker->ri_relid = RelationGetRelid(OldHeap);
+				worker->ri_toastrelid = OldHeap->rd_rel->reltoastrelid;
+				break;
+			}
+			if (!freefound)
+				elog(ERROR, "could not find free repack entry");
+			LWLockRelease(RepackLock);
 		}
+
+		rebuild_relation(OldHeap, index, verbose, ident_idx);
 	}
-	PG_END_TRY();
+	PG_END_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
+
+	RepackCleanup(&context);
 
 	/* rebuild_relation closes OldHeap, and index if valid */
 
@@ -691,6 +749,117 @@ out:
 	pgstat_progress_end_command();
 }
 
+/*
+ * Return whether any backend is running concurrent REPACK on the given table
+ * (which could be a toast table).
+ */
+bool
+is_table_under_repack(Oid databaseId, Oid relid)
+{
+	bool		retval = false;
+
+	LWLockAcquire(RepackLock, LW_SHARED);
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		RepackWorkerInfo *rworker;
+
+		if (!RepackShmem->re_workerinfo[i].ri_in_use)
+			continue;
+
+		rworker = &RepackShmem->re_workerinfo[i];
+		if (rworker->ri_dbid == MyDatabaseId &&
+			(rworker->ri_relid == relid ||
+			 rworker->ri_toastrelid == relid))
+			retval = true;
+	}
+	LWLockRelease(RepackLock);
+
+	return retval;
+}
+
+/*
+ * Remove ourselves from the workerinfo array.
+ */
+static void
+RepackCleanup(RepackCleanupContext *context)
+{
+	if (context->concurrent)
+	{
+		RepackWorkerInfo *worker;
+
+		/*
+		 * The worker would normally terminate on its own when the work is
+		 * done, but make sure we signal it just in case.
+		 */
+		stop_repack_decoding_worker();
+
+		/*
+		 * also, make sure we stop advertising the relation we were repacking,
+		 * so that autovacuum reverts to handling it normally.
+		 */
+		LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+
+		worker = &RepackShmem->re_workerinfo[context->workerindex];
+		Assert(worker->ri_backendpid == MyProcPid);
+		worker->ri_in_use = false;
+		worker->ri_backendpid = 0;
+		worker->ri_dbid = InvalidOid;
+		worker->ri_relid = InvalidOid;
+		worker->ri_toastrelid = InvalidOid;
+		LWLockRelease(RepackLock);
+	}
+}
+
+/*
+ * RepackCleanup wrapped as an on_shmem_exit callback function
+ */
+static void
+RepackCleanupCb(int code, Datum arg)
+{
+	RepackCleanup((RepackCleanupContext *) DatumGetPointer(arg));
+}
+
+/*
+ * RepackShmemRequest
+ *		Register shared memory space needed for repack
+ */
+static void
+RepackShmemRequest(void *arg)
+{
+	Size		size;
+
+	/*
+	 * Need the fixed struct and the array of RepackWorkerInfo.
+	 */
+	size = sizeof(RepackShmemStruct);
+	size = MAXALIGN(size);
+	size = add_size(size, mul_size(max_repack_replication_slots,
+								   sizeof(RepackWorkerInfo)));
+
+	ShmemRequestStruct(.name = "Repack Data",
+					   .size = size,
+					   .ptr = (void **) &RepackShmem,
+		);
+}
+
+static void
+RepackShmemInit(void *arg)
+{
+	RepackWorkerInfo *reinfo;
+
+	reinfo = (RepackWorkerInfo *) ((char *) RepackShmem +
+								   MAXALIGN(sizeof(RepackShmemStruct)));
+
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		reinfo[i].ri_in_use = false;
+		reinfo[i].ri_backendpid = 0;
+		reinfo[i].ri_dbid = InvalidOid;
+		reinfo[i].ri_relid = InvalidOid;
+		reinfo[i].ri_toastrelid = InvalidOid;
+	}
+}
+
 /*
  * Check if the table (and its index) still meets the requirements of
  * cluster_rel().
diff --git a/src/backend/postmaster/autovacuum.c b/src/backend/postmaster/autovacuum.c
index bd626a16363..080c64ea3c8 100644
--- a/src/backend/postmaster/autovacuum.c
+++ b/src/backend/postmaster/autovacuum.c
@@ -78,6 +78,7 @@
 #include "catalog/namespace.h"
 #include "catalog/pg_database.h"
 #include "catalog/pg_namespace.h"
+#include "commands/repack.h"
 #include "commands/vacuum.h"
 #include "common/int.h"
 #include "funcapi.h"
@@ -2422,6 +2423,25 @@ do_autovacuum(void)
 			}
 		}
 		LWLockRelease(AutovacuumLock);
+
+		/*
+		 * Similarly, if the table is being processed by concurrent repack,
+		 * skip it (but make a note of that).  We wouldn't be able to acquire
+		 * its lock anyway.
+		 */
+		if (!skipit)
+		{
+			MemoryContextSwitchTo(PortalContext);
+
+			skipit = is_table_under_repack(MyDatabaseId, relid);
+			if (skipit)
+				ereport(LOG,
+						errmsg("skipping table \"%s.%s.%s\" because it's being repacked in concurrent mode",
+							   get_database_name(MyDatabaseId),
+							   get_namespace_name(get_rel_namespace(relid)),
+							   get_rel_name(relid)));
+		}
+
 		if (skipit)
 		{
 			LWLockRelease(AutovacuumScheduleLock);
diff --git a/src/backend/utils/activity/wait_event_names.txt b/src/backend/utils/activity/wait_event_names.txt
index 7bda5298558..e206304f204 100644
--- a/src/backend/utils/activity/wait_event_names.txt
+++ b/src/backend/utils/activity/wait_event_names.txt
@@ -332,6 +332,7 @@ SInvalWrite	"Waiting to add a message to the shared catalog invalidation queue."
 WALBufMapping	"Waiting to replace a page in WAL buffers."
 WALWrite	"Waiting for WAL buffers to be written to disk."
 ControlFile	"Waiting to read or update the <filename>pg_control</filename> file or create a new WAL file."
+Repack	"Waiting to read or update tables in process by concurrent repack."
 MultiXactGen	"Waiting to read or update shared multixact state."
 RelCacheInit	"Waiting to read or update a <filename>pg_internal.init</filename> relation cache initialization file."
 CheckpointerComm	"Waiting to manage fsync requests."
diff --git a/src/include/commands/repack.h b/src/include/commands/repack.h
index fd16e74b179..be7d38b5fae 100644
--- a/src/include/commands/repack.h
+++ b/src/include/commands/repack.h
@@ -42,6 +42,8 @@ extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel);
 
 extern void cluster_rel(RepackCommand command, Relation OldHeap, Oid indexOid,
 						ClusterParams *params, bool isTopLevel);
+extern bool is_table_under_repack(Oid databaseId, Oid relid);
+
 extern void check_index_is_clusterable(Relation OldHeap, Oid indexOid,
 									   LOCKMODE lockmode);
 extern void mark_index_clustered(Relation rel, Oid indexOid, bool is_internal);
diff --git a/src/include/storage/lwlocklist.h b/src/include/storage/lwlocklist.h
index af8553bcb6c..3f08f4a15d4 100644
--- a/src/include/storage/lwlocklist.h
+++ b/src/include/storage/lwlocklist.h
@@ -41,7 +41,7 @@ PG_LWLOCK(6, SInvalWrite)
 PG_LWLOCK(7, WALBufMapping)
 PG_LWLOCK(8, WALWrite)
 PG_LWLOCK(9, ControlFile)
-/* 10 was CheckpointLock */
+PG_LWLOCK(10, Repack)
 /* 11 was XactSLRULock */
 /* 12 was SubtransSLRULock */
 PG_LWLOCK(13, MultiXactGen)
diff --git a/src/include/storage/subsystemlist.h b/src/include/storage/subsystemlist.h
index 9ad619080be..4e683b8b0a8 100644
--- a/src/include/storage/subsystemlist.h
+++ b/src/include/storage/subsystemlist.h
@@ -72,6 +72,7 @@ PG_SHMEM_SUBSYSTEM(WalSummarizerShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(PgArchShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(ApplyLauncherShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(SlotSyncShmemCallbacks)
+PG_SHMEM_SUBSYSTEM(RepackShmemCallbacks)
 
 /* other modules that need some shared memory space */
 PG_SHMEM_SUBSYSTEM(BTreeShmemCallbacks)
diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list
index 637c669a146..d019e03aaf1 100644
--- a/src/tools/pgindent/typedefs.list
+++ b/src/tools/pgindent/typedefs.list
@@ -2639,9 +2639,12 @@ ReorderBufferTupleCidEnt
 ReorderBufferTupleCidKey
 ReorderBufferUpdateProgressTxnCB
 ReorderTuple
+RepackCleanupContext
 RepackCommand
 RepackDecodingState
+RepackShmemStruct
 RepackStmt
+RepackWorkerInfo
 ReparameterizeForeignPathByChild_function
 ReplOriginId
 ReplOriginXactState
-- 
2.47.3


--kdrcpfmkbkc4lqhu--





^ permalink  raw  reply  [nested|flat] 223+ messages in thread

* [PATCH 2/2] Publish list of tables being repacked in shared memory
@ 2026-04-07 20:29 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 223+ messages in thread

From: Álvaro Herrera @ 2026-04-07 20:29 UTC (permalink / raw)

Use it in autovacuum to skip processing tables that are being repacked.
This is mostly to avoid repeated attempts to process such tables, which
would fail due to the special deadlock checker behavior for repack.

Author: Álvaro Herrera <[email protected]>
Discussion: https://postgr.es/m/[email protected]
---
 src/backend/commands/repack.c                 | 195 ++++++++++++++++--
 src/backend/postmaster/autovacuum.c           |  20 ++
 .../utils/activity/wait_event_names.txt       |   1 +
 src/include/commands/repack.h                 |   2 +
 src/include/storage/lwlocklist.h              |   2 +-
 src/include/storage/subsystemlist.h           |   1 +
 src/tools/pgindent/typedefs.list              |   3 +
 7 files changed, 210 insertions(+), 14 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index a5f5df77291..ee7072dce6a 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -63,9 +63,11 @@
 #include "optimizer/optimizer.h"
 #include "pgstat.h"
 #include "storage/bufmgr.h"
+#include "storage/ipc.h"
 #include "storage/lmgr.h"
 #include "storage/predicate.h"
 #include "storage/proc.h"
+#include "storage/subsystems.h"
 #include "utils/acl.h"
 #include "utils/fmgroids.h"
 #include "utils/guc.h"
@@ -79,6 +81,32 @@
 #include "utils/syscache.h"
 #include "utils/wait_event_types.h"
 
+
+/* Shared memory layout for REPACK */
+typedef struct RepackWorkerInfo
+{
+	bool		ri_in_use;
+	pid_t		ri_backendpid;
+	Oid			ri_dbid;
+	Oid			ri_relid;
+	Oid			ri_toastrelid;
+} RepackWorkerInfo;
+
+typedef struct
+{
+	bool		re_useless;
+	RepackWorkerInfo re_workerinfo[FLEXIBLE_ARRAY_MEMBER];
+} RepackShmemStruct;
+
+static RepackShmemStruct *RepackShmem;
+
+typedef struct RepackCleanupContext
+{
+	bool		concurrent;
+	int			workerindex;
+} RepackCleanupContext;
+
+
 /*
  * This struct is used to pass around the information on tables to be
  * clustered. We need this so we can make a list of them when invoked without
@@ -90,6 +118,7 @@ typedef struct
 	Oid			indexOid;
 } RelToCluster;
 
+
 /*
  * The first file exported by the decoding worker must contain a snapshot, the
  * following ones contain the data changes.
@@ -166,6 +195,10 @@ static List *get_tables_to_repack_partitioned(RepackCommand cmd,
 											  MemoryContext permcxt);
 static bool repack_is_permitted_for_relation(RepackCommand cmd,
 											 Oid relid, Oid userid);
+static void RepackCleanup(RepackCleanupContext *context);
+static void RepackCleanupCb(int code, Datum arg);
+static void RepackShmemRequest(void *arg);
+static void RepackShmemInit(void *arg);
 
 static void apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt);
 static void apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
@@ -210,6 +243,11 @@ static void ProcessRepackMessage(StringInfo msg);
 static const char *RepackCommandAsString(RepackCommand cmd);
 
 
+const ShmemCallbacks RepackShmemCallbacks = {
+	.request_fn = RepackShmemRequest,
+	.init_fn = RepackShmemInit,
+};
+
 /*
  * The repack code allows for processing multiple tables at once. Because
  * of this, we cannot just run everything on a single transaction, or we
@@ -514,6 +552,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 	Oid			tableOid = RelationGetRelid(OldHeap);
 	Relation	index;
 	LOCKMODE	lmode;
+	RepackCleanupContext context;
 	Oid			save_userid;
 	int			save_sec_context;
 	int			save_nestlevel;
@@ -660,24 +699,43 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 		TransferPredicateLocksToHeapRelation(OldHeap);
 
 	/* rebuild_relation does all the dirty work */
-	PG_TRY();
-	{
-		rebuild_relation(OldHeap, index, verbose, ident_idx);
-	}
-	PG_FINALLY();
+	context.concurrent = concurrent;
+
+	PG_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
 	{
 		if (concurrent)
 		{
-			/*
-			 * Since during normal operation the worker was already asked to
-			 * exit, stopping it explicitly is especially important on ERROR.
-			 * However it still seems a good practice to make sure that the
-			 * worker never survives the REPACK command.
-			 */
-			stop_repack_decoding_worker();
+			bool		freefound = false;
+
+			LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+			for (int i = 0; i < max_repack_replication_slots; i++)
+			{
+				RepackWorkerInfo *worker;
+
+				if (RepackShmem->re_workerinfo[i].ri_in_use)
+					continue;
+
+				freefound = true;
+				worker = &RepackShmem->re_workerinfo[i];
+				context.workerindex = i;
+
+				worker->ri_in_use = true;
+				worker->ri_backendpid = MyProcPid;
+				worker->ri_dbid = MyDatabaseId;
+				worker->ri_relid = RelationGetRelid(OldHeap);
+				worker->ri_toastrelid = OldHeap->rd_rel->reltoastrelid;
+				break;
+			}
+			if (!freefound)
+				elog(ERROR, "could not find free repack entry");
+			LWLockRelease(RepackLock);
 		}
+
+		rebuild_relation(OldHeap, index, verbose, ident_idx);
 	}
-	PG_END_TRY();
+	PG_END_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
+
+	RepackCleanup(&context);
 
 	/* rebuild_relation closes OldHeap, and index if valid */
 
@@ -691,6 +749,117 @@ out:
 	pgstat_progress_end_command();
 }
 
+/*
+ * Return whether any backend is running concurrent REPACK on the given table
+ * (which could be a toast table).
+ */
+bool
+is_table_under_repack(Oid databaseId, Oid relid)
+{
+	bool		retval = false;
+
+	LWLockAcquire(RepackLock, LW_SHARED);
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		RepackWorkerInfo *rworker;
+
+		if (!RepackShmem->re_workerinfo[i].ri_in_use)
+			continue;
+
+		rworker = &RepackShmem->re_workerinfo[i];
+		if (rworker->ri_dbid == MyDatabaseId &&
+			(rworker->ri_relid == relid ||
+			 rworker->ri_toastrelid == relid))
+			retval = true;
+	}
+	LWLockRelease(RepackLock);
+
+	return retval;
+}
+
+/*
+ * Remove ourselves from the workerinfo array.
+ */
+static void
+RepackCleanup(RepackCleanupContext *context)
+{
+	if (context->concurrent)
+	{
+		RepackWorkerInfo *worker;
+
+		/*
+		 * The worker would normally terminate on its own when the work is
+		 * done, but make sure we signal it just in case.
+		 */
+		stop_repack_decoding_worker();
+
+		/*
+		 * also, make sure we stop advertising the relation we were repacking,
+		 * so that autovacuum reverts to handling it normally.
+		 */
+		LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+
+		worker = &RepackShmem->re_workerinfo[context->workerindex];
+		Assert(worker->ri_backendpid == MyProcPid);
+		worker->ri_in_use = false;
+		worker->ri_backendpid = 0;
+		worker->ri_dbid = InvalidOid;
+		worker->ri_relid = InvalidOid;
+		worker->ri_toastrelid = InvalidOid;
+		LWLockRelease(RepackLock);
+	}
+}
+
+/*
+ * RepackCleanup wrapped as an on_shmem_exit callback function
+ */
+static void
+RepackCleanupCb(int code, Datum arg)
+{
+	RepackCleanup((RepackCleanupContext *) DatumGetPointer(arg));
+}
+
+/*
+ * RepackShmemRequest
+ *		Register shared memory space needed for repack
+ */
+static void
+RepackShmemRequest(void *arg)
+{
+	Size		size;
+
+	/*
+	 * Need the fixed struct and the array of RepackWorkerInfo.
+	 */
+	size = sizeof(RepackShmemStruct);
+	size = MAXALIGN(size);
+	size = add_size(size, mul_size(max_repack_replication_slots,
+								   sizeof(RepackWorkerInfo)));
+
+	ShmemRequestStruct(.name = "Repack Data",
+					   .size = size,
+					   .ptr = (void **) &RepackShmem,
+		);
+}
+
+static void
+RepackShmemInit(void *arg)
+{
+	RepackWorkerInfo *reinfo;
+
+	reinfo = (RepackWorkerInfo *) ((char *) RepackShmem +
+								   MAXALIGN(sizeof(RepackShmemStruct)));
+
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		reinfo[i].ri_in_use = false;
+		reinfo[i].ri_backendpid = 0;
+		reinfo[i].ri_dbid = InvalidOid;
+		reinfo[i].ri_relid = InvalidOid;
+		reinfo[i].ri_toastrelid = InvalidOid;
+	}
+}
+
 /*
  * Check if the table (and its index) still meets the requirements of
  * cluster_rel().
diff --git a/src/backend/postmaster/autovacuum.c b/src/backend/postmaster/autovacuum.c
index bd626a16363..080c64ea3c8 100644
--- a/src/backend/postmaster/autovacuum.c
+++ b/src/backend/postmaster/autovacuum.c
@@ -78,6 +78,7 @@
 #include "catalog/namespace.h"
 #include "catalog/pg_database.h"
 #include "catalog/pg_namespace.h"
+#include "commands/repack.h"
 #include "commands/vacuum.h"
 #include "common/int.h"
 #include "funcapi.h"
@@ -2422,6 +2423,25 @@ do_autovacuum(void)
 			}
 		}
 		LWLockRelease(AutovacuumLock);
+
+		/*
+		 * Similarly, if the table is being processed by concurrent repack,
+		 * skip it (but make a note of that).  We wouldn't be able to acquire
+		 * its lock anyway.
+		 */
+		if (!skipit)
+		{
+			MemoryContextSwitchTo(PortalContext);
+
+			skipit = is_table_under_repack(MyDatabaseId, relid);
+			if (skipit)
+				ereport(LOG,
+						errmsg("skipping table \"%s.%s.%s\" because it's being repacked in concurrent mode",
+							   get_database_name(MyDatabaseId),
+							   get_namespace_name(get_rel_namespace(relid)),
+							   get_rel_name(relid)));
+		}
+
 		if (skipit)
 		{
 			LWLockRelease(AutovacuumScheduleLock);
diff --git a/src/backend/utils/activity/wait_event_names.txt b/src/backend/utils/activity/wait_event_names.txt
index 7bda5298558..e206304f204 100644
--- a/src/backend/utils/activity/wait_event_names.txt
+++ b/src/backend/utils/activity/wait_event_names.txt
@@ -332,6 +332,7 @@ SInvalWrite	"Waiting to add a message to the shared catalog invalidation queue."
 WALBufMapping	"Waiting to replace a page in WAL buffers."
 WALWrite	"Waiting for WAL buffers to be written to disk."
 ControlFile	"Waiting to read or update the <filename>pg_control</filename> file or create a new WAL file."
+Repack	"Waiting to read or update tables in process by concurrent repack."
 MultiXactGen	"Waiting to read or update shared multixact state."
 RelCacheInit	"Waiting to read or update a <filename>pg_internal.init</filename> relation cache initialization file."
 CheckpointerComm	"Waiting to manage fsync requests."
diff --git a/src/include/commands/repack.h b/src/include/commands/repack.h
index fd16e74b179..be7d38b5fae 100644
--- a/src/include/commands/repack.h
+++ b/src/include/commands/repack.h
@@ -42,6 +42,8 @@ extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel);
 
 extern void cluster_rel(RepackCommand command, Relation OldHeap, Oid indexOid,
 						ClusterParams *params, bool isTopLevel);
+extern bool is_table_under_repack(Oid databaseId, Oid relid);
+
 extern void check_index_is_clusterable(Relation OldHeap, Oid indexOid,
 									   LOCKMODE lockmode);
 extern void mark_index_clustered(Relation rel, Oid indexOid, bool is_internal);
diff --git a/src/include/storage/lwlocklist.h b/src/include/storage/lwlocklist.h
index af8553bcb6c..3f08f4a15d4 100644
--- a/src/include/storage/lwlocklist.h
+++ b/src/include/storage/lwlocklist.h
@@ -41,7 +41,7 @@ PG_LWLOCK(6, SInvalWrite)
 PG_LWLOCK(7, WALBufMapping)
 PG_LWLOCK(8, WALWrite)
 PG_LWLOCK(9, ControlFile)
-/* 10 was CheckpointLock */
+PG_LWLOCK(10, Repack)
 /* 11 was XactSLRULock */
 /* 12 was SubtransSLRULock */
 PG_LWLOCK(13, MultiXactGen)
diff --git a/src/include/storage/subsystemlist.h b/src/include/storage/subsystemlist.h
index 9ad619080be..4e683b8b0a8 100644
--- a/src/include/storage/subsystemlist.h
+++ b/src/include/storage/subsystemlist.h
@@ -72,6 +72,7 @@ PG_SHMEM_SUBSYSTEM(WalSummarizerShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(PgArchShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(ApplyLauncherShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(SlotSyncShmemCallbacks)
+PG_SHMEM_SUBSYSTEM(RepackShmemCallbacks)
 
 /* other modules that need some shared memory space */
 PG_SHMEM_SUBSYSTEM(BTreeShmemCallbacks)
diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list
index 637c669a146..d019e03aaf1 100644
--- a/src/tools/pgindent/typedefs.list
+++ b/src/tools/pgindent/typedefs.list
@@ -2639,9 +2639,12 @@ ReorderBufferTupleCidEnt
 ReorderBufferTupleCidKey
 ReorderBufferUpdateProgressTxnCB
 ReorderTuple
+RepackCleanupContext
 RepackCommand
 RepackDecodingState
+RepackShmemStruct
 RepackStmt
+RepackWorkerInfo
 ReparameterizeForeignPathByChild_function
 ReplOriginId
 ReplOriginXactState
-- 
2.47.3


--kdrcpfmkbkc4lqhu--





^ permalink  raw  reply  [nested|flat] 223+ messages in thread

* [PATCH 2/2] Publish list of tables being repacked in shared memory
@ 2026-04-07 20:29 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 223+ messages in thread

From: Álvaro Herrera @ 2026-04-07 20:29 UTC (permalink / raw)

Use it in autovacuum to skip processing tables that are being repacked.
This is mostly to avoid repeated attempts to process such tables, which
would fail due to the special deadlock checker behavior for repack.

Author: Álvaro Herrera <[email protected]>
Discussion: https://postgr.es/m/[email protected]
---
 src/backend/commands/repack.c                 | 195 ++++++++++++++++--
 src/backend/postmaster/autovacuum.c           |  20 ++
 .../utils/activity/wait_event_names.txt       |   1 +
 src/include/commands/repack.h                 |   2 +
 src/include/storage/lwlocklist.h              |   2 +-
 src/include/storage/subsystemlist.h           |   1 +
 src/tools/pgindent/typedefs.list              |   3 +
 7 files changed, 210 insertions(+), 14 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index a5f5df77291..ee7072dce6a 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -63,9 +63,11 @@
 #include "optimizer/optimizer.h"
 #include "pgstat.h"
 #include "storage/bufmgr.h"
+#include "storage/ipc.h"
 #include "storage/lmgr.h"
 #include "storage/predicate.h"
 #include "storage/proc.h"
+#include "storage/subsystems.h"
 #include "utils/acl.h"
 #include "utils/fmgroids.h"
 #include "utils/guc.h"
@@ -79,6 +81,32 @@
 #include "utils/syscache.h"
 #include "utils/wait_event_types.h"
 
+
+/* Shared memory layout for REPACK */
+typedef struct RepackWorkerInfo
+{
+	bool		ri_in_use;
+	pid_t		ri_backendpid;
+	Oid			ri_dbid;
+	Oid			ri_relid;
+	Oid			ri_toastrelid;
+} RepackWorkerInfo;
+
+typedef struct
+{
+	bool		re_useless;
+	RepackWorkerInfo re_workerinfo[FLEXIBLE_ARRAY_MEMBER];
+} RepackShmemStruct;
+
+static RepackShmemStruct *RepackShmem;
+
+typedef struct RepackCleanupContext
+{
+	bool		concurrent;
+	int			workerindex;
+} RepackCleanupContext;
+
+
 /*
  * This struct is used to pass around the information on tables to be
  * clustered. We need this so we can make a list of them when invoked without
@@ -90,6 +118,7 @@ typedef struct
 	Oid			indexOid;
 } RelToCluster;
 
+
 /*
  * The first file exported by the decoding worker must contain a snapshot, the
  * following ones contain the data changes.
@@ -166,6 +195,10 @@ static List *get_tables_to_repack_partitioned(RepackCommand cmd,
 											  MemoryContext permcxt);
 static bool repack_is_permitted_for_relation(RepackCommand cmd,
 											 Oid relid, Oid userid);
+static void RepackCleanup(RepackCleanupContext *context);
+static void RepackCleanupCb(int code, Datum arg);
+static void RepackShmemRequest(void *arg);
+static void RepackShmemInit(void *arg);
 
 static void apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt);
 static void apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
@@ -210,6 +243,11 @@ static void ProcessRepackMessage(StringInfo msg);
 static const char *RepackCommandAsString(RepackCommand cmd);
 
 
+const ShmemCallbacks RepackShmemCallbacks = {
+	.request_fn = RepackShmemRequest,
+	.init_fn = RepackShmemInit,
+};
+
 /*
  * The repack code allows for processing multiple tables at once. Because
  * of this, we cannot just run everything on a single transaction, or we
@@ -514,6 +552,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 	Oid			tableOid = RelationGetRelid(OldHeap);
 	Relation	index;
 	LOCKMODE	lmode;
+	RepackCleanupContext context;
 	Oid			save_userid;
 	int			save_sec_context;
 	int			save_nestlevel;
@@ -660,24 +699,43 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 		TransferPredicateLocksToHeapRelation(OldHeap);
 
 	/* rebuild_relation does all the dirty work */
-	PG_TRY();
-	{
-		rebuild_relation(OldHeap, index, verbose, ident_idx);
-	}
-	PG_FINALLY();
+	context.concurrent = concurrent;
+
+	PG_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
 	{
 		if (concurrent)
 		{
-			/*
-			 * Since during normal operation the worker was already asked to
-			 * exit, stopping it explicitly is especially important on ERROR.
-			 * However it still seems a good practice to make sure that the
-			 * worker never survives the REPACK command.
-			 */
-			stop_repack_decoding_worker();
+			bool		freefound = false;
+
+			LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+			for (int i = 0; i < max_repack_replication_slots; i++)
+			{
+				RepackWorkerInfo *worker;
+
+				if (RepackShmem->re_workerinfo[i].ri_in_use)
+					continue;
+
+				freefound = true;
+				worker = &RepackShmem->re_workerinfo[i];
+				context.workerindex = i;
+
+				worker->ri_in_use = true;
+				worker->ri_backendpid = MyProcPid;
+				worker->ri_dbid = MyDatabaseId;
+				worker->ri_relid = RelationGetRelid(OldHeap);
+				worker->ri_toastrelid = OldHeap->rd_rel->reltoastrelid;
+				break;
+			}
+			if (!freefound)
+				elog(ERROR, "could not find free repack entry");
+			LWLockRelease(RepackLock);
 		}
+
+		rebuild_relation(OldHeap, index, verbose, ident_idx);
 	}
-	PG_END_TRY();
+	PG_END_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
+
+	RepackCleanup(&context);
 
 	/* rebuild_relation closes OldHeap, and index if valid */
 
@@ -691,6 +749,117 @@ out:
 	pgstat_progress_end_command();
 }
 
+/*
+ * Return whether any backend is running concurrent REPACK on the given table
+ * (which could be a toast table).
+ */
+bool
+is_table_under_repack(Oid databaseId, Oid relid)
+{
+	bool		retval = false;
+
+	LWLockAcquire(RepackLock, LW_SHARED);
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		RepackWorkerInfo *rworker;
+
+		if (!RepackShmem->re_workerinfo[i].ri_in_use)
+			continue;
+
+		rworker = &RepackShmem->re_workerinfo[i];
+		if (rworker->ri_dbid == MyDatabaseId &&
+			(rworker->ri_relid == relid ||
+			 rworker->ri_toastrelid == relid))
+			retval = true;
+	}
+	LWLockRelease(RepackLock);
+
+	return retval;
+}
+
+/*
+ * Remove ourselves from the workerinfo array.
+ */
+static void
+RepackCleanup(RepackCleanupContext *context)
+{
+	if (context->concurrent)
+	{
+		RepackWorkerInfo *worker;
+
+		/*
+		 * The worker would normally terminate on its own when the work is
+		 * done, but make sure we signal it just in case.
+		 */
+		stop_repack_decoding_worker();
+
+		/*
+		 * also, make sure we stop advertising the relation we were repacking,
+		 * so that autovacuum reverts to handling it normally.
+		 */
+		LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+
+		worker = &RepackShmem->re_workerinfo[context->workerindex];
+		Assert(worker->ri_backendpid == MyProcPid);
+		worker->ri_in_use = false;
+		worker->ri_backendpid = 0;
+		worker->ri_dbid = InvalidOid;
+		worker->ri_relid = InvalidOid;
+		worker->ri_toastrelid = InvalidOid;
+		LWLockRelease(RepackLock);
+	}
+}
+
+/*
+ * RepackCleanup wrapped as an on_shmem_exit callback function
+ */
+static void
+RepackCleanupCb(int code, Datum arg)
+{
+	RepackCleanup((RepackCleanupContext *) DatumGetPointer(arg));
+}
+
+/*
+ * RepackShmemRequest
+ *		Register shared memory space needed for repack
+ */
+static void
+RepackShmemRequest(void *arg)
+{
+	Size		size;
+
+	/*
+	 * Need the fixed struct and the array of RepackWorkerInfo.
+	 */
+	size = sizeof(RepackShmemStruct);
+	size = MAXALIGN(size);
+	size = add_size(size, mul_size(max_repack_replication_slots,
+								   sizeof(RepackWorkerInfo)));
+
+	ShmemRequestStruct(.name = "Repack Data",
+					   .size = size,
+					   .ptr = (void **) &RepackShmem,
+		);
+}
+
+static void
+RepackShmemInit(void *arg)
+{
+	RepackWorkerInfo *reinfo;
+
+	reinfo = (RepackWorkerInfo *) ((char *) RepackShmem +
+								   MAXALIGN(sizeof(RepackShmemStruct)));
+
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		reinfo[i].ri_in_use = false;
+		reinfo[i].ri_backendpid = 0;
+		reinfo[i].ri_dbid = InvalidOid;
+		reinfo[i].ri_relid = InvalidOid;
+		reinfo[i].ri_toastrelid = InvalidOid;
+	}
+}
+
 /*
  * Check if the table (and its index) still meets the requirements of
  * cluster_rel().
diff --git a/src/backend/postmaster/autovacuum.c b/src/backend/postmaster/autovacuum.c
index bd626a16363..080c64ea3c8 100644
--- a/src/backend/postmaster/autovacuum.c
+++ b/src/backend/postmaster/autovacuum.c
@@ -78,6 +78,7 @@
 #include "catalog/namespace.h"
 #include "catalog/pg_database.h"
 #include "catalog/pg_namespace.h"
+#include "commands/repack.h"
 #include "commands/vacuum.h"
 #include "common/int.h"
 #include "funcapi.h"
@@ -2422,6 +2423,25 @@ do_autovacuum(void)
 			}
 		}
 		LWLockRelease(AutovacuumLock);
+
+		/*
+		 * Similarly, if the table is being processed by concurrent repack,
+		 * skip it (but make a note of that).  We wouldn't be able to acquire
+		 * its lock anyway.
+		 */
+		if (!skipit)
+		{
+			MemoryContextSwitchTo(PortalContext);
+
+			skipit = is_table_under_repack(MyDatabaseId, relid);
+			if (skipit)
+				ereport(LOG,
+						errmsg("skipping table \"%s.%s.%s\" because it's being repacked in concurrent mode",
+							   get_database_name(MyDatabaseId),
+							   get_namespace_name(get_rel_namespace(relid)),
+							   get_rel_name(relid)));
+		}
+
 		if (skipit)
 		{
 			LWLockRelease(AutovacuumScheduleLock);
diff --git a/src/backend/utils/activity/wait_event_names.txt b/src/backend/utils/activity/wait_event_names.txt
index 7bda5298558..e206304f204 100644
--- a/src/backend/utils/activity/wait_event_names.txt
+++ b/src/backend/utils/activity/wait_event_names.txt
@@ -332,6 +332,7 @@ SInvalWrite	"Waiting to add a message to the shared catalog invalidation queue."
 WALBufMapping	"Waiting to replace a page in WAL buffers."
 WALWrite	"Waiting for WAL buffers to be written to disk."
 ControlFile	"Waiting to read or update the <filename>pg_control</filename> file or create a new WAL file."
+Repack	"Waiting to read or update tables in process by concurrent repack."
 MultiXactGen	"Waiting to read or update shared multixact state."
 RelCacheInit	"Waiting to read or update a <filename>pg_internal.init</filename> relation cache initialization file."
 CheckpointerComm	"Waiting to manage fsync requests."
diff --git a/src/include/commands/repack.h b/src/include/commands/repack.h
index fd16e74b179..be7d38b5fae 100644
--- a/src/include/commands/repack.h
+++ b/src/include/commands/repack.h
@@ -42,6 +42,8 @@ extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel);
 
 extern void cluster_rel(RepackCommand command, Relation OldHeap, Oid indexOid,
 						ClusterParams *params, bool isTopLevel);
+extern bool is_table_under_repack(Oid databaseId, Oid relid);
+
 extern void check_index_is_clusterable(Relation OldHeap, Oid indexOid,
 									   LOCKMODE lockmode);
 extern void mark_index_clustered(Relation rel, Oid indexOid, bool is_internal);
diff --git a/src/include/storage/lwlocklist.h b/src/include/storage/lwlocklist.h
index af8553bcb6c..3f08f4a15d4 100644
--- a/src/include/storage/lwlocklist.h
+++ b/src/include/storage/lwlocklist.h
@@ -41,7 +41,7 @@ PG_LWLOCK(6, SInvalWrite)
 PG_LWLOCK(7, WALBufMapping)
 PG_LWLOCK(8, WALWrite)
 PG_LWLOCK(9, ControlFile)
-/* 10 was CheckpointLock */
+PG_LWLOCK(10, Repack)
 /* 11 was XactSLRULock */
 /* 12 was SubtransSLRULock */
 PG_LWLOCK(13, MultiXactGen)
diff --git a/src/include/storage/subsystemlist.h b/src/include/storage/subsystemlist.h
index 9ad619080be..4e683b8b0a8 100644
--- a/src/include/storage/subsystemlist.h
+++ b/src/include/storage/subsystemlist.h
@@ -72,6 +72,7 @@ PG_SHMEM_SUBSYSTEM(WalSummarizerShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(PgArchShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(ApplyLauncherShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(SlotSyncShmemCallbacks)
+PG_SHMEM_SUBSYSTEM(RepackShmemCallbacks)
 
 /* other modules that need some shared memory space */
 PG_SHMEM_SUBSYSTEM(BTreeShmemCallbacks)
diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list
index 637c669a146..d019e03aaf1 100644
--- a/src/tools/pgindent/typedefs.list
+++ b/src/tools/pgindent/typedefs.list
@@ -2639,9 +2639,12 @@ ReorderBufferTupleCidEnt
 ReorderBufferTupleCidKey
 ReorderBufferUpdateProgressTxnCB
 ReorderTuple
+RepackCleanupContext
 RepackCommand
 RepackDecodingState
+RepackShmemStruct
 RepackStmt
+RepackWorkerInfo
 ReparameterizeForeignPathByChild_function
 ReplOriginId
 ReplOriginXactState
-- 
2.47.3


--kdrcpfmkbkc4lqhu--





^ permalink  raw  reply  [nested|flat] 223+ messages in thread

* [PATCH 2/2] Publish list of tables being repacked in shared memory
@ 2026-04-07 20:29 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 223+ messages in thread

From: Álvaro Herrera @ 2026-04-07 20:29 UTC (permalink / raw)

Use it in autovacuum to skip processing tables that are being repacked.
This is mostly to avoid repeated attempts to process such tables, which
would fail due to the special deadlock checker behavior for repack.

Author: Álvaro Herrera <[email protected]>
Discussion: https://postgr.es/m/[email protected]
---
 src/backend/commands/repack.c                 | 195 ++++++++++++++++--
 src/backend/postmaster/autovacuum.c           |  20 ++
 .../utils/activity/wait_event_names.txt       |   1 +
 src/include/commands/repack.h                 |   2 +
 src/include/storage/lwlocklist.h              |   2 +-
 src/include/storage/subsystemlist.h           |   1 +
 src/tools/pgindent/typedefs.list              |   3 +
 7 files changed, 210 insertions(+), 14 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index a5f5df77291..ee7072dce6a 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -63,9 +63,11 @@
 #include "optimizer/optimizer.h"
 #include "pgstat.h"
 #include "storage/bufmgr.h"
+#include "storage/ipc.h"
 #include "storage/lmgr.h"
 #include "storage/predicate.h"
 #include "storage/proc.h"
+#include "storage/subsystems.h"
 #include "utils/acl.h"
 #include "utils/fmgroids.h"
 #include "utils/guc.h"
@@ -79,6 +81,32 @@
 #include "utils/syscache.h"
 #include "utils/wait_event_types.h"
 
+
+/* Shared memory layout for REPACK */
+typedef struct RepackWorkerInfo
+{
+	bool		ri_in_use;
+	pid_t		ri_backendpid;
+	Oid			ri_dbid;
+	Oid			ri_relid;
+	Oid			ri_toastrelid;
+} RepackWorkerInfo;
+
+typedef struct
+{
+	bool		re_useless;
+	RepackWorkerInfo re_workerinfo[FLEXIBLE_ARRAY_MEMBER];
+} RepackShmemStruct;
+
+static RepackShmemStruct *RepackShmem;
+
+typedef struct RepackCleanupContext
+{
+	bool		concurrent;
+	int			workerindex;
+} RepackCleanupContext;
+
+
 /*
  * This struct is used to pass around the information on tables to be
  * clustered. We need this so we can make a list of them when invoked without
@@ -90,6 +118,7 @@ typedef struct
 	Oid			indexOid;
 } RelToCluster;
 
+
 /*
  * The first file exported by the decoding worker must contain a snapshot, the
  * following ones contain the data changes.
@@ -166,6 +195,10 @@ static List *get_tables_to_repack_partitioned(RepackCommand cmd,
 											  MemoryContext permcxt);
 static bool repack_is_permitted_for_relation(RepackCommand cmd,
 											 Oid relid, Oid userid);
+static void RepackCleanup(RepackCleanupContext *context);
+static void RepackCleanupCb(int code, Datum arg);
+static void RepackShmemRequest(void *arg);
+static void RepackShmemInit(void *arg);
 
 static void apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt);
 static void apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
@@ -210,6 +243,11 @@ static void ProcessRepackMessage(StringInfo msg);
 static const char *RepackCommandAsString(RepackCommand cmd);
 
 
+const ShmemCallbacks RepackShmemCallbacks = {
+	.request_fn = RepackShmemRequest,
+	.init_fn = RepackShmemInit,
+};
+
 /*
  * The repack code allows for processing multiple tables at once. Because
  * of this, we cannot just run everything on a single transaction, or we
@@ -514,6 +552,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 	Oid			tableOid = RelationGetRelid(OldHeap);
 	Relation	index;
 	LOCKMODE	lmode;
+	RepackCleanupContext context;
 	Oid			save_userid;
 	int			save_sec_context;
 	int			save_nestlevel;
@@ -660,24 +699,43 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 		TransferPredicateLocksToHeapRelation(OldHeap);
 
 	/* rebuild_relation does all the dirty work */
-	PG_TRY();
-	{
-		rebuild_relation(OldHeap, index, verbose, ident_idx);
-	}
-	PG_FINALLY();
+	context.concurrent = concurrent;
+
+	PG_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
 	{
 		if (concurrent)
 		{
-			/*
-			 * Since during normal operation the worker was already asked to
-			 * exit, stopping it explicitly is especially important on ERROR.
-			 * However it still seems a good practice to make sure that the
-			 * worker never survives the REPACK command.
-			 */
-			stop_repack_decoding_worker();
+			bool		freefound = false;
+
+			LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+			for (int i = 0; i < max_repack_replication_slots; i++)
+			{
+				RepackWorkerInfo *worker;
+
+				if (RepackShmem->re_workerinfo[i].ri_in_use)
+					continue;
+
+				freefound = true;
+				worker = &RepackShmem->re_workerinfo[i];
+				context.workerindex = i;
+
+				worker->ri_in_use = true;
+				worker->ri_backendpid = MyProcPid;
+				worker->ri_dbid = MyDatabaseId;
+				worker->ri_relid = RelationGetRelid(OldHeap);
+				worker->ri_toastrelid = OldHeap->rd_rel->reltoastrelid;
+				break;
+			}
+			if (!freefound)
+				elog(ERROR, "could not find free repack entry");
+			LWLockRelease(RepackLock);
 		}
+
+		rebuild_relation(OldHeap, index, verbose, ident_idx);
 	}
-	PG_END_TRY();
+	PG_END_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
+
+	RepackCleanup(&context);
 
 	/* rebuild_relation closes OldHeap, and index if valid */
 
@@ -691,6 +749,117 @@ out:
 	pgstat_progress_end_command();
 }
 
+/*
+ * Return whether any backend is running concurrent REPACK on the given table
+ * (which could be a toast table).
+ */
+bool
+is_table_under_repack(Oid databaseId, Oid relid)
+{
+	bool		retval = false;
+
+	LWLockAcquire(RepackLock, LW_SHARED);
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		RepackWorkerInfo *rworker;
+
+		if (!RepackShmem->re_workerinfo[i].ri_in_use)
+			continue;
+
+		rworker = &RepackShmem->re_workerinfo[i];
+		if (rworker->ri_dbid == MyDatabaseId &&
+			(rworker->ri_relid == relid ||
+			 rworker->ri_toastrelid == relid))
+			retval = true;
+	}
+	LWLockRelease(RepackLock);
+
+	return retval;
+}
+
+/*
+ * Remove ourselves from the workerinfo array.
+ */
+static void
+RepackCleanup(RepackCleanupContext *context)
+{
+	if (context->concurrent)
+	{
+		RepackWorkerInfo *worker;
+
+		/*
+		 * The worker would normally terminate on its own when the work is
+		 * done, but make sure we signal it just in case.
+		 */
+		stop_repack_decoding_worker();
+
+		/*
+		 * also, make sure we stop advertising the relation we were repacking,
+		 * so that autovacuum reverts to handling it normally.
+		 */
+		LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+
+		worker = &RepackShmem->re_workerinfo[context->workerindex];
+		Assert(worker->ri_backendpid == MyProcPid);
+		worker->ri_in_use = false;
+		worker->ri_backendpid = 0;
+		worker->ri_dbid = InvalidOid;
+		worker->ri_relid = InvalidOid;
+		worker->ri_toastrelid = InvalidOid;
+		LWLockRelease(RepackLock);
+	}
+}
+
+/*
+ * RepackCleanup wrapped as an on_shmem_exit callback function
+ */
+static void
+RepackCleanupCb(int code, Datum arg)
+{
+	RepackCleanup((RepackCleanupContext *) DatumGetPointer(arg));
+}
+
+/*
+ * RepackShmemRequest
+ *		Register shared memory space needed for repack
+ */
+static void
+RepackShmemRequest(void *arg)
+{
+	Size		size;
+
+	/*
+	 * Need the fixed struct and the array of RepackWorkerInfo.
+	 */
+	size = sizeof(RepackShmemStruct);
+	size = MAXALIGN(size);
+	size = add_size(size, mul_size(max_repack_replication_slots,
+								   sizeof(RepackWorkerInfo)));
+
+	ShmemRequestStruct(.name = "Repack Data",
+					   .size = size,
+					   .ptr = (void **) &RepackShmem,
+		);
+}
+
+static void
+RepackShmemInit(void *arg)
+{
+	RepackWorkerInfo *reinfo;
+
+	reinfo = (RepackWorkerInfo *) ((char *) RepackShmem +
+								   MAXALIGN(sizeof(RepackShmemStruct)));
+
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		reinfo[i].ri_in_use = false;
+		reinfo[i].ri_backendpid = 0;
+		reinfo[i].ri_dbid = InvalidOid;
+		reinfo[i].ri_relid = InvalidOid;
+		reinfo[i].ri_toastrelid = InvalidOid;
+	}
+}
+
 /*
  * Check if the table (and its index) still meets the requirements of
  * cluster_rel().
diff --git a/src/backend/postmaster/autovacuum.c b/src/backend/postmaster/autovacuum.c
index bd626a16363..080c64ea3c8 100644
--- a/src/backend/postmaster/autovacuum.c
+++ b/src/backend/postmaster/autovacuum.c
@@ -78,6 +78,7 @@
 #include "catalog/namespace.h"
 #include "catalog/pg_database.h"
 #include "catalog/pg_namespace.h"
+#include "commands/repack.h"
 #include "commands/vacuum.h"
 #include "common/int.h"
 #include "funcapi.h"
@@ -2422,6 +2423,25 @@ do_autovacuum(void)
 			}
 		}
 		LWLockRelease(AutovacuumLock);
+
+		/*
+		 * Similarly, if the table is being processed by concurrent repack,
+		 * skip it (but make a note of that).  We wouldn't be able to acquire
+		 * its lock anyway.
+		 */
+		if (!skipit)
+		{
+			MemoryContextSwitchTo(PortalContext);
+
+			skipit = is_table_under_repack(MyDatabaseId, relid);
+			if (skipit)
+				ereport(LOG,
+						errmsg("skipping table \"%s.%s.%s\" because it's being repacked in concurrent mode",
+							   get_database_name(MyDatabaseId),
+							   get_namespace_name(get_rel_namespace(relid)),
+							   get_rel_name(relid)));
+		}
+
 		if (skipit)
 		{
 			LWLockRelease(AutovacuumScheduleLock);
diff --git a/src/backend/utils/activity/wait_event_names.txt b/src/backend/utils/activity/wait_event_names.txt
index 7bda5298558..e206304f204 100644
--- a/src/backend/utils/activity/wait_event_names.txt
+++ b/src/backend/utils/activity/wait_event_names.txt
@@ -332,6 +332,7 @@ SInvalWrite	"Waiting to add a message to the shared catalog invalidation queue."
 WALBufMapping	"Waiting to replace a page in WAL buffers."
 WALWrite	"Waiting for WAL buffers to be written to disk."
 ControlFile	"Waiting to read or update the <filename>pg_control</filename> file or create a new WAL file."
+Repack	"Waiting to read or update tables in process by concurrent repack."
 MultiXactGen	"Waiting to read or update shared multixact state."
 RelCacheInit	"Waiting to read or update a <filename>pg_internal.init</filename> relation cache initialization file."
 CheckpointerComm	"Waiting to manage fsync requests."
diff --git a/src/include/commands/repack.h b/src/include/commands/repack.h
index fd16e74b179..be7d38b5fae 100644
--- a/src/include/commands/repack.h
+++ b/src/include/commands/repack.h
@@ -42,6 +42,8 @@ extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel);
 
 extern void cluster_rel(RepackCommand command, Relation OldHeap, Oid indexOid,
 						ClusterParams *params, bool isTopLevel);
+extern bool is_table_under_repack(Oid databaseId, Oid relid);
+
 extern void check_index_is_clusterable(Relation OldHeap, Oid indexOid,
 									   LOCKMODE lockmode);
 extern void mark_index_clustered(Relation rel, Oid indexOid, bool is_internal);
diff --git a/src/include/storage/lwlocklist.h b/src/include/storage/lwlocklist.h
index af8553bcb6c..3f08f4a15d4 100644
--- a/src/include/storage/lwlocklist.h
+++ b/src/include/storage/lwlocklist.h
@@ -41,7 +41,7 @@ PG_LWLOCK(6, SInvalWrite)
 PG_LWLOCK(7, WALBufMapping)
 PG_LWLOCK(8, WALWrite)
 PG_LWLOCK(9, ControlFile)
-/* 10 was CheckpointLock */
+PG_LWLOCK(10, Repack)
 /* 11 was XactSLRULock */
 /* 12 was SubtransSLRULock */
 PG_LWLOCK(13, MultiXactGen)
diff --git a/src/include/storage/subsystemlist.h b/src/include/storage/subsystemlist.h
index 9ad619080be..4e683b8b0a8 100644
--- a/src/include/storage/subsystemlist.h
+++ b/src/include/storage/subsystemlist.h
@@ -72,6 +72,7 @@ PG_SHMEM_SUBSYSTEM(WalSummarizerShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(PgArchShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(ApplyLauncherShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(SlotSyncShmemCallbacks)
+PG_SHMEM_SUBSYSTEM(RepackShmemCallbacks)
 
 /* other modules that need some shared memory space */
 PG_SHMEM_SUBSYSTEM(BTreeShmemCallbacks)
diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list
index 637c669a146..d019e03aaf1 100644
--- a/src/tools/pgindent/typedefs.list
+++ b/src/tools/pgindent/typedefs.list
@@ -2639,9 +2639,12 @@ ReorderBufferTupleCidEnt
 ReorderBufferTupleCidKey
 ReorderBufferUpdateProgressTxnCB
 ReorderTuple
+RepackCleanupContext
 RepackCommand
 RepackDecodingState
+RepackShmemStruct
 RepackStmt
+RepackWorkerInfo
 ReparameterizeForeignPathByChild_function
 ReplOriginId
 ReplOriginXactState
-- 
2.47.3


--kdrcpfmkbkc4lqhu--





^ permalink  raw  reply  [nested|flat] 223+ messages in thread

* [PATCH 2/2] Publish list of tables being repacked in shared memory
@ 2026-04-07 20:29 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 223+ messages in thread

From: Álvaro Herrera @ 2026-04-07 20:29 UTC (permalink / raw)

Use it in autovacuum to skip processing tables that are being repacked.
This is mostly to avoid repeated attempts to process such tables, which
would fail due to the special deadlock checker behavior for repack.

Author: Álvaro Herrera <[email protected]>
Discussion: https://postgr.es/m/[email protected]
---
 src/backend/commands/repack.c                 | 195 ++++++++++++++++--
 src/backend/postmaster/autovacuum.c           |  20 ++
 .../utils/activity/wait_event_names.txt       |   1 +
 src/include/commands/repack.h                 |   2 +
 src/include/storage/lwlocklist.h              |   2 +-
 src/include/storage/subsystemlist.h           |   1 +
 src/tools/pgindent/typedefs.list              |   3 +
 7 files changed, 210 insertions(+), 14 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index a5f5df77291..ee7072dce6a 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -63,9 +63,11 @@
 #include "optimizer/optimizer.h"
 #include "pgstat.h"
 #include "storage/bufmgr.h"
+#include "storage/ipc.h"
 #include "storage/lmgr.h"
 #include "storage/predicate.h"
 #include "storage/proc.h"
+#include "storage/subsystems.h"
 #include "utils/acl.h"
 #include "utils/fmgroids.h"
 #include "utils/guc.h"
@@ -79,6 +81,32 @@
 #include "utils/syscache.h"
 #include "utils/wait_event_types.h"
 
+
+/* Shared memory layout for REPACK */
+typedef struct RepackWorkerInfo
+{
+	bool		ri_in_use;
+	pid_t		ri_backendpid;
+	Oid			ri_dbid;
+	Oid			ri_relid;
+	Oid			ri_toastrelid;
+} RepackWorkerInfo;
+
+typedef struct
+{
+	bool		re_useless;
+	RepackWorkerInfo re_workerinfo[FLEXIBLE_ARRAY_MEMBER];
+} RepackShmemStruct;
+
+static RepackShmemStruct *RepackShmem;
+
+typedef struct RepackCleanupContext
+{
+	bool		concurrent;
+	int			workerindex;
+} RepackCleanupContext;
+
+
 /*
  * This struct is used to pass around the information on tables to be
  * clustered. We need this so we can make a list of them when invoked without
@@ -90,6 +118,7 @@ typedef struct
 	Oid			indexOid;
 } RelToCluster;
 
+
 /*
  * The first file exported by the decoding worker must contain a snapshot, the
  * following ones contain the data changes.
@@ -166,6 +195,10 @@ static List *get_tables_to_repack_partitioned(RepackCommand cmd,
 											  MemoryContext permcxt);
 static bool repack_is_permitted_for_relation(RepackCommand cmd,
 											 Oid relid, Oid userid);
+static void RepackCleanup(RepackCleanupContext *context);
+static void RepackCleanupCb(int code, Datum arg);
+static void RepackShmemRequest(void *arg);
+static void RepackShmemInit(void *arg);
 
 static void apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt);
 static void apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
@@ -210,6 +243,11 @@ static void ProcessRepackMessage(StringInfo msg);
 static const char *RepackCommandAsString(RepackCommand cmd);
 
 
+const ShmemCallbacks RepackShmemCallbacks = {
+	.request_fn = RepackShmemRequest,
+	.init_fn = RepackShmemInit,
+};
+
 /*
  * The repack code allows for processing multiple tables at once. Because
  * of this, we cannot just run everything on a single transaction, or we
@@ -514,6 +552,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 	Oid			tableOid = RelationGetRelid(OldHeap);
 	Relation	index;
 	LOCKMODE	lmode;
+	RepackCleanupContext context;
 	Oid			save_userid;
 	int			save_sec_context;
 	int			save_nestlevel;
@@ -660,24 +699,43 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 		TransferPredicateLocksToHeapRelation(OldHeap);
 
 	/* rebuild_relation does all the dirty work */
-	PG_TRY();
-	{
-		rebuild_relation(OldHeap, index, verbose, ident_idx);
-	}
-	PG_FINALLY();
+	context.concurrent = concurrent;
+
+	PG_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
 	{
 		if (concurrent)
 		{
-			/*
-			 * Since during normal operation the worker was already asked to
-			 * exit, stopping it explicitly is especially important on ERROR.
-			 * However it still seems a good practice to make sure that the
-			 * worker never survives the REPACK command.
-			 */
-			stop_repack_decoding_worker();
+			bool		freefound = false;
+
+			LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+			for (int i = 0; i < max_repack_replication_slots; i++)
+			{
+				RepackWorkerInfo *worker;
+
+				if (RepackShmem->re_workerinfo[i].ri_in_use)
+					continue;
+
+				freefound = true;
+				worker = &RepackShmem->re_workerinfo[i];
+				context.workerindex = i;
+
+				worker->ri_in_use = true;
+				worker->ri_backendpid = MyProcPid;
+				worker->ri_dbid = MyDatabaseId;
+				worker->ri_relid = RelationGetRelid(OldHeap);
+				worker->ri_toastrelid = OldHeap->rd_rel->reltoastrelid;
+				break;
+			}
+			if (!freefound)
+				elog(ERROR, "could not find free repack entry");
+			LWLockRelease(RepackLock);
 		}
+
+		rebuild_relation(OldHeap, index, verbose, ident_idx);
 	}
-	PG_END_TRY();
+	PG_END_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
+
+	RepackCleanup(&context);
 
 	/* rebuild_relation closes OldHeap, and index if valid */
 
@@ -691,6 +749,117 @@ out:
 	pgstat_progress_end_command();
 }
 
+/*
+ * Return whether any backend is running concurrent REPACK on the given table
+ * (which could be a toast table).
+ */
+bool
+is_table_under_repack(Oid databaseId, Oid relid)
+{
+	bool		retval = false;
+
+	LWLockAcquire(RepackLock, LW_SHARED);
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		RepackWorkerInfo *rworker;
+
+		if (!RepackShmem->re_workerinfo[i].ri_in_use)
+			continue;
+
+		rworker = &RepackShmem->re_workerinfo[i];
+		if (rworker->ri_dbid == MyDatabaseId &&
+			(rworker->ri_relid == relid ||
+			 rworker->ri_toastrelid == relid))
+			retval = true;
+	}
+	LWLockRelease(RepackLock);
+
+	return retval;
+}
+
+/*
+ * Remove ourselves from the workerinfo array.
+ */
+static void
+RepackCleanup(RepackCleanupContext *context)
+{
+	if (context->concurrent)
+	{
+		RepackWorkerInfo *worker;
+
+		/*
+		 * The worker would normally terminate on its own when the work is
+		 * done, but make sure we signal it just in case.
+		 */
+		stop_repack_decoding_worker();
+
+		/*
+		 * also, make sure we stop advertising the relation we were repacking,
+		 * so that autovacuum reverts to handling it normally.
+		 */
+		LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+
+		worker = &RepackShmem->re_workerinfo[context->workerindex];
+		Assert(worker->ri_backendpid == MyProcPid);
+		worker->ri_in_use = false;
+		worker->ri_backendpid = 0;
+		worker->ri_dbid = InvalidOid;
+		worker->ri_relid = InvalidOid;
+		worker->ri_toastrelid = InvalidOid;
+		LWLockRelease(RepackLock);
+	}
+}
+
+/*
+ * RepackCleanup wrapped as an on_shmem_exit callback function
+ */
+static void
+RepackCleanupCb(int code, Datum arg)
+{
+	RepackCleanup((RepackCleanupContext *) DatumGetPointer(arg));
+}
+
+/*
+ * RepackShmemRequest
+ *		Register shared memory space needed for repack
+ */
+static void
+RepackShmemRequest(void *arg)
+{
+	Size		size;
+
+	/*
+	 * Need the fixed struct and the array of RepackWorkerInfo.
+	 */
+	size = sizeof(RepackShmemStruct);
+	size = MAXALIGN(size);
+	size = add_size(size, mul_size(max_repack_replication_slots,
+								   sizeof(RepackWorkerInfo)));
+
+	ShmemRequestStruct(.name = "Repack Data",
+					   .size = size,
+					   .ptr = (void **) &RepackShmem,
+		);
+}
+
+static void
+RepackShmemInit(void *arg)
+{
+	RepackWorkerInfo *reinfo;
+
+	reinfo = (RepackWorkerInfo *) ((char *) RepackShmem +
+								   MAXALIGN(sizeof(RepackShmemStruct)));
+
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		reinfo[i].ri_in_use = false;
+		reinfo[i].ri_backendpid = 0;
+		reinfo[i].ri_dbid = InvalidOid;
+		reinfo[i].ri_relid = InvalidOid;
+		reinfo[i].ri_toastrelid = InvalidOid;
+	}
+}
+
 /*
  * Check if the table (and its index) still meets the requirements of
  * cluster_rel().
diff --git a/src/backend/postmaster/autovacuum.c b/src/backend/postmaster/autovacuum.c
index bd626a16363..080c64ea3c8 100644
--- a/src/backend/postmaster/autovacuum.c
+++ b/src/backend/postmaster/autovacuum.c
@@ -78,6 +78,7 @@
 #include "catalog/namespace.h"
 #include "catalog/pg_database.h"
 #include "catalog/pg_namespace.h"
+#include "commands/repack.h"
 #include "commands/vacuum.h"
 #include "common/int.h"
 #include "funcapi.h"
@@ -2422,6 +2423,25 @@ do_autovacuum(void)
 			}
 		}
 		LWLockRelease(AutovacuumLock);
+
+		/*
+		 * Similarly, if the table is being processed by concurrent repack,
+		 * skip it (but make a note of that).  We wouldn't be able to acquire
+		 * its lock anyway.
+		 */
+		if (!skipit)
+		{
+			MemoryContextSwitchTo(PortalContext);
+
+			skipit = is_table_under_repack(MyDatabaseId, relid);
+			if (skipit)
+				ereport(LOG,
+						errmsg("skipping table \"%s.%s.%s\" because it's being repacked in concurrent mode",
+							   get_database_name(MyDatabaseId),
+							   get_namespace_name(get_rel_namespace(relid)),
+							   get_rel_name(relid)));
+		}
+
 		if (skipit)
 		{
 			LWLockRelease(AutovacuumScheduleLock);
diff --git a/src/backend/utils/activity/wait_event_names.txt b/src/backend/utils/activity/wait_event_names.txt
index 7bda5298558..e206304f204 100644
--- a/src/backend/utils/activity/wait_event_names.txt
+++ b/src/backend/utils/activity/wait_event_names.txt
@@ -332,6 +332,7 @@ SInvalWrite	"Waiting to add a message to the shared catalog invalidation queue."
 WALBufMapping	"Waiting to replace a page in WAL buffers."
 WALWrite	"Waiting for WAL buffers to be written to disk."
 ControlFile	"Waiting to read or update the <filename>pg_control</filename> file or create a new WAL file."
+Repack	"Waiting to read or update tables in process by concurrent repack."
 MultiXactGen	"Waiting to read or update shared multixact state."
 RelCacheInit	"Waiting to read or update a <filename>pg_internal.init</filename> relation cache initialization file."
 CheckpointerComm	"Waiting to manage fsync requests."
diff --git a/src/include/commands/repack.h b/src/include/commands/repack.h
index fd16e74b179..be7d38b5fae 100644
--- a/src/include/commands/repack.h
+++ b/src/include/commands/repack.h
@@ -42,6 +42,8 @@ extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel);
 
 extern void cluster_rel(RepackCommand command, Relation OldHeap, Oid indexOid,
 						ClusterParams *params, bool isTopLevel);
+extern bool is_table_under_repack(Oid databaseId, Oid relid);
+
 extern void check_index_is_clusterable(Relation OldHeap, Oid indexOid,
 									   LOCKMODE lockmode);
 extern void mark_index_clustered(Relation rel, Oid indexOid, bool is_internal);
diff --git a/src/include/storage/lwlocklist.h b/src/include/storage/lwlocklist.h
index af8553bcb6c..3f08f4a15d4 100644
--- a/src/include/storage/lwlocklist.h
+++ b/src/include/storage/lwlocklist.h
@@ -41,7 +41,7 @@ PG_LWLOCK(6, SInvalWrite)
 PG_LWLOCK(7, WALBufMapping)
 PG_LWLOCK(8, WALWrite)
 PG_LWLOCK(9, ControlFile)
-/* 10 was CheckpointLock */
+PG_LWLOCK(10, Repack)
 /* 11 was XactSLRULock */
 /* 12 was SubtransSLRULock */
 PG_LWLOCK(13, MultiXactGen)
diff --git a/src/include/storage/subsystemlist.h b/src/include/storage/subsystemlist.h
index 9ad619080be..4e683b8b0a8 100644
--- a/src/include/storage/subsystemlist.h
+++ b/src/include/storage/subsystemlist.h
@@ -72,6 +72,7 @@ PG_SHMEM_SUBSYSTEM(WalSummarizerShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(PgArchShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(ApplyLauncherShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(SlotSyncShmemCallbacks)
+PG_SHMEM_SUBSYSTEM(RepackShmemCallbacks)
 
 /* other modules that need some shared memory space */
 PG_SHMEM_SUBSYSTEM(BTreeShmemCallbacks)
diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list
index 637c669a146..d019e03aaf1 100644
--- a/src/tools/pgindent/typedefs.list
+++ b/src/tools/pgindent/typedefs.list
@@ -2639,9 +2639,12 @@ ReorderBufferTupleCidEnt
 ReorderBufferTupleCidKey
 ReorderBufferUpdateProgressTxnCB
 ReorderTuple
+RepackCleanupContext
 RepackCommand
 RepackDecodingState
+RepackShmemStruct
 RepackStmt
+RepackWorkerInfo
 ReparameterizeForeignPathByChild_function
 ReplOriginId
 ReplOriginXactState
-- 
2.47.3


--kdrcpfmkbkc4lqhu--





^ permalink  raw  reply  [nested|flat] 223+ messages in thread

* [PATCH 2/2] Publish list of tables being repacked in shared memory
@ 2026-04-07 20:29 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 223+ messages in thread

From: Álvaro Herrera @ 2026-04-07 20:29 UTC (permalink / raw)

Use it in autovacuum to skip processing tables that are being repacked.
This is mostly to avoid repeated attempts to process such tables, which
would fail due to the special deadlock checker behavior for repack.

Author: Álvaro Herrera <[email protected]>
Discussion: https://postgr.es/m/[email protected]
---
 src/backend/commands/repack.c                 | 195 ++++++++++++++++--
 src/backend/postmaster/autovacuum.c           |  20 ++
 .../utils/activity/wait_event_names.txt       |   1 +
 src/include/commands/repack.h                 |   2 +
 src/include/storage/lwlocklist.h              |   2 +-
 src/include/storage/subsystemlist.h           |   1 +
 src/tools/pgindent/typedefs.list              |   3 +
 7 files changed, 210 insertions(+), 14 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index a5f5df77291..ee7072dce6a 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -63,9 +63,11 @@
 #include "optimizer/optimizer.h"
 #include "pgstat.h"
 #include "storage/bufmgr.h"
+#include "storage/ipc.h"
 #include "storage/lmgr.h"
 #include "storage/predicate.h"
 #include "storage/proc.h"
+#include "storage/subsystems.h"
 #include "utils/acl.h"
 #include "utils/fmgroids.h"
 #include "utils/guc.h"
@@ -79,6 +81,32 @@
 #include "utils/syscache.h"
 #include "utils/wait_event_types.h"
 
+
+/* Shared memory layout for REPACK */
+typedef struct RepackWorkerInfo
+{
+	bool		ri_in_use;
+	pid_t		ri_backendpid;
+	Oid			ri_dbid;
+	Oid			ri_relid;
+	Oid			ri_toastrelid;
+} RepackWorkerInfo;
+
+typedef struct
+{
+	bool		re_useless;
+	RepackWorkerInfo re_workerinfo[FLEXIBLE_ARRAY_MEMBER];
+} RepackShmemStruct;
+
+static RepackShmemStruct *RepackShmem;
+
+typedef struct RepackCleanupContext
+{
+	bool		concurrent;
+	int			workerindex;
+} RepackCleanupContext;
+
+
 /*
  * This struct is used to pass around the information on tables to be
  * clustered. We need this so we can make a list of them when invoked without
@@ -90,6 +118,7 @@ typedef struct
 	Oid			indexOid;
 } RelToCluster;
 
+
 /*
  * The first file exported by the decoding worker must contain a snapshot, the
  * following ones contain the data changes.
@@ -166,6 +195,10 @@ static List *get_tables_to_repack_partitioned(RepackCommand cmd,
 											  MemoryContext permcxt);
 static bool repack_is_permitted_for_relation(RepackCommand cmd,
 											 Oid relid, Oid userid);
+static void RepackCleanup(RepackCleanupContext *context);
+static void RepackCleanupCb(int code, Datum arg);
+static void RepackShmemRequest(void *arg);
+static void RepackShmemInit(void *arg);
 
 static void apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt);
 static void apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
@@ -210,6 +243,11 @@ static void ProcessRepackMessage(StringInfo msg);
 static const char *RepackCommandAsString(RepackCommand cmd);
 
 
+const ShmemCallbacks RepackShmemCallbacks = {
+	.request_fn = RepackShmemRequest,
+	.init_fn = RepackShmemInit,
+};
+
 /*
  * The repack code allows for processing multiple tables at once. Because
  * of this, we cannot just run everything on a single transaction, or we
@@ -514,6 +552,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 	Oid			tableOid = RelationGetRelid(OldHeap);
 	Relation	index;
 	LOCKMODE	lmode;
+	RepackCleanupContext context;
 	Oid			save_userid;
 	int			save_sec_context;
 	int			save_nestlevel;
@@ -660,24 +699,43 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 		TransferPredicateLocksToHeapRelation(OldHeap);
 
 	/* rebuild_relation does all the dirty work */
-	PG_TRY();
-	{
-		rebuild_relation(OldHeap, index, verbose, ident_idx);
-	}
-	PG_FINALLY();
+	context.concurrent = concurrent;
+
+	PG_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
 	{
 		if (concurrent)
 		{
-			/*
-			 * Since during normal operation the worker was already asked to
-			 * exit, stopping it explicitly is especially important on ERROR.
-			 * However it still seems a good practice to make sure that the
-			 * worker never survives the REPACK command.
-			 */
-			stop_repack_decoding_worker();
+			bool		freefound = false;
+
+			LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+			for (int i = 0; i < max_repack_replication_slots; i++)
+			{
+				RepackWorkerInfo *worker;
+
+				if (RepackShmem->re_workerinfo[i].ri_in_use)
+					continue;
+
+				freefound = true;
+				worker = &RepackShmem->re_workerinfo[i];
+				context.workerindex = i;
+
+				worker->ri_in_use = true;
+				worker->ri_backendpid = MyProcPid;
+				worker->ri_dbid = MyDatabaseId;
+				worker->ri_relid = RelationGetRelid(OldHeap);
+				worker->ri_toastrelid = OldHeap->rd_rel->reltoastrelid;
+				break;
+			}
+			if (!freefound)
+				elog(ERROR, "could not find free repack entry");
+			LWLockRelease(RepackLock);
 		}
+
+		rebuild_relation(OldHeap, index, verbose, ident_idx);
 	}
-	PG_END_TRY();
+	PG_END_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
+
+	RepackCleanup(&context);
 
 	/* rebuild_relation closes OldHeap, and index if valid */
 
@@ -691,6 +749,117 @@ out:
 	pgstat_progress_end_command();
 }
 
+/*
+ * Return whether any backend is running concurrent REPACK on the given table
+ * (which could be a toast table).
+ */
+bool
+is_table_under_repack(Oid databaseId, Oid relid)
+{
+	bool		retval = false;
+
+	LWLockAcquire(RepackLock, LW_SHARED);
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		RepackWorkerInfo *rworker;
+
+		if (!RepackShmem->re_workerinfo[i].ri_in_use)
+			continue;
+
+		rworker = &RepackShmem->re_workerinfo[i];
+		if (rworker->ri_dbid == MyDatabaseId &&
+			(rworker->ri_relid == relid ||
+			 rworker->ri_toastrelid == relid))
+			retval = true;
+	}
+	LWLockRelease(RepackLock);
+
+	return retval;
+}
+
+/*
+ * Remove ourselves from the workerinfo array.
+ */
+static void
+RepackCleanup(RepackCleanupContext *context)
+{
+	if (context->concurrent)
+	{
+		RepackWorkerInfo *worker;
+
+		/*
+		 * The worker would normally terminate on its own when the work is
+		 * done, but make sure we signal it just in case.
+		 */
+		stop_repack_decoding_worker();
+
+		/*
+		 * also, make sure we stop advertising the relation we were repacking,
+		 * so that autovacuum reverts to handling it normally.
+		 */
+		LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+
+		worker = &RepackShmem->re_workerinfo[context->workerindex];
+		Assert(worker->ri_backendpid == MyProcPid);
+		worker->ri_in_use = false;
+		worker->ri_backendpid = 0;
+		worker->ri_dbid = InvalidOid;
+		worker->ri_relid = InvalidOid;
+		worker->ri_toastrelid = InvalidOid;
+		LWLockRelease(RepackLock);
+	}
+}
+
+/*
+ * RepackCleanup wrapped as an on_shmem_exit callback function
+ */
+static void
+RepackCleanupCb(int code, Datum arg)
+{
+	RepackCleanup((RepackCleanupContext *) DatumGetPointer(arg));
+}
+
+/*
+ * RepackShmemRequest
+ *		Register shared memory space needed for repack
+ */
+static void
+RepackShmemRequest(void *arg)
+{
+	Size		size;
+
+	/*
+	 * Need the fixed struct and the array of RepackWorkerInfo.
+	 */
+	size = sizeof(RepackShmemStruct);
+	size = MAXALIGN(size);
+	size = add_size(size, mul_size(max_repack_replication_slots,
+								   sizeof(RepackWorkerInfo)));
+
+	ShmemRequestStruct(.name = "Repack Data",
+					   .size = size,
+					   .ptr = (void **) &RepackShmem,
+		);
+}
+
+static void
+RepackShmemInit(void *arg)
+{
+	RepackWorkerInfo *reinfo;
+
+	reinfo = (RepackWorkerInfo *) ((char *) RepackShmem +
+								   MAXALIGN(sizeof(RepackShmemStruct)));
+
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		reinfo[i].ri_in_use = false;
+		reinfo[i].ri_backendpid = 0;
+		reinfo[i].ri_dbid = InvalidOid;
+		reinfo[i].ri_relid = InvalidOid;
+		reinfo[i].ri_toastrelid = InvalidOid;
+	}
+}
+
 /*
  * Check if the table (and its index) still meets the requirements of
  * cluster_rel().
diff --git a/src/backend/postmaster/autovacuum.c b/src/backend/postmaster/autovacuum.c
index bd626a16363..080c64ea3c8 100644
--- a/src/backend/postmaster/autovacuum.c
+++ b/src/backend/postmaster/autovacuum.c
@@ -78,6 +78,7 @@
 #include "catalog/namespace.h"
 #include "catalog/pg_database.h"
 #include "catalog/pg_namespace.h"
+#include "commands/repack.h"
 #include "commands/vacuum.h"
 #include "common/int.h"
 #include "funcapi.h"
@@ -2422,6 +2423,25 @@ do_autovacuum(void)
 			}
 		}
 		LWLockRelease(AutovacuumLock);
+
+		/*
+		 * Similarly, if the table is being processed by concurrent repack,
+		 * skip it (but make a note of that).  We wouldn't be able to acquire
+		 * its lock anyway.
+		 */
+		if (!skipit)
+		{
+			MemoryContextSwitchTo(PortalContext);
+
+			skipit = is_table_under_repack(MyDatabaseId, relid);
+			if (skipit)
+				ereport(LOG,
+						errmsg("skipping table \"%s.%s.%s\" because it's being repacked in concurrent mode",
+							   get_database_name(MyDatabaseId),
+							   get_namespace_name(get_rel_namespace(relid)),
+							   get_rel_name(relid)));
+		}
+
 		if (skipit)
 		{
 			LWLockRelease(AutovacuumScheduleLock);
diff --git a/src/backend/utils/activity/wait_event_names.txt b/src/backend/utils/activity/wait_event_names.txt
index 7bda5298558..e206304f204 100644
--- a/src/backend/utils/activity/wait_event_names.txt
+++ b/src/backend/utils/activity/wait_event_names.txt
@@ -332,6 +332,7 @@ SInvalWrite	"Waiting to add a message to the shared catalog invalidation queue."
 WALBufMapping	"Waiting to replace a page in WAL buffers."
 WALWrite	"Waiting for WAL buffers to be written to disk."
 ControlFile	"Waiting to read or update the <filename>pg_control</filename> file or create a new WAL file."
+Repack	"Waiting to read or update tables in process by concurrent repack."
 MultiXactGen	"Waiting to read or update shared multixact state."
 RelCacheInit	"Waiting to read or update a <filename>pg_internal.init</filename> relation cache initialization file."
 CheckpointerComm	"Waiting to manage fsync requests."
diff --git a/src/include/commands/repack.h b/src/include/commands/repack.h
index fd16e74b179..be7d38b5fae 100644
--- a/src/include/commands/repack.h
+++ b/src/include/commands/repack.h
@@ -42,6 +42,8 @@ extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel);
 
 extern void cluster_rel(RepackCommand command, Relation OldHeap, Oid indexOid,
 						ClusterParams *params, bool isTopLevel);
+extern bool is_table_under_repack(Oid databaseId, Oid relid);
+
 extern void check_index_is_clusterable(Relation OldHeap, Oid indexOid,
 									   LOCKMODE lockmode);
 extern void mark_index_clustered(Relation rel, Oid indexOid, bool is_internal);
diff --git a/src/include/storage/lwlocklist.h b/src/include/storage/lwlocklist.h
index af8553bcb6c..3f08f4a15d4 100644
--- a/src/include/storage/lwlocklist.h
+++ b/src/include/storage/lwlocklist.h
@@ -41,7 +41,7 @@ PG_LWLOCK(6, SInvalWrite)
 PG_LWLOCK(7, WALBufMapping)
 PG_LWLOCK(8, WALWrite)
 PG_LWLOCK(9, ControlFile)
-/* 10 was CheckpointLock */
+PG_LWLOCK(10, Repack)
 /* 11 was XactSLRULock */
 /* 12 was SubtransSLRULock */
 PG_LWLOCK(13, MultiXactGen)
diff --git a/src/include/storage/subsystemlist.h b/src/include/storage/subsystemlist.h
index 9ad619080be..4e683b8b0a8 100644
--- a/src/include/storage/subsystemlist.h
+++ b/src/include/storage/subsystemlist.h
@@ -72,6 +72,7 @@ PG_SHMEM_SUBSYSTEM(WalSummarizerShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(PgArchShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(ApplyLauncherShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(SlotSyncShmemCallbacks)
+PG_SHMEM_SUBSYSTEM(RepackShmemCallbacks)
 
 /* other modules that need some shared memory space */
 PG_SHMEM_SUBSYSTEM(BTreeShmemCallbacks)
diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list
index 637c669a146..d019e03aaf1 100644
--- a/src/tools/pgindent/typedefs.list
+++ b/src/tools/pgindent/typedefs.list
@@ -2639,9 +2639,12 @@ ReorderBufferTupleCidEnt
 ReorderBufferTupleCidKey
 ReorderBufferUpdateProgressTxnCB
 ReorderTuple
+RepackCleanupContext
 RepackCommand
 RepackDecodingState
+RepackShmemStruct
 RepackStmt
+RepackWorkerInfo
 ReparameterizeForeignPathByChild_function
 ReplOriginId
 ReplOriginXactState
-- 
2.47.3


--kdrcpfmkbkc4lqhu--





^ permalink  raw  reply  [nested|flat] 223+ messages in thread

* [PATCH 2/2] Publish list of tables being repacked in shared memory
@ 2026-04-07 20:29 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 223+ messages in thread

From: Álvaro Herrera @ 2026-04-07 20:29 UTC (permalink / raw)

Use it in autovacuum to skip processing tables that are being repacked.
This is mostly to avoid repeated attempts to process such tables, which
would fail due to the special deadlock checker behavior for repack.

Author: Álvaro Herrera <[email protected]>
Discussion: https://postgr.es/m/[email protected]
---
 src/backend/commands/repack.c                 | 195 ++++++++++++++++--
 src/backend/postmaster/autovacuum.c           |  20 ++
 .../utils/activity/wait_event_names.txt       |   1 +
 src/include/commands/repack.h                 |   2 +
 src/include/storage/lwlocklist.h              |   2 +-
 src/include/storage/subsystemlist.h           |   1 +
 src/tools/pgindent/typedefs.list              |   3 +
 7 files changed, 210 insertions(+), 14 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index a5f5df77291..ee7072dce6a 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -63,9 +63,11 @@
 #include "optimizer/optimizer.h"
 #include "pgstat.h"
 #include "storage/bufmgr.h"
+#include "storage/ipc.h"
 #include "storage/lmgr.h"
 #include "storage/predicate.h"
 #include "storage/proc.h"
+#include "storage/subsystems.h"
 #include "utils/acl.h"
 #include "utils/fmgroids.h"
 #include "utils/guc.h"
@@ -79,6 +81,32 @@
 #include "utils/syscache.h"
 #include "utils/wait_event_types.h"
 
+
+/* Shared memory layout for REPACK */
+typedef struct RepackWorkerInfo
+{
+	bool		ri_in_use;
+	pid_t		ri_backendpid;
+	Oid			ri_dbid;
+	Oid			ri_relid;
+	Oid			ri_toastrelid;
+} RepackWorkerInfo;
+
+typedef struct
+{
+	bool		re_useless;
+	RepackWorkerInfo re_workerinfo[FLEXIBLE_ARRAY_MEMBER];
+} RepackShmemStruct;
+
+static RepackShmemStruct *RepackShmem;
+
+typedef struct RepackCleanupContext
+{
+	bool		concurrent;
+	int			workerindex;
+} RepackCleanupContext;
+
+
 /*
  * This struct is used to pass around the information on tables to be
  * clustered. We need this so we can make a list of them when invoked without
@@ -90,6 +118,7 @@ typedef struct
 	Oid			indexOid;
 } RelToCluster;
 
+
 /*
  * The first file exported by the decoding worker must contain a snapshot, the
  * following ones contain the data changes.
@@ -166,6 +195,10 @@ static List *get_tables_to_repack_partitioned(RepackCommand cmd,
 											  MemoryContext permcxt);
 static bool repack_is_permitted_for_relation(RepackCommand cmd,
 											 Oid relid, Oid userid);
+static void RepackCleanup(RepackCleanupContext *context);
+static void RepackCleanupCb(int code, Datum arg);
+static void RepackShmemRequest(void *arg);
+static void RepackShmemInit(void *arg);
 
 static void apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt);
 static void apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
@@ -210,6 +243,11 @@ static void ProcessRepackMessage(StringInfo msg);
 static const char *RepackCommandAsString(RepackCommand cmd);
 
 
+const ShmemCallbacks RepackShmemCallbacks = {
+	.request_fn = RepackShmemRequest,
+	.init_fn = RepackShmemInit,
+};
+
 /*
  * The repack code allows for processing multiple tables at once. Because
  * of this, we cannot just run everything on a single transaction, or we
@@ -514,6 +552,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 	Oid			tableOid = RelationGetRelid(OldHeap);
 	Relation	index;
 	LOCKMODE	lmode;
+	RepackCleanupContext context;
 	Oid			save_userid;
 	int			save_sec_context;
 	int			save_nestlevel;
@@ -660,24 +699,43 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 		TransferPredicateLocksToHeapRelation(OldHeap);
 
 	/* rebuild_relation does all the dirty work */
-	PG_TRY();
-	{
-		rebuild_relation(OldHeap, index, verbose, ident_idx);
-	}
-	PG_FINALLY();
+	context.concurrent = concurrent;
+
+	PG_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
 	{
 		if (concurrent)
 		{
-			/*
-			 * Since during normal operation the worker was already asked to
-			 * exit, stopping it explicitly is especially important on ERROR.
-			 * However it still seems a good practice to make sure that the
-			 * worker never survives the REPACK command.
-			 */
-			stop_repack_decoding_worker();
+			bool		freefound = false;
+
+			LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+			for (int i = 0; i < max_repack_replication_slots; i++)
+			{
+				RepackWorkerInfo *worker;
+
+				if (RepackShmem->re_workerinfo[i].ri_in_use)
+					continue;
+
+				freefound = true;
+				worker = &RepackShmem->re_workerinfo[i];
+				context.workerindex = i;
+
+				worker->ri_in_use = true;
+				worker->ri_backendpid = MyProcPid;
+				worker->ri_dbid = MyDatabaseId;
+				worker->ri_relid = RelationGetRelid(OldHeap);
+				worker->ri_toastrelid = OldHeap->rd_rel->reltoastrelid;
+				break;
+			}
+			if (!freefound)
+				elog(ERROR, "could not find free repack entry");
+			LWLockRelease(RepackLock);
 		}
+
+		rebuild_relation(OldHeap, index, verbose, ident_idx);
 	}
-	PG_END_TRY();
+	PG_END_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
+
+	RepackCleanup(&context);
 
 	/* rebuild_relation closes OldHeap, and index if valid */
 
@@ -691,6 +749,117 @@ out:
 	pgstat_progress_end_command();
 }
 
+/*
+ * Return whether any backend is running concurrent REPACK on the given table
+ * (which could be a toast table).
+ */
+bool
+is_table_under_repack(Oid databaseId, Oid relid)
+{
+	bool		retval = false;
+
+	LWLockAcquire(RepackLock, LW_SHARED);
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		RepackWorkerInfo *rworker;
+
+		if (!RepackShmem->re_workerinfo[i].ri_in_use)
+			continue;
+
+		rworker = &RepackShmem->re_workerinfo[i];
+		if (rworker->ri_dbid == MyDatabaseId &&
+			(rworker->ri_relid == relid ||
+			 rworker->ri_toastrelid == relid))
+			retval = true;
+	}
+	LWLockRelease(RepackLock);
+
+	return retval;
+}
+
+/*
+ * Remove ourselves from the workerinfo array.
+ */
+static void
+RepackCleanup(RepackCleanupContext *context)
+{
+	if (context->concurrent)
+	{
+		RepackWorkerInfo *worker;
+
+		/*
+		 * The worker would normally terminate on its own when the work is
+		 * done, but make sure we signal it just in case.
+		 */
+		stop_repack_decoding_worker();
+
+		/*
+		 * also, make sure we stop advertising the relation we were repacking,
+		 * so that autovacuum reverts to handling it normally.
+		 */
+		LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+
+		worker = &RepackShmem->re_workerinfo[context->workerindex];
+		Assert(worker->ri_backendpid == MyProcPid);
+		worker->ri_in_use = false;
+		worker->ri_backendpid = 0;
+		worker->ri_dbid = InvalidOid;
+		worker->ri_relid = InvalidOid;
+		worker->ri_toastrelid = InvalidOid;
+		LWLockRelease(RepackLock);
+	}
+}
+
+/*
+ * RepackCleanup wrapped as an on_shmem_exit callback function
+ */
+static void
+RepackCleanupCb(int code, Datum arg)
+{
+	RepackCleanup((RepackCleanupContext *) DatumGetPointer(arg));
+}
+
+/*
+ * RepackShmemRequest
+ *		Register shared memory space needed for repack
+ */
+static void
+RepackShmemRequest(void *arg)
+{
+	Size		size;
+
+	/*
+	 * Need the fixed struct and the array of RepackWorkerInfo.
+	 */
+	size = sizeof(RepackShmemStruct);
+	size = MAXALIGN(size);
+	size = add_size(size, mul_size(max_repack_replication_slots,
+								   sizeof(RepackWorkerInfo)));
+
+	ShmemRequestStruct(.name = "Repack Data",
+					   .size = size,
+					   .ptr = (void **) &RepackShmem,
+		);
+}
+
+static void
+RepackShmemInit(void *arg)
+{
+	RepackWorkerInfo *reinfo;
+
+	reinfo = (RepackWorkerInfo *) ((char *) RepackShmem +
+								   MAXALIGN(sizeof(RepackShmemStruct)));
+
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		reinfo[i].ri_in_use = false;
+		reinfo[i].ri_backendpid = 0;
+		reinfo[i].ri_dbid = InvalidOid;
+		reinfo[i].ri_relid = InvalidOid;
+		reinfo[i].ri_toastrelid = InvalidOid;
+	}
+}
+
 /*
  * Check if the table (and its index) still meets the requirements of
  * cluster_rel().
diff --git a/src/backend/postmaster/autovacuum.c b/src/backend/postmaster/autovacuum.c
index bd626a16363..080c64ea3c8 100644
--- a/src/backend/postmaster/autovacuum.c
+++ b/src/backend/postmaster/autovacuum.c
@@ -78,6 +78,7 @@
 #include "catalog/namespace.h"
 #include "catalog/pg_database.h"
 #include "catalog/pg_namespace.h"
+#include "commands/repack.h"
 #include "commands/vacuum.h"
 #include "common/int.h"
 #include "funcapi.h"
@@ -2422,6 +2423,25 @@ do_autovacuum(void)
 			}
 		}
 		LWLockRelease(AutovacuumLock);
+
+		/*
+		 * Similarly, if the table is being processed by concurrent repack,
+		 * skip it (but make a note of that).  We wouldn't be able to acquire
+		 * its lock anyway.
+		 */
+		if (!skipit)
+		{
+			MemoryContextSwitchTo(PortalContext);
+
+			skipit = is_table_under_repack(MyDatabaseId, relid);
+			if (skipit)
+				ereport(LOG,
+						errmsg("skipping table \"%s.%s.%s\" because it's being repacked in concurrent mode",
+							   get_database_name(MyDatabaseId),
+							   get_namespace_name(get_rel_namespace(relid)),
+							   get_rel_name(relid)));
+		}
+
 		if (skipit)
 		{
 			LWLockRelease(AutovacuumScheduleLock);
diff --git a/src/backend/utils/activity/wait_event_names.txt b/src/backend/utils/activity/wait_event_names.txt
index 7bda5298558..e206304f204 100644
--- a/src/backend/utils/activity/wait_event_names.txt
+++ b/src/backend/utils/activity/wait_event_names.txt
@@ -332,6 +332,7 @@ SInvalWrite	"Waiting to add a message to the shared catalog invalidation queue."
 WALBufMapping	"Waiting to replace a page in WAL buffers."
 WALWrite	"Waiting for WAL buffers to be written to disk."
 ControlFile	"Waiting to read or update the <filename>pg_control</filename> file or create a new WAL file."
+Repack	"Waiting to read or update tables in process by concurrent repack."
 MultiXactGen	"Waiting to read or update shared multixact state."
 RelCacheInit	"Waiting to read or update a <filename>pg_internal.init</filename> relation cache initialization file."
 CheckpointerComm	"Waiting to manage fsync requests."
diff --git a/src/include/commands/repack.h b/src/include/commands/repack.h
index fd16e74b179..be7d38b5fae 100644
--- a/src/include/commands/repack.h
+++ b/src/include/commands/repack.h
@@ -42,6 +42,8 @@ extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel);
 
 extern void cluster_rel(RepackCommand command, Relation OldHeap, Oid indexOid,
 						ClusterParams *params, bool isTopLevel);
+extern bool is_table_under_repack(Oid databaseId, Oid relid);
+
 extern void check_index_is_clusterable(Relation OldHeap, Oid indexOid,
 									   LOCKMODE lockmode);
 extern void mark_index_clustered(Relation rel, Oid indexOid, bool is_internal);
diff --git a/src/include/storage/lwlocklist.h b/src/include/storage/lwlocklist.h
index af8553bcb6c..3f08f4a15d4 100644
--- a/src/include/storage/lwlocklist.h
+++ b/src/include/storage/lwlocklist.h
@@ -41,7 +41,7 @@ PG_LWLOCK(6, SInvalWrite)
 PG_LWLOCK(7, WALBufMapping)
 PG_LWLOCK(8, WALWrite)
 PG_LWLOCK(9, ControlFile)
-/* 10 was CheckpointLock */
+PG_LWLOCK(10, Repack)
 /* 11 was XactSLRULock */
 /* 12 was SubtransSLRULock */
 PG_LWLOCK(13, MultiXactGen)
diff --git a/src/include/storage/subsystemlist.h b/src/include/storage/subsystemlist.h
index 9ad619080be..4e683b8b0a8 100644
--- a/src/include/storage/subsystemlist.h
+++ b/src/include/storage/subsystemlist.h
@@ -72,6 +72,7 @@ PG_SHMEM_SUBSYSTEM(WalSummarizerShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(PgArchShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(ApplyLauncherShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(SlotSyncShmemCallbacks)
+PG_SHMEM_SUBSYSTEM(RepackShmemCallbacks)
 
 /* other modules that need some shared memory space */
 PG_SHMEM_SUBSYSTEM(BTreeShmemCallbacks)
diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list
index 637c669a146..d019e03aaf1 100644
--- a/src/tools/pgindent/typedefs.list
+++ b/src/tools/pgindent/typedefs.list
@@ -2639,9 +2639,12 @@ ReorderBufferTupleCidEnt
 ReorderBufferTupleCidKey
 ReorderBufferUpdateProgressTxnCB
 ReorderTuple
+RepackCleanupContext
 RepackCommand
 RepackDecodingState
+RepackShmemStruct
 RepackStmt
+RepackWorkerInfo
 ReparameterizeForeignPathByChild_function
 ReplOriginId
 ReplOriginXactState
-- 
2.47.3


--kdrcpfmkbkc4lqhu--





^ permalink  raw  reply  [nested|flat] 223+ messages in thread

* [PATCH 2/2] Publish list of tables being repacked in shared memory
@ 2026-04-07 20:29 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 223+ messages in thread

From: Álvaro Herrera @ 2026-04-07 20:29 UTC (permalink / raw)

Use it in autovacuum to skip processing tables that are being repacked.
This is mostly to avoid repeated attempts to process such tables, which
would fail due to the special deadlock checker behavior for repack.

Author: Álvaro Herrera <[email protected]>
Discussion: https://postgr.es/m/[email protected]
---
 src/backend/commands/repack.c                 | 195 ++++++++++++++++--
 src/backend/postmaster/autovacuum.c           |  20 ++
 .../utils/activity/wait_event_names.txt       |   1 +
 src/include/commands/repack.h                 |   2 +
 src/include/storage/lwlocklist.h              |   2 +-
 src/include/storage/subsystemlist.h           |   1 +
 src/tools/pgindent/typedefs.list              |   3 +
 7 files changed, 210 insertions(+), 14 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index a5f5df77291..ee7072dce6a 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -63,9 +63,11 @@
 #include "optimizer/optimizer.h"
 #include "pgstat.h"
 #include "storage/bufmgr.h"
+#include "storage/ipc.h"
 #include "storage/lmgr.h"
 #include "storage/predicate.h"
 #include "storage/proc.h"
+#include "storage/subsystems.h"
 #include "utils/acl.h"
 #include "utils/fmgroids.h"
 #include "utils/guc.h"
@@ -79,6 +81,32 @@
 #include "utils/syscache.h"
 #include "utils/wait_event_types.h"
 
+
+/* Shared memory layout for REPACK */
+typedef struct RepackWorkerInfo
+{
+	bool		ri_in_use;
+	pid_t		ri_backendpid;
+	Oid			ri_dbid;
+	Oid			ri_relid;
+	Oid			ri_toastrelid;
+} RepackWorkerInfo;
+
+typedef struct
+{
+	bool		re_useless;
+	RepackWorkerInfo re_workerinfo[FLEXIBLE_ARRAY_MEMBER];
+} RepackShmemStruct;
+
+static RepackShmemStruct *RepackShmem;
+
+typedef struct RepackCleanupContext
+{
+	bool		concurrent;
+	int			workerindex;
+} RepackCleanupContext;
+
+
 /*
  * This struct is used to pass around the information on tables to be
  * clustered. We need this so we can make a list of them when invoked without
@@ -90,6 +118,7 @@ typedef struct
 	Oid			indexOid;
 } RelToCluster;
 
+
 /*
  * The first file exported by the decoding worker must contain a snapshot, the
  * following ones contain the data changes.
@@ -166,6 +195,10 @@ static List *get_tables_to_repack_partitioned(RepackCommand cmd,
 											  MemoryContext permcxt);
 static bool repack_is_permitted_for_relation(RepackCommand cmd,
 											 Oid relid, Oid userid);
+static void RepackCleanup(RepackCleanupContext *context);
+static void RepackCleanupCb(int code, Datum arg);
+static void RepackShmemRequest(void *arg);
+static void RepackShmemInit(void *arg);
 
 static void apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt);
 static void apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
@@ -210,6 +243,11 @@ static void ProcessRepackMessage(StringInfo msg);
 static const char *RepackCommandAsString(RepackCommand cmd);
 
 
+const ShmemCallbacks RepackShmemCallbacks = {
+	.request_fn = RepackShmemRequest,
+	.init_fn = RepackShmemInit,
+};
+
 /*
  * The repack code allows for processing multiple tables at once. Because
  * of this, we cannot just run everything on a single transaction, or we
@@ -514,6 +552,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 	Oid			tableOid = RelationGetRelid(OldHeap);
 	Relation	index;
 	LOCKMODE	lmode;
+	RepackCleanupContext context;
 	Oid			save_userid;
 	int			save_sec_context;
 	int			save_nestlevel;
@@ -660,24 +699,43 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 		TransferPredicateLocksToHeapRelation(OldHeap);
 
 	/* rebuild_relation does all the dirty work */
-	PG_TRY();
-	{
-		rebuild_relation(OldHeap, index, verbose, ident_idx);
-	}
-	PG_FINALLY();
+	context.concurrent = concurrent;
+
+	PG_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
 	{
 		if (concurrent)
 		{
-			/*
-			 * Since during normal operation the worker was already asked to
-			 * exit, stopping it explicitly is especially important on ERROR.
-			 * However it still seems a good practice to make sure that the
-			 * worker never survives the REPACK command.
-			 */
-			stop_repack_decoding_worker();
+			bool		freefound = false;
+
+			LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+			for (int i = 0; i < max_repack_replication_slots; i++)
+			{
+				RepackWorkerInfo *worker;
+
+				if (RepackShmem->re_workerinfo[i].ri_in_use)
+					continue;
+
+				freefound = true;
+				worker = &RepackShmem->re_workerinfo[i];
+				context.workerindex = i;
+
+				worker->ri_in_use = true;
+				worker->ri_backendpid = MyProcPid;
+				worker->ri_dbid = MyDatabaseId;
+				worker->ri_relid = RelationGetRelid(OldHeap);
+				worker->ri_toastrelid = OldHeap->rd_rel->reltoastrelid;
+				break;
+			}
+			if (!freefound)
+				elog(ERROR, "could not find free repack entry");
+			LWLockRelease(RepackLock);
 		}
+
+		rebuild_relation(OldHeap, index, verbose, ident_idx);
 	}
-	PG_END_TRY();
+	PG_END_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
+
+	RepackCleanup(&context);
 
 	/* rebuild_relation closes OldHeap, and index if valid */
 
@@ -691,6 +749,117 @@ out:
 	pgstat_progress_end_command();
 }
 
+/*
+ * Return whether any backend is running concurrent REPACK on the given table
+ * (which could be a toast table).
+ */
+bool
+is_table_under_repack(Oid databaseId, Oid relid)
+{
+	bool		retval = false;
+
+	LWLockAcquire(RepackLock, LW_SHARED);
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		RepackWorkerInfo *rworker;
+
+		if (!RepackShmem->re_workerinfo[i].ri_in_use)
+			continue;
+
+		rworker = &RepackShmem->re_workerinfo[i];
+		if (rworker->ri_dbid == MyDatabaseId &&
+			(rworker->ri_relid == relid ||
+			 rworker->ri_toastrelid == relid))
+			retval = true;
+	}
+	LWLockRelease(RepackLock);
+
+	return retval;
+}
+
+/*
+ * Remove ourselves from the workerinfo array.
+ */
+static void
+RepackCleanup(RepackCleanupContext *context)
+{
+	if (context->concurrent)
+	{
+		RepackWorkerInfo *worker;
+
+		/*
+		 * The worker would normally terminate on its own when the work is
+		 * done, but make sure we signal it just in case.
+		 */
+		stop_repack_decoding_worker();
+
+		/*
+		 * also, make sure we stop advertising the relation we were repacking,
+		 * so that autovacuum reverts to handling it normally.
+		 */
+		LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+
+		worker = &RepackShmem->re_workerinfo[context->workerindex];
+		Assert(worker->ri_backendpid == MyProcPid);
+		worker->ri_in_use = false;
+		worker->ri_backendpid = 0;
+		worker->ri_dbid = InvalidOid;
+		worker->ri_relid = InvalidOid;
+		worker->ri_toastrelid = InvalidOid;
+		LWLockRelease(RepackLock);
+	}
+}
+
+/*
+ * RepackCleanup wrapped as an on_shmem_exit callback function
+ */
+static void
+RepackCleanupCb(int code, Datum arg)
+{
+	RepackCleanup((RepackCleanupContext *) DatumGetPointer(arg));
+}
+
+/*
+ * RepackShmemRequest
+ *		Register shared memory space needed for repack
+ */
+static void
+RepackShmemRequest(void *arg)
+{
+	Size		size;
+
+	/*
+	 * Need the fixed struct and the array of RepackWorkerInfo.
+	 */
+	size = sizeof(RepackShmemStruct);
+	size = MAXALIGN(size);
+	size = add_size(size, mul_size(max_repack_replication_slots,
+								   sizeof(RepackWorkerInfo)));
+
+	ShmemRequestStruct(.name = "Repack Data",
+					   .size = size,
+					   .ptr = (void **) &RepackShmem,
+		);
+}
+
+static void
+RepackShmemInit(void *arg)
+{
+	RepackWorkerInfo *reinfo;
+
+	reinfo = (RepackWorkerInfo *) ((char *) RepackShmem +
+								   MAXALIGN(sizeof(RepackShmemStruct)));
+
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		reinfo[i].ri_in_use = false;
+		reinfo[i].ri_backendpid = 0;
+		reinfo[i].ri_dbid = InvalidOid;
+		reinfo[i].ri_relid = InvalidOid;
+		reinfo[i].ri_toastrelid = InvalidOid;
+	}
+}
+
 /*
  * Check if the table (and its index) still meets the requirements of
  * cluster_rel().
diff --git a/src/backend/postmaster/autovacuum.c b/src/backend/postmaster/autovacuum.c
index bd626a16363..080c64ea3c8 100644
--- a/src/backend/postmaster/autovacuum.c
+++ b/src/backend/postmaster/autovacuum.c
@@ -78,6 +78,7 @@
 #include "catalog/namespace.h"
 #include "catalog/pg_database.h"
 #include "catalog/pg_namespace.h"
+#include "commands/repack.h"
 #include "commands/vacuum.h"
 #include "common/int.h"
 #include "funcapi.h"
@@ -2422,6 +2423,25 @@ do_autovacuum(void)
 			}
 		}
 		LWLockRelease(AutovacuumLock);
+
+		/*
+		 * Similarly, if the table is being processed by concurrent repack,
+		 * skip it (but make a note of that).  We wouldn't be able to acquire
+		 * its lock anyway.
+		 */
+		if (!skipit)
+		{
+			MemoryContextSwitchTo(PortalContext);
+
+			skipit = is_table_under_repack(MyDatabaseId, relid);
+			if (skipit)
+				ereport(LOG,
+						errmsg("skipping table \"%s.%s.%s\" because it's being repacked in concurrent mode",
+							   get_database_name(MyDatabaseId),
+							   get_namespace_name(get_rel_namespace(relid)),
+							   get_rel_name(relid)));
+		}
+
 		if (skipit)
 		{
 			LWLockRelease(AutovacuumScheduleLock);
diff --git a/src/backend/utils/activity/wait_event_names.txt b/src/backend/utils/activity/wait_event_names.txt
index 7bda5298558..e206304f204 100644
--- a/src/backend/utils/activity/wait_event_names.txt
+++ b/src/backend/utils/activity/wait_event_names.txt
@@ -332,6 +332,7 @@ SInvalWrite	"Waiting to add a message to the shared catalog invalidation queue."
 WALBufMapping	"Waiting to replace a page in WAL buffers."
 WALWrite	"Waiting for WAL buffers to be written to disk."
 ControlFile	"Waiting to read or update the <filename>pg_control</filename> file or create a new WAL file."
+Repack	"Waiting to read or update tables in process by concurrent repack."
 MultiXactGen	"Waiting to read or update shared multixact state."
 RelCacheInit	"Waiting to read or update a <filename>pg_internal.init</filename> relation cache initialization file."
 CheckpointerComm	"Waiting to manage fsync requests."
diff --git a/src/include/commands/repack.h b/src/include/commands/repack.h
index fd16e74b179..be7d38b5fae 100644
--- a/src/include/commands/repack.h
+++ b/src/include/commands/repack.h
@@ -42,6 +42,8 @@ extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel);
 
 extern void cluster_rel(RepackCommand command, Relation OldHeap, Oid indexOid,
 						ClusterParams *params, bool isTopLevel);
+extern bool is_table_under_repack(Oid databaseId, Oid relid);
+
 extern void check_index_is_clusterable(Relation OldHeap, Oid indexOid,
 									   LOCKMODE lockmode);
 extern void mark_index_clustered(Relation rel, Oid indexOid, bool is_internal);
diff --git a/src/include/storage/lwlocklist.h b/src/include/storage/lwlocklist.h
index af8553bcb6c..3f08f4a15d4 100644
--- a/src/include/storage/lwlocklist.h
+++ b/src/include/storage/lwlocklist.h
@@ -41,7 +41,7 @@ PG_LWLOCK(6, SInvalWrite)
 PG_LWLOCK(7, WALBufMapping)
 PG_LWLOCK(8, WALWrite)
 PG_LWLOCK(9, ControlFile)
-/* 10 was CheckpointLock */
+PG_LWLOCK(10, Repack)
 /* 11 was XactSLRULock */
 /* 12 was SubtransSLRULock */
 PG_LWLOCK(13, MultiXactGen)
diff --git a/src/include/storage/subsystemlist.h b/src/include/storage/subsystemlist.h
index 9ad619080be..4e683b8b0a8 100644
--- a/src/include/storage/subsystemlist.h
+++ b/src/include/storage/subsystemlist.h
@@ -72,6 +72,7 @@ PG_SHMEM_SUBSYSTEM(WalSummarizerShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(PgArchShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(ApplyLauncherShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(SlotSyncShmemCallbacks)
+PG_SHMEM_SUBSYSTEM(RepackShmemCallbacks)
 
 /* other modules that need some shared memory space */
 PG_SHMEM_SUBSYSTEM(BTreeShmemCallbacks)
diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list
index 637c669a146..d019e03aaf1 100644
--- a/src/tools/pgindent/typedefs.list
+++ b/src/tools/pgindent/typedefs.list
@@ -2639,9 +2639,12 @@ ReorderBufferTupleCidEnt
 ReorderBufferTupleCidKey
 ReorderBufferUpdateProgressTxnCB
 ReorderTuple
+RepackCleanupContext
 RepackCommand
 RepackDecodingState
+RepackShmemStruct
 RepackStmt
+RepackWorkerInfo
 ReparameterizeForeignPathByChild_function
 ReplOriginId
 ReplOriginXactState
-- 
2.47.3


--kdrcpfmkbkc4lqhu--





^ permalink  raw  reply  [nested|flat] 223+ messages in thread

* [PATCH 2/2] Publish list of tables being repacked in shared memory
@ 2026-04-07 20:29 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 223+ messages in thread

From: Álvaro Herrera @ 2026-04-07 20:29 UTC (permalink / raw)

Use it in autovacuum to skip processing tables that are being repacked.
This is mostly to avoid repeated attempts to process such tables, which
would fail due to the special deadlock checker behavior for repack.

Author: Álvaro Herrera <[email protected]>
Discussion: https://postgr.es/m/[email protected]
---
 src/backend/commands/repack.c                 | 195 ++++++++++++++++--
 src/backend/postmaster/autovacuum.c           |  20 ++
 .../utils/activity/wait_event_names.txt       |   1 +
 src/include/commands/repack.h                 |   2 +
 src/include/storage/lwlocklist.h              |   2 +-
 src/include/storage/subsystemlist.h           |   1 +
 src/tools/pgindent/typedefs.list              |   3 +
 7 files changed, 210 insertions(+), 14 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index a5f5df77291..ee7072dce6a 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -63,9 +63,11 @@
 #include "optimizer/optimizer.h"
 #include "pgstat.h"
 #include "storage/bufmgr.h"
+#include "storage/ipc.h"
 #include "storage/lmgr.h"
 #include "storage/predicate.h"
 #include "storage/proc.h"
+#include "storage/subsystems.h"
 #include "utils/acl.h"
 #include "utils/fmgroids.h"
 #include "utils/guc.h"
@@ -79,6 +81,32 @@
 #include "utils/syscache.h"
 #include "utils/wait_event_types.h"
 
+
+/* Shared memory layout for REPACK */
+typedef struct RepackWorkerInfo
+{
+	bool		ri_in_use;
+	pid_t		ri_backendpid;
+	Oid			ri_dbid;
+	Oid			ri_relid;
+	Oid			ri_toastrelid;
+} RepackWorkerInfo;
+
+typedef struct
+{
+	bool		re_useless;
+	RepackWorkerInfo re_workerinfo[FLEXIBLE_ARRAY_MEMBER];
+} RepackShmemStruct;
+
+static RepackShmemStruct *RepackShmem;
+
+typedef struct RepackCleanupContext
+{
+	bool		concurrent;
+	int			workerindex;
+} RepackCleanupContext;
+
+
 /*
  * This struct is used to pass around the information on tables to be
  * clustered. We need this so we can make a list of them when invoked without
@@ -90,6 +118,7 @@ typedef struct
 	Oid			indexOid;
 } RelToCluster;
 
+
 /*
  * The first file exported by the decoding worker must contain a snapshot, the
  * following ones contain the data changes.
@@ -166,6 +195,10 @@ static List *get_tables_to_repack_partitioned(RepackCommand cmd,
 											  MemoryContext permcxt);
 static bool repack_is_permitted_for_relation(RepackCommand cmd,
 											 Oid relid, Oid userid);
+static void RepackCleanup(RepackCleanupContext *context);
+static void RepackCleanupCb(int code, Datum arg);
+static void RepackShmemRequest(void *arg);
+static void RepackShmemInit(void *arg);
 
 static void apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt);
 static void apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
@@ -210,6 +243,11 @@ static void ProcessRepackMessage(StringInfo msg);
 static const char *RepackCommandAsString(RepackCommand cmd);
 
 
+const ShmemCallbacks RepackShmemCallbacks = {
+	.request_fn = RepackShmemRequest,
+	.init_fn = RepackShmemInit,
+};
+
 /*
  * The repack code allows for processing multiple tables at once. Because
  * of this, we cannot just run everything on a single transaction, or we
@@ -514,6 +552,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 	Oid			tableOid = RelationGetRelid(OldHeap);
 	Relation	index;
 	LOCKMODE	lmode;
+	RepackCleanupContext context;
 	Oid			save_userid;
 	int			save_sec_context;
 	int			save_nestlevel;
@@ -660,24 +699,43 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 		TransferPredicateLocksToHeapRelation(OldHeap);
 
 	/* rebuild_relation does all the dirty work */
-	PG_TRY();
-	{
-		rebuild_relation(OldHeap, index, verbose, ident_idx);
-	}
-	PG_FINALLY();
+	context.concurrent = concurrent;
+
+	PG_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
 	{
 		if (concurrent)
 		{
-			/*
-			 * Since during normal operation the worker was already asked to
-			 * exit, stopping it explicitly is especially important on ERROR.
-			 * However it still seems a good practice to make sure that the
-			 * worker never survives the REPACK command.
-			 */
-			stop_repack_decoding_worker();
+			bool		freefound = false;
+
+			LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+			for (int i = 0; i < max_repack_replication_slots; i++)
+			{
+				RepackWorkerInfo *worker;
+
+				if (RepackShmem->re_workerinfo[i].ri_in_use)
+					continue;
+
+				freefound = true;
+				worker = &RepackShmem->re_workerinfo[i];
+				context.workerindex = i;
+
+				worker->ri_in_use = true;
+				worker->ri_backendpid = MyProcPid;
+				worker->ri_dbid = MyDatabaseId;
+				worker->ri_relid = RelationGetRelid(OldHeap);
+				worker->ri_toastrelid = OldHeap->rd_rel->reltoastrelid;
+				break;
+			}
+			if (!freefound)
+				elog(ERROR, "could not find free repack entry");
+			LWLockRelease(RepackLock);
 		}
+
+		rebuild_relation(OldHeap, index, verbose, ident_idx);
 	}
-	PG_END_TRY();
+	PG_END_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
+
+	RepackCleanup(&context);
 
 	/* rebuild_relation closes OldHeap, and index if valid */
 
@@ -691,6 +749,117 @@ out:
 	pgstat_progress_end_command();
 }
 
+/*
+ * Return whether any backend is running concurrent REPACK on the given table
+ * (which could be a toast table).
+ */
+bool
+is_table_under_repack(Oid databaseId, Oid relid)
+{
+	bool		retval = false;
+
+	LWLockAcquire(RepackLock, LW_SHARED);
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		RepackWorkerInfo *rworker;
+
+		if (!RepackShmem->re_workerinfo[i].ri_in_use)
+			continue;
+
+		rworker = &RepackShmem->re_workerinfo[i];
+		if (rworker->ri_dbid == MyDatabaseId &&
+			(rworker->ri_relid == relid ||
+			 rworker->ri_toastrelid == relid))
+			retval = true;
+	}
+	LWLockRelease(RepackLock);
+
+	return retval;
+}
+
+/*
+ * Remove ourselves from the workerinfo array.
+ */
+static void
+RepackCleanup(RepackCleanupContext *context)
+{
+	if (context->concurrent)
+	{
+		RepackWorkerInfo *worker;
+
+		/*
+		 * The worker would normally terminate on its own when the work is
+		 * done, but make sure we signal it just in case.
+		 */
+		stop_repack_decoding_worker();
+
+		/*
+		 * also, make sure we stop advertising the relation we were repacking,
+		 * so that autovacuum reverts to handling it normally.
+		 */
+		LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+
+		worker = &RepackShmem->re_workerinfo[context->workerindex];
+		Assert(worker->ri_backendpid == MyProcPid);
+		worker->ri_in_use = false;
+		worker->ri_backendpid = 0;
+		worker->ri_dbid = InvalidOid;
+		worker->ri_relid = InvalidOid;
+		worker->ri_toastrelid = InvalidOid;
+		LWLockRelease(RepackLock);
+	}
+}
+
+/*
+ * RepackCleanup wrapped as an on_shmem_exit callback function
+ */
+static void
+RepackCleanupCb(int code, Datum arg)
+{
+	RepackCleanup((RepackCleanupContext *) DatumGetPointer(arg));
+}
+
+/*
+ * RepackShmemRequest
+ *		Register shared memory space needed for repack
+ */
+static void
+RepackShmemRequest(void *arg)
+{
+	Size		size;
+
+	/*
+	 * Need the fixed struct and the array of RepackWorkerInfo.
+	 */
+	size = sizeof(RepackShmemStruct);
+	size = MAXALIGN(size);
+	size = add_size(size, mul_size(max_repack_replication_slots,
+								   sizeof(RepackWorkerInfo)));
+
+	ShmemRequestStruct(.name = "Repack Data",
+					   .size = size,
+					   .ptr = (void **) &RepackShmem,
+		);
+}
+
+static void
+RepackShmemInit(void *arg)
+{
+	RepackWorkerInfo *reinfo;
+
+	reinfo = (RepackWorkerInfo *) ((char *) RepackShmem +
+								   MAXALIGN(sizeof(RepackShmemStruct)));
+
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		reinfo[i].ri_in_use = false;
+		reinfo[i].ri_backendpid = 0;
+		reinfo[i].ri_dbid = InvalidOid;
+		reinfo[i].ri_relid = InvalidOid;
+		reinfo[i].ri_toastrelid = InvalidOid;
+	}
+}
+
 /*
  * Check if the table (and its index) still meets the requirements of
  * cluster_rel().
diff --git a/src/backend/postmaster/autovacuum.c b/src/backend/postmaster/autovacuum.c
index bd626a16363..080c64ea3c8 100644
--- a/src/backend/postmaster/autovacuum.c
+++ b/src/backend/postmaster/autovacuum.c
@@ -78,6 +78,7 @@
 #include "catalog/namespace.h"
 #include "catalog/pg_database.h"
 #include "catalog/pg_namespace.h"
+#include "commands/repack.h"
 #include "commands/vacuum.h"
 #include "common/int.h"
 #include "funcapi.h"
@@ -2422,6 +2423,25 @@ do_autovacuum(void)
 			}
 		}
 		LWLockRelease(AutovacuumLock);
+
+		/*
+		 * Similarly, if the table is being processed by concurrent repack,
+		 * skip it (but make a note of that).  We wouldn't be able to acquire
+		 * its lock anyway.
+		 */
+		if (!skipit)
+		{
+			MemoryContextSwitchTo(PortalContext);
+
+			skipit = is_table_under_repack(MyDatabaseId, relid);
+			if (skipit)
+				ereport(LOG,
+						errmsg("skipping table \"%s.%s.%s\" because it's being repacked in concurrent mode",
+							   get_database_name(MyDatabaseId),
+							   get_namespace_name(get_rel_namespace(relid)),
+							   get_rel_name(relid)));
+		}
+
 		if (skipit)
 		{
 			LWLockRelease(AutovacuumScheduleLock);
diff --git a/src/backend/utils/activity/wait_event_names.txt b/src/backend/utils/activity/wait_event_names.txt
index 7bda5298558..e206304f204 100644
--- a/src/backend/utils/activity/wait_event_names.txt
+++ b/src/backend/utils/activity/wait_event_names.txt
@@ -332,6 +332,7 @@ SInvalWrite	"Waiting to add a message to the shared catalog invalidation queue."
 WALBufMapping	"Waiting to replace a page in WAL buffers."
 WALWrite	"Waiting for WAL buffers to be written to disk."
 ControlFile	"Waiting to read or update the <filename>pg_control</filename> file or create a new WAL file."
+Repack	"Waiting to read or update tables in process by concurrent repack."
 MultiXactGen	"Waiting to read or update shared multixact state."
 RelCacheInit	"Waiting to read or update a <filename>pg_internal.init</filename> relation cache initialization file."
 CheckpointerComm	"Waiting to manage fsync requests."
diff --git a/src/include/commands/repack.h b/src/include/commands/repack.h
index fd16e74b179..be7d38b5fae 100644
--- a/src/include/commands/repack.h
+++ b/src/include/commands/repack.h
@@ -42,6 +42,8 @@ extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel);
 
 extern void cluster_rel(RepackCommand command, Relation OldHeap, Oid indexOid,
 						ClusterParams *params, bool isTopLevel);
+extern bool is_table_under_repack(Oid databaseId, Oid relid);
+
 extern void check_index_is_clusterable(Relation OldHeap, Oid indexOid,
 									   LOCKMODE lockmode);
 extern void mark_index_clustered(Relation rel, Oid indexOid, bool is_internal);
diff --git a/src/include/storage/lwlocklist.h b/src/include/storage/lwlocklist.h
index af8553bcb6c..3f08f4a15d4 100644
--- a/src/include/storage/lwlocklist.h
+++ b/src/include/storage/lwlocklist.h
@@ -41,7 +41,7 @@ PG_LWLOCK(6, SInvalWrite)
 PG_LWLOCK(7, WALBufMapping)
 PG_LWLOCK(8, WALWrite)
 PG_LWLOCK(9, ControlFile)
-/* 10 was CheckpointLock */
+PG_LWLOCK(10, Repack)
 /* 11 was XactSLRULock */
 /* 12 was SubtransSLRULock */
 PG_LWLOCK(13, MultiXactGen)
diff --git a/src/include/storage/subsystemlist.h b/src/include/storage/subsystemlist.h
index 9ad619080be..4e683b8b0a8 100644
--- a/src/include/storage/subsystemlist.h
+++ b/src/include/storage/subsystemlist.h
@@ -72,6 +72,7 @@ PG_SHMEM_SUBSYSTEM(WalSummarizerShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(PgArchShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(ApplyLauncherShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(SlotSyncShmemCallbacks)
+PG_SHMEM_SUBSYSTEM(RepackShmemCallbacks)
 
 /* other modules that need some shared memory space */
 PG_SHMEM_SUBSYSTEM(BTreeShmemCallbacks)
diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list
index 637c669a146..d019e03aaf1 100644
--- a/src/tools/pgindent/typedefs.list
+++ b/src/tools/pgindent/typedefs.list
@@ -2639,9 +2639,12 @@ ReorderBufferTupleCidEnt
 ReorderBufferTupleCidKey
 ReorderBufferUpdateProgressTxnCB
 ReorderTuple
+RepackCleanupContext
 RepackCommand
 RepackDecodingState
+RepackShmemStruct
 RepackStmt
+RepackWorkerInfo
 ReparameterizeForeignPathByChild_function
 ReplOriginId
 ReplOriginXactState
-- 
2.47.3


--kdrcpfmkbkc4lqhu--





^ permalink  raw  reply  [nested|flat] 223+ messages in thread

* [PATCH 2/2] Publish list of tables being repacked in shared memory
@ 2026-04-07 20:29 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 223+ messages in thread

From: Álvaro Herrera @ 2026-04-07 20:29 UTC (permalink / raw)

Use it in autovacuum to skip processing tables that are being repacked.
This is mostly to avoid repeated attempts to process such tables, which
would fail due to the special deadlock checker behavior for repack.

Author: Álvaro Herrera <[email protected]>
Discussion: https://postgr.es/m/[email protected]
---
 src/backend/commands/repack.c                 | 195 ++++++++++++++++--
 src/backend/postmaster/autovacuum.c           |  20 ++
 .../utils/activity/wait_event_names.txt       |   1 +
 src/include/commands/repack.h                 |   2 +
 src/include/storage/lwlocklist.h              |   2 +-
 src/include/storage/subsystemlist.h           |   1 +
 src/tools/pgindent/typedefs.list              |   3 +
 7 files changed, 210 insertions(+), 14 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index a5f5df77291..ee7072dce6a 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -63,9 +63,11 @@
 #include "optimizer/optimizer.h"
 #include "pgstat.h"
 #include "storage/bufmgr.h"
+#include "storage/ipc.h"
 #include "storage/lmgr.h"
 #include "storage/predicate.h"
 #include "storage/proc.h"
+#include "storage/subsystems.h"
 #include "utils/acl.h"
 #include "utils/fmgroids.h"
 #include "utils/guc.h"
@@ -79,6 +81,32 @@
 #include "utils/syscache.h"
 #include "utils/wait_event_types.h"
 
+
+/* Shared memory layout for REPACK */
+typedef struct RepackWorkerInfo
+{
+	bool		ri_in_use;
+	pid_t		ri_backendpid;
+	Oid			ri_dbid;
+	Oid			ri_relid;
+	Oid			ri_toastrelid;
+} RepackWorkerInfo;
+
+typedef struct
+{
+	bool		re_useless;
+	RepackWorkerInfo re_workerinfo[FLEXIBLE_ARRAY_MEMBER];
+} RepackShmemStruct;
+
+static RepackShmemStruct *RepackShmem;
+
+typedef struct RepackCleanupContext
+{
+	bool		concurrent;
+	int			workerindex;
+} RepackCleanupContext;
+
+
 /*
  * This struct is used to pass around the information on tables to be
  * clustered. We need this so we can make a list of them when invoked without
@@ -90,6 +118,7 @@ typedef struct
 	Oid			indexOid;
 } RelToCluster;
 
+
 /*
  * The first file exported by the decoding worker must contain a snapshot, the
  * following ones contain the data changes.
@@ -166,6 +195,10 @@ static List *get_tables_to_repack_partitioned(RepackCommand cmd,
 											  MemoryContext permcxt);
 static bool repack_is_permitted_for_relation(RepackCommand cmd,
 											 Oid relid, Oid userid);
+static void RepackCleanup(RepackCleanupContext *context);
+static void RepackCleanupCb(int code, Datum arg);
+static void RepackShmemRequest(void *arg);
+static void RepackShmemInit(void *arg);
 
 static void apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt);
 static void apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
@@ -210,6 +243,11 @@ static void ProcessRepackMessage(StringInfo msg);
 static const char *RepackCommandAsString(RepackCommand cmd);
 
 
+const ShmemCallbacks RepackShmemCallbacks = {
+	.request_fn = RepackShmemRequest,
+	.init_fn = RepackShmemInit,
+};
+
 /*
  * The repack code allows for processing multiple tables at once. Because
  * of this, we cannot just run everything on a single transaction, or we
@@ -514,6 +552,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 	Oid			tableOid = RelationGetRelid(OldHeap);
 	Relation	index;
 	LOCKMODE	lmode;
+	RepackCleanupContext context;
 	Oid			save_userid;
 	int			save_sec_context;
 	int			save_nestlevel;
@@ -660,24 +699,43 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 		TransferPredicateLocksToHeapRelation(OldHeap);
 
 	/* rebuild_relation does all the dirty work */
-	PG_TRY();
-	{
-		rebuild_relation(OldHeap, index, verbose, ident_idx);
-	}
-	PG_FINALLY();
+	context.concurrent = concurrent;
+
+	PG_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
 	{
 		if (concurrent)
 		{
-			/*
-			 * Since during normal operation the worker was already asked to
-			 * exit, stopping it explicitly is especially important on ERROR.
-			 * However it still seems a good practice to make sure that the
-			 * worker never survives the REPACK command.
-			 */
-			stop_repack_decoding_worker();
+			bool		freefound = false;
+
+			LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+			for (int i = 0; i < max_repack_replication_slots; i++)
+			{
+				RepackWorkerInfo *worker;
+
+				if (RepackShmem->re_workerinfo[i].ri_in_use)
+					continue;
+
+				freefound = true;
+				worker = &RepackShmem->re_workerinfo[i];
+				context.workerindex = i;
+
+				worker->ri_in_use = true;
+				worker->ri_backendpid = MyProcPid;
+				worker->ri_dbid = MyDatabaseId;
+				worker->ri_relid = RelationGetRelid(OldHeap);
+				worker->ri_toastrelid = OldHeap->rd_rel->reltoastrelid;
+				break;
+			}
+			if (!freefound)
+				elog(ERROR, "could not find free repack entry");
+			LWLockRelease(RepackLock);
 		}
+
+		rebuild_relation(OldHeap, index, verbose, ident_idx);
 	}
-	PG_END_TRY();
+	PG_END_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
+
+	RepackCleanup(&context);
 
 	/* rebuild_relation closes OldHeap, and index if valid */
 
@@ -691,6 +749,117 @@ out:
 	pgstat_progress_end_command();
 }
 
+/*
+ * Return whether any backend is running concurrent REPACK on the given table
+ * (which could be a toast table).
+ */
+bool
+is_table_under_repack(Oid databaseId, Oid relid)
+{
+	bool		retval = false;
+
+	LWLockAcquire(RepackLock, LW_SHARED);
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		RepackWorkerInfo *rworker;
+
+		if (!RepackShmem->re_workerinfo[i].ri_in_use)
+			continue;
+
+		rworker = &RepackShmem->re_workerinfo[i];
+		if (rworker->ri_dbid == MyDatabaseId &&
+			(rworker->ri_relid == relid ||
+			 rworker->ri_toastrelid == relid))
+			retval = true;
+	}
+	LWLockRelease(RepackLock);
+
+	return retval;
+}
+
+/*
+ * Remove ourselves from the workerinfo array.
+ */
+static void
+RepackCleanup(RepackCleanupContext *context)
+{
+	if (context->concurrent)
+	{
+		RepackWorkerInfo *worker;
+
+		/*
+		 * The worker would normally terminate on its own when the work is
+		 * done, but make sure we signal it just in case.
+		 */
+		stop_repack_decoding_worker();
+
+		/*
+		 * also, make sure we stop advertising the relation we were repacking,
+		 * so that autovacuum reverts to handling it normally.
+		 */
+		LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+
+		worker = &RepackShmem->re_workerinfo[context->workerindex];
+		Assert(worker->ri_backendpid == MyProcPid);
+		worker->ri_in_use = false;
+		worker->ri_backendpid = 0;
+		worker->ri_dbid = InvalidOid;
+		worker->ri_relid = InvalidOid;
+		worker->ri_toastrelid = InvalidOid;
+		LWLockRelease(RepackLock);
+	}
+}
+
+/*
+ * RepackCleanup wrapped as an on_shmem_exit callback function
+ */
+static void
+RepackCleanupCb(int code, Datum arg)
+{
+	RepackCleanup((RepackCleanupContext *) DatumGetPointer(arg));
+}
+
+/*
+ * RepackShmemRequest
+ *		Register shared memory space needed for repack
+ */
+static void
+RepackShmemRequest(void *arg)
+{
+	Size		size;
+
+	/*
+	 * Need the fixed struct and the array of RepackWorkerInfo.
+	 */
+	size = sizeof(RepackShmemStruct);
+	size = MAXALIGN(size);
+	size = add_size(size, mul_size(max_repack_replication_slots,
+								   sizeof(RepackWorkerInfo)));
+
+	ShmemRequestStruct(.name = "Repack Data",
+					   .size = size,
+					   .ptr = (void **) &RepackShmem,
+		);
+}
+
+static void
+RepackShmemInit(void *arg)
+{
+	RepackWorkerInfo *reinfo;
+
+	reinfo = (RepackWorkerInfo *) ((char *) RepackShmem +
+								   MAXALIGN(sizeof(RepackShmemStruct)));
+
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		reinfo[i].ri_in_use = false;
+		reinfo[i].ri_backendpid = 0;
+		reinfo[i].ri_dbid = InvalidOid;
+		reinfo[i].ri_relid = InvalidOid;
+		reinfo[i].ri_toastrelid = InvalidOid;
+	}
+}
+
 /*
  * Check if the table (and its index) still meets the requirements of
  * cluster_rel().
diff --git a/src/backend/postmaster/autovacuum.c b/src/backend/postmaster/autovacuum.c
index bd626a16363..080c64ea3c8 100644
--- a/src/backend/postmaster/autovacuum.c
+++ b/src/backend/postmaster/autovacuum.c
@@ -78,6 +78,7 @@
 #include "catalog/namespace.h"
 #include "catalog/pg_database.h"
 #include "catalog/pg_namespace.h"
+#include "commands/repack.h"
 #include "commands/vacuum.h"
 #include "common/int.h"
 #include "funcapi.h"
@@ -2422,6 +2423,25 @@ do_autovacuum(void)
 			}
 		}
 		LWLockRelease(AutovacuumLock);
+
+		/*
+		 * Similarly, if the table is being processed by concurrent repack,
+		 * skip it (but make a note of that).  We wouldn't be able to acquire
+		 * its lock anyway.
+		 */
+		if (!skipit)
+		{
+			MemoryContextSwitchTo(PortalContext);
+
+			skipit = is_table_under_repack(MyDatabaseId, relid);
+			if (skipit)
+				ereport(LOG,
+						errmsg("skipping table \"%s.%s.%s\" because it's being repacked in concurrent mode",
+							   get_database_name(MyDatabaseId),
+							   get_namespace_name(get_rel_namespace(relid)),
+							   get_rel_name(relid)));
+		}
+
 		if (skipit)
 		{
 			LWLockRelease(AutovacuumScheduleLock);
diff --git a/src/backend/utils/activity/wait_event_names.txt b/src/backend/utils/activity/wait_event_names.txt
index 7bda5298558..e206304f204 100644
--- a/src/backend/utils/activity/wait_event_names.txt
+++ b/src/backend/utils/activity/wait_event_names.txt
@@ -332,6 +332,7 @@ SInvalWrite	"Waiting to add a message to the shared catalog invalidation queue."
 WALBufMapping	"Waiting to replace a page in WAL buffers."
 WALWrite	"Waiting for WAL buffers to be written to disk."
 ControlFile	"Waiting to read or update the <filename>pg_control</filename> file or create a new WAL file."
+Repack	"Waiting to read or update tables in process by concurrent repack."
 MultiXactGen	"Waiting to read or update shared multixact state."
 RelCacheInit	"Waiting to read or update a <filename>pg_internal.init</filename> relation cache initialization file."
 CheckpointerComm	"Waiting to manage fsync requests."
diff --git a/src/include/commands/repack.h b/src/include/commands/repack.h
index fd16e74b179..be7d38b5fae 100644
--- a/src/include/commands/repack.h
+++ b/src/include/commands/repack.h
@@ -42,6 +42,8 @@ extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel);
 
 extern void cluster_rel(RepackCommand command, Relation OldHeap, Oid indexOid,
 						ClusterParams *params, bool isTopLevel);
+extern bool is_table_under_repack(Oid databaseId, Oid relid);
+
 extern void check_index_is_clusterable(Relation OldHeap, Oid indexOid,
 									   LOCKMODE lockmode);
 extern void mark_index_clustered(Relation rel, Oid indexOid, bool is_internal);
diff --git a/src/include/storage/lwlocklist.h b/src/include/storage/lwlocklist.h
index af8553bcb6c..3f08f4a15d4 100644
--- a/src/include/storage/lwlocklist.h
+++ b/src/include/storage/lwlocklist.h
@@ -41,7 +41,7 @@ PG_LWLOCK(6, SInvalWrite)
 PG_LWLOCK(7, WALBufMapping)
 PG_LWLOCK(8, WALWrite)
 PG_LWLOCK(9, ControlFile)
-/* 10 was CheckpointLock */
+PG_LWLOCK(10, Repack)
 /* 11 was XactSLRULock */
 /* 12 was SubtransSLRULock */
 PG_LWLOCK(13, MultiXactGen)
diff --git a/src/include/storage/subsystemlist.h b/src/include/storage/subsystemlist.h
index 9ad619080be..4e683b8b0a8 100644
--- a/src/include/storage/subsystemlist.h
+++ b/src/include/storage/subsystemlist.h
@@ -72,6 +72,7 @@ PG_SHMEM_SUBSYSTEM(WalSummarizerShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(PgArchShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(ApplyLauncherShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(SlotSyncShmemCallbacks)
+PG_SHMEM_SUBSYSTEM(RepackShmemCallbacks)
 
 /* other modules that need some shared memory space */
 PG_SHMEM_SUBSYSTEM(BTreeShmemCallbacks)
diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list
index 637c669a146..d019e03aaf1 100644
--- a/src/tools/pgindent/typedefs.list
+++ b/src/tools/pgindent/typedefs.list
@@ -2639,9 +2639,12 @@ ReorderBufferTupleCidEnt
 ReorderBufferTupleCidKey
 ReorderBufferUpdateProgressTxnCB
 ReorderTuple
+RepackCleanupContext
 RepackCommand
 RepackDecodingState
+RepackShmemStruct
 RepackStmt
+RepackWorkerInfo
 ReparameterizeForeignPathByChild_function
 ReplOriginId
 ReplOriginXactState
-- 
2.47.3


--kdrcpfmkbkc4lqhu--





^ permalink  raw  reply  [nested|flat] 223+ messages in thread

* [PATCH 2/2] Publish list of tables being repacked in shared memory
@ 2026-04-07 20:29 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 223+ messages in thread

From: Álvaro Herrera @ 2026-04-07 20:29 UTC (permalink / raw)

Use it in autovacuum to skip processing tables that are being repacked.
This is mostly to avoid repeated attempts to process such tables, which
would fail due to the special deadlock checker behavior for repack.

Author: Álvaro Herrera <[email protected]>
Discussion: https://postgr.es/m/[email protected]
---
 src/backend/commands/repack.c                 | 195 ++++++++++++++++--
 src/backend/postmaster/autovacuum.c           |  20 ++
 .../utils/activity/wait_event_names.txt       |   1 +
 src/include/commands/repack.h                 |   2 +
 src/include/storage/lwlocklist.h              |   2 +-
 src/include/storage/subsystemlist.h           |   1 +
 src/tools/pgindent/typedefs.list              |   3 +
 7 files changed, 210 insertions(+), 14 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index a5f5df77291..ee7072dce6a 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -63,9 +63,11 @@
 #include "optimizer/optimizer.h"
 #include "pgstat.h"
 #include "storage/bufmgr.h"
+#include "storage/ipc.h"
 #include "storage/lmgr.h"
 #include "storage/predicate.h"
 #include "storage/proc.h"
+#include "storage/subsystems.h"
 #include "utils/acl.h"
 #include "utils/fmgroids.h"
 #include "utils/guc.h"
@@ -79,6 +81,32 @@
 #include "utils/syscache.h"
 #include "utils/wait_event_types.h"
 
+
+/* Shared memory layout for REPACK */
+typedef struct RepackWorkerInfo
+{
+	bool		ri_in_use;
+	pid_t		ri_backendpid;
+	Oid			ri_dbid;
+	Oid			ri_relid;
+	Oid			ri_toastrelid;
+} RepackWorkerInfo;
+
+typedef struct
+{
+	bool		re_useless;
+	RepackWorkerInfo re_workerinfo[FLEXIBLE_ARRAY_MEMBER];
+} RepackShmemStruct;
+
+static RepackShmemStruct *RepackShmem;
+
+typedef struct RepackCleanupContext
+{
+	bool		concurrent;
+	int			workerindex;
+} RepackCleanupContext;
+
+
 /*
  * This struct is used to pass around the information on tables to be
  * clustered. We need this so we can make a list of them when invoked without
@@ -90,6 +118,7 @@ typedef struct
 	Oid			indexOid;
 } RelToCluster;
 
+
 /*
  * The first file exported by the decoding worker must contain a snapshot, the
  * following ones contain the data changes.
@@ -166,6 +195,10 @@ static List *get_tables_to_repack_partitioned(RepackCommand cmd,
 											  MemoryContext permcxt);
 static bool repack_is_permitted_for_relation(RepackCommand cmd,
 											 Oid relid, Oid userid);
+static void RepackCleanup(RepackCleanupContext *context);
+static void RepackCleanupCb(int code, Datum arg);
+static void RepackShmemRequest(void *arg);
+static void RepackShmemInit(void *arg);
 
 static void apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt);
 static void apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
@@ -210,6 +243,11 @@ static void ProcessRepackMessage(StringInfo msg);
 static const char *RepackCommandAsString(RepackCommand cmd);
 
 
+const ShmemCallbacks RepackShmemCallbacks = {
+	.request_fn = RepackShmemRequest,
+	.init_fn = RepackShmemInit,
+};
+
 /*
  * The repack code allows for processing multiple tables at once. Because
  * of this, we cannot just run everything on a single transaction, or we
@@ -514,6 +552,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 	Oid			tableOid = RelationGetRelid(OldHeap);
 	Relation	index;
 	LOCKMODE	lmode;
+	RepackCleanupContext context;
 	Oid			save_userid;
 	int			save_sec_context;
 	int			save_nestlevel;
@@ -660,24 +699,43 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 		TransferPredicateLocksToHeapRelation(OldHeap);
 
 	/* rebuild_relation does all the dirty work */
-	PG_TRY();
-	{
-		rebuild_relation(OldHeap, index, verbose, ident_idx);
-	}
-	PG_FINALLY();
+	context.concurrent = concurrent;
+
+	PG_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
 	{
 		if (concurrent)
 		{
-			/*
-			 * Since during normal operation the worker was already asked to
-			 * exit, stopping it explicitly is especially important on ERROR.
-			 * However it still seems a good practice to make sure that the
-			 * worker never survives the REPACK command.
-			 */
-			stop_repack_decoding_worker();
+			bool		freefound = false;
+
+			LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+			for (int i = 0; i < max_repack_replication_slots; i++)
+			{
+				RepackWorkerInfo *worker;
+
+				if (RepackShmem->re_workerinfo[i].ri_in_use)
+					continue;
+
+				freefound = true;
+				worker = &RepackShmem->re_workerinfo[i];
+				context.workerindex = i;
+
+				worker->ri_in_use = true;
+				worker->ri_backendpid = MyProcPid;
+				worker->ri_dbid = MyDatabaseId;
+				worker->ri_relid = RelationGetRelid(OldHeap);
+				worker->ri_toastrelid = OldHeap->rd_rel->reltoastrelid;
+				break;
+			}
+			if (!freefound)
+				elog(ERROR, "could not find free repack entry");
+			LWLockRelease(RepackLock);
 		}
+
+		rebuild_relation(OldHeap, index, verbose, ident_idx);
 	}
-	PG_END_TRY();
+	PG_END_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
+
+	RepackCleanup(&context);
 
 	/* rebuild_relation closes OldHeap, and index if valid */
 
@@ -691,6 +749,117 @@ out:
 	pgstat_progress_end_command();
 }
 
+/*
+ * Return whether any backend is running concurrent REPACK on the given table
+ * (which could be a toast table).
+ */
+bool
+is_table_under_repack(Oid databaseId, Oid relid)
+{
+	bool		retval = false;
+
+	LWLockAcquire(RepackLock, LW_SHARED);
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		RepackWorkerInfo *rworker;
+
+		if (!RepackShmem->re_workerinfo[i].ri_in_use)
+			continue;
+
+		rworker = &RepackShmem->re_workerinfo[i];
+		if (rworker->ri_dbid == MyDatabaseId &&
+			(rworker->ri_relid == relid ||
+			 rworker->ri_toastrelid == relid))
+			retval = true;
+	}
+	LWLockRelease(RepackLock);
+
+	return retval;
+}
+
+/*
+ * Remove ourselves from the workerinfo array.
+ */
+static void
+RepackCleanup(RepackCleanupContext *context)
+{
+	if (context->concurrent)
+	{
+		RepackWorkerInfo *worker;
+
+		/*
+		 * The worker would normally terminate on its own when the work is
+		 * done, but make sure we signal it just in case.
+		 */
+		stop_repack_decoding_worker();
+
+		/*
+		 * also, make sure we stop advertising the relation we were repacking,
+		 * so that autovacuum reverts to handling it normally.
+		 */
+		LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+
+		worker = &RepackShmem->re_workerinfo[context->workerindex];
+		Assert(worker->ri_backendpid == MyProcPid);
+		worker->ri_in_use = false;
+		worker->ri_backendpid = 0;
+		worker->ri_dbid = InvalidOid;
+		worker->ri_relid = InvalidOid;
+		worker->ri_toastrelid = InvalidOid;
+		LWLockRelease(RepackLock);
+	}
+}
+
+/*
+ * RepackCleanup wrapped as an on_shmem_exit callback function
+ */
+static void
+RepackCleanupCb(int code, Datum arg)
+{
+	RepackCleanup((RepackCleanupContext *) DatumGetPointer(arg));
+}
+
+/*
+ * RepackShmemRequest
+ *		Register shared memory space needed for repack
+ */
+static void
+RepackShmemRequest(void *arg)
+{
+	Size		size;
+
+	/*
+	 * Need the fixed struct and the array of RepackWorkerInfo.
+	 */
+	size = sizeof(RepackShmemStruct);
+	size = MAXALIGN(size);
+	size = add_size(size, mul_size(max_repack_replication_slots,
+								   sizeof(RepackWorkerInfo)));
+
+	ShmemRequestStruct(.name = "Repack Data",
+					   .size = size,
+					   .ptr = (void **) &RepackShmem,
+		);
+}
+
+static void
+RepackShmemInit(void *arg)
+{
+	RepackWorkerInfo *reinfo;
+
+	reinfo = (RepackWorkerInfo *) ((char *) RepackShmem +
+								   MAXALIGN(sizeof(RepackShmemStruct)));
+
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		reinfo[i].ri_in_use = false;
+		reinfo[i].ri_backendpid = 0;
+		reinfo[i].ri_dbid = InvalidOid;
+		reinfo[i].ri_relid = InvalidOid;
+		reinfo[i].ri_toastrelid = InvalidOid;
+	}
+}
+
 /*
  * Check if the table (and its index) still meets the requirements of
  * cluster_rel().
diff --git a/src/backend/postmaster/autovacuum.c b/src/backend/postmaster/autovacuum.c
index bd626a16363..080c64ea3c8 100644
--- a/src/backend/postmaster/autovacuum.c
+++ b/src/backend/postmaster/autovacuum.c
@@ -78,6 +78,7 @@
 #include "catalog/namespace.h"
 #include "catalog/pg_database.h"
 #include "catalog/pg_namespace.h"
+#include "commands/repack.h"
 #include "commands/vacuum.h"
 #include "common/int.h"
 #include "funcapi.h"
@@ -2422,6 +2423,25 @@ do_autovacuum(void)
 			}
 		}
 		LWLockRelease(AutovacuumLock);
+
+		/*
+		 * Similarly, if the table is being processed by concurrent repack,
+		 * skip it (but make a note of that).  We wouldn't be able to acquire
+		 * its lock anyway.
+		 */
+		if (!skipit)
+		{
+			MemoryContextSwitchTo(PortalContext);
+
+			skipit = is_table_under_repack(MyDatabaseId, relid);
+			if (skipit)
+				ereport(LOG,
+						errmsg("skipping table \"%s.%s.%s\" because it's being repacked in concurrent mode",
+							   get_database_name(MyDatabaseId),
+							   get_namespace_name(get_rel_namespace(relid)),
+							   get_rel_name(relid)));
+		}
+
 		if (skipit)
 		{
 			LWLockRelease(AutovacuumScheduleLock);
diff --git a/src/backend/utils/activity/wait_event_names.txt b/src/backend/utils/activity/wait_event_names.txt
index 7bda5298558..e206304f204 100644
--- a/src/backend/utils/activity/wait_event_names.txt
+++ b/src/backend/utils/activity/wait_event_names.txt
@@ -332,6 +332,7 @@ SInvalWrite	"Waiting to add a message to the shared catalog invalidation queue."
 WALBufMapping	"Waiting to replace a page in WAL buffers."
 WALWrite	"Waiting for WAL buffers to be written to disk."
 ControlFile	"Waiting to read or update the <filename>pg_control</filename> file or create a new WAL file."
+Repack	"Waiting to read or update tables in process by concurrent repack."
 MultiXactGen	"Waiting to read or update shared multixact state."
 RelCacheInit	"Waiting to read or update a <filename>pg_internal.init</filename> relation cache initialization file."
 CheckpointerComm	"Waiting to manage fsync requests."
diff --git a/src/include/commands/repack.h b/src/include/commands/repack.h
index fd16e74b179..be7d38b5fae 100644
--- a/src/include/commands/repack.h
+++ b/src/include/commands/repack.h
@@ -42,6 +42,8 @@ extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel);
 
 extern void cluster_rel(RepackCommand command, Relation OldHeap, Oid indexOid,
 						ClusterParams *params, bool isTopLevel);
+extern bool is_table_under_repack(Oid databaseId, Oid relid);
+
 extern void check_index_is_clusterable(Relation OldHeap, Oid indexOid,
 									   LOCKMODE lockmode);
 extern void mark_index_clustered(Relation rel, Oid indexOid, bool is_internal);
diff --git a/src/include/storage/lwlocklist.h b/src/include/storage/lwlocklist.h
index af8553bcb6c..3f08f4a15d4 100644
--- a/src/include/storage/lwlocklist.h
+++ b/src/include/storage/lwlocklist.h
@@ -41,7 +41,7 @@ PG_LWLOCK(6, SInvalWrite)
 PG_LWLOCK(7, WALBufMapping)
 PG_LWLOCK(8, WALWrite)
 PG_LWLOCK(9, ControlFile)
-/* 10 was CheckpointLock */
+PG_LWLOCK(10, Repack)
 /* 11 was XactSLRULock */
 /* 12 was SubtransSLRULock */
 PG_LWLOCK(13, MultiXactGen)
diff --git a/src/include/storage/subsystemlist.h b/src/include/storage/subsystemlist.h
index 9ad619080be..4e683b8b0a8 100644
--- a/src/include/storage/subsystemlist.h
+++ b/src/include/storage/subsystemlist.h
@@ -72,6 +72,7 @@ PG_SHMEM_SUBSYSTEM(WalSummarizerShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(PgArchShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(ApplyLauncherShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(SlotSyncShmemCallbacks)
+PG_SHMEM_SUBSYSTEM(RepackShmemCallbacks)
 
 /* other modules that need some shared memory space */
 PG_SHMEM_SUBSYSTEM(BTreeShmemCallbacks)
diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list
index 637c669a146..d019e03aaf1 100644
--- a/src/tools/pgindent/typedefs.list
+++ b/src/tools/pgindent/typedefs.list
@@ -2639,9 +2639,12 @@ ReorderBufferTupleCidEnt
 ReorderBufferTupleCidKey
 ReorderBufferUpdateProgressTxnCB
 ReorderTuple
+RepackCleanupContext
 RepackCommand
 RepackDecodingState
+RepackShmemStruct
 RepackStmt
+RepackWorkerInfo
 ReparameterizeForeignPathByChild_function
 ReplOriginId
 ReplOriginXactState
-- 
2.47.3


--kdrcpfmkbkc4lqhu--





^ permalink  raw  reply  [nested|flat] 223+ messages in thread

* [PATCH 2/2] Publish list of tables being repacked in shared memory
@ 2026-04-07 20:29 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 223+ messages in thread

From: Álvaro Herrera @ 2026-04-07 20:29 UTC (permalink / raw)

Use it in autovacuum to skip processing tables that are being repacked.
This is mostly to avoid repeated attempts to process such tables, which
would fail due to the special deadlock checker behavior for repack.

Author: Álvaro Herrera <[email protected]>
Discussion: https://postgr.es/m/[email protected]
---
 src/backend/commands/repack.c                 | 195 ++++++++++++++++--
 src/backend/postmaster/autovacuum.c           |  20 ++
 .../utils/activity/wait_event_names.txt       |   1 +
 src/include/commands/repack.h                 |   2 +
 src/include/storage/lwlocklist.h              |   2 +-
 src/include/storage/subsystemlist.h           |   1 +
 src/tools/pgindent/typedefs.list              |   3 +
 7 files changed, 210 insertions(+), 14 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index a5f5df77291..ee7072dce6a 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -63,9 +63,11 @@
 #include "optimizer/optimizer.h"
 #include "pgstat.h"
 #include "storage/bufmgr.h"
+#include "storage/ipc.h"
 #include "storage/lmgr.h"
 #include "storage/predicate.h"
 #include "storage/proc.h"
+#include "storage/subsystems.h"
 #include "utils/acl.h"
 #include "utils/fmgroids.h"
 #include "utils/guc.h"
@@ -79,6 +81,32 @@
 #include "utils/syscache.h"
 #include "utils/wait_event_types.h"
 
+
+/* Shared memory layout for REPACK */
+typedef struct RepackWorkerInfo
+{
+	bool		ri_in_use;
+	pid_t		ri_backendpid;
+	Oid			ri_dbid;
+	Oid			ri_relid;
+	Oid			ri_toastrelid;
+} RepackWorkerInfo;
+
+typedef struct
+{
+	bool		re_useless;
+	RepackWorkerInfo re_workerinfo[FLEXIBLE_ARRAY_MEMBER];
+} RepackShmemStruct;
+
+static RepackShmemStruct *RepackShmem;
+
+typedef struct RepackCleanupContext
+{
+	bool		concurrent;
+	int			workerindex;
+} RepackCleanupContext;
+
+
 /*
  * This struct is used to pass around the information on tables to be
  * clustered. We need this so we can make a list of them when invoked without
@@ -90,6 +118,7 @@ typedef struct
 	Oid			indexOid;
 } RelToCluster;
 
+
 /*
  * The first file exported by the decoding worker must contain a snapshot, the
  * following ones contain the data changes.
@@ -166,6 +195,10 @@ static List *get_tables_to_repack_partitioned(RepackCommand cmd,
 											  MemoryContext permcxt);
 static bool repack_is_permitted_for_relation(RepackCommand cmd,
 											 Oid relid, Oid userid);
+static void RepackCleanup(RepackCleanupContext *context);
+static void RepackCleanupCb(int code, Datum arg);
+static void RepackShmemRequest(void *arg);
+static void RepackShmemInit(void *arg);
 
 static void apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt);
 static void apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
@@ -210,6 +243,11 @@ static void ProcessRepackMessage(StringInfo msg);
 static const char *RepackCommandAsString(RepackCommand cmd);
 
 
+const ShmemCallbacks RepackShmemCallbacks = {
+	.request_fn = RepackShmemRequest,
+	.init_fn = RepackShmemInit,
+};
+
 /*
  * The repack code allows for processing multiple tables at once. Because
  * of this, we cannot just run everything on a single transaction, or we
@@ -514,6 +552,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 	Oid			tableOid = RelationGetRelid(OldHeap);
 	Relation	index;
 	LOCKMODE	lmode;
+	RepackCleanupContext context;
 	Oid			save_userid;
 	int			save_sec_context;
 	int			save_nestlevel;
@@ -660,24 +699,43 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 		TransferPredicateLocksToHeapRelation(OldHeap);
 
 	/* rebuild_relation does all the dirty work */
-	PG_TRY();
-	{
-		rebuild_relation(OldHeap, index, verbose, ident_idx);
-	}
-	PG_FINALLY();
+	context.concurrent = concurrent;
+
+	PG_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
 	{
 		if (concurrent)
 		{
-			/*
-			 * Since during normal operation the worker was already asked to
-			 * exit, stopping it explicitly is especially important on ERROR.
-			 * However it still seems a good practice to make sure that the
-			 * worker never survives the REPACK command.
-			 */
-			stop_repack_decoding_worker();
+			bool		freefound = false;
+
+			LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+			for (int i = 0; i < max_repack_replication_slots; i++)
+			{
+				RepackWorkerInfo *worker;
+
+				if (RepackShmem->re_workerinfo[i].ri_in_use)
+					continue;
+
+				freefound = true;
+				worker = &RepackShmem->re_workerinfo[i];
+				context.workerindex = i;
+
+				worker->ri_in_use = true;
+				worker->ri_backendpid = MyProcPid;
+				worker->ri_dbid = MyDatabaseId;
+				worker->ri_relid = RelationGetRelid(OldHeap);
+				worker->ri_toastrelid = OldHeap->rd_rel->reltoastrelid;
+				break;
+			}
+			if (!freefound)
+				elog(ERROR, "could not find free repack entry");
+			LWLockRelease(RepackLock);
 		}
+
+		rebuild_relation(OldHeap, index, verbose, ident_idx);
 	}
-	PG_END_TRY();
+	PG_END_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
+
+	RepackCleanup(&context);
 
 	/* rebuild_relation closes OldHeap, and index if valid */
 
@@ -691,6 +749,117 @@ out:
 	pgstat_progress_end_command();
 }
 
+/*
+ * Return whether any backend is running concurrent REPACK on the given table
+ * (which could be a toast table).
+ */
+bool
+is_table_under_repack(Oid databaseId, Oid relid)
+{
+	bool		retval = false;
+
+	LWLockAcquire(RepackLock, LW_SHARED);
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		RepackWorkerInfo *rworker;
+
+		if (!RepackShmem->re_workerinfo[i].ri_in_use)
+			continue;
+
+		rworker = &RepackShmem->re_workerinfo[i];
+		if (rworker->ri_dbid == MyDatabaseId &&
+			(rworker->ri_relid == relid ||
+			 rworker->ri_toastrelid == relid))
+			retval = true;
+	}
+	LWLockRelease(RepackLock);
+
+	return retval;
+}
+
+/*
+ * Remove ourselves from the workerinfo array.
+ */
+static void
+RepackCleanup(RepackCleanupContext *context)
+{
+	if (context->concurrent)
+	{
+		RepackWorkerInfo *worker;
+
+		/*
+		 * The worker would normally terminate on its own when the work is
+		 * done, but make sure we signal it just in case.
+		 */
+		stop_repack_decoding_worker();
+
+		/*
+		 * also, make sure we stop advertising the relation we were repacking,
+		 * so that autovacuum reverts to handling it normally.
+		 */
+		LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+
+		worker = &RepackShmem->re_workerinfo[context->workerindex];
+		Assert(worker->ri_backendpid == MyProcPid);
+		worker->ri_in_use = false;
+		worker->ri_backendpid = 0;
+		worker->ri_dbid = InvalidOid;
+		worker->ri_relid = InvalidOid;
+		worker->ri_toastrelid = InvalidOid;
+		LWLockRelease(RepackLock);
+	}
+}
+
+/*
+ * RepackCleanup wrapped as an on_shmem_exit callback function
+ */
+static void
+RepackCleanupCb(int code, Datum arg)
+{
+	RepackCleanup((RepackCleanupContext *) DatumGetPointer(arg));
+}
+
+/*
+ * RepackShmemRequest
+ *		Register shared memory space needed for repack
+ */
+static void
+RepackShmemRequest(void *arg)
+{
+	Size		size;
+
+	/*
+	 * Need the fixed struct and the array of RepackWorkerInfo.
+	 */
+	size = sizeof(RepackShmemStruct);
+	size = MAXALIGN(size);
+	size = add_size(size, mul_size(max_repack_replication_slots,
+								   sizeof(RepackWorkerInfo)));
+
+	ShmemRequestStruct(.name = "Repack Data",
+					   .size = size,
+					   .ptr = (void **) &RepackShmem,
+		);
+}
+
+static void
+RepackShmemInit(void *arg)
+{
+	RepackWorkerInfo *reinfo;
+
+	reinfo = (RepackWorkerInfo *) ((char *) RepackShmem +
+								   MAXALIGN(sizeof(RepackShmemStruct)));
+
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		reinfo[i].ri_in_use = false;
+		reinfo[i].ri_backendpid = 0;
+		reinfo[i].ri_dbid = InvalidOid;
+		reinfo[i].ri_relid = InvalidOid;
+		reinfo[i].ri_toastrelid = InvalidOid;
+	}
+}
+
 /*
  * Check if the table (and its index) still meets the requirements of
  * cluster_rel().
diff --git a/src/backend/postmaster/autovacuum.c b/src/backend/postmaster/autovacuum.c
index bd626a16363..080c64ea3c8 100644
--- a/src/backend/postmaster/autovacuum.c
+++ b/src/backend/postmaster/autovacuum.c
@@ -78,6 +78,7 @@
 #include "catalog/namespace.h"
 #include "catalog/pg_database.h"
 #include "catalog/pg_namespace.h"
+#include "commands/repack.h"
 #include "commands/vacuum.h"
 #include "common/int.h"
 #include "funcapi.h"
@@ -2422,6 +2423,25 @@ do_autovacuum(void)
 			}
 		}
 		LWLockRelease(AutovacuumLock);
+
+		/*
+		 * Similarly, if the table is being processed by concurrent repack,
+		 * skip it (but make a note of that).  We wouldn't be able to acquire
+		 * its lock anyway.
+		 */
+		if (!skipit)
+		{
+			MemoryContextSwitchTo(PortalContext);
+
+			skipit = is_table_under_repack(MyDatabaseId, relid);
+			if (skipit)
+				ereport(LOG,
+						errmsg("skipping table \"%s.%s.%s\" because it's being repacked in concurrent mode",
+							   get_database_name(MyDatabaseId),
+							   get_namespace_name(get_rel_namespace(relid)),
+							   get_rel_name(relid)));
+		}
+
 		if (skipit)
 		{
 			LWLockRelease(AutovacuumScheduleLock);
diff --git a/src/backend/utils/activity/wait_event_names.txt b/src/backend/utils/activity/wait_event_names.txt
index 7bda5298558..e206304f204 100644
--- a/src/backend/utils/activity/wait_event_names.txt
+++ b/src/backend/utils/activity/wait_event_names.txt
@@ -332,6 +332,7 @@ SInvalWrite	"Waiting to add a message to the shared catalog invalidation queue."
 WALBufMapping	"Waiting to replace a page in WAL buffers."
 WALWrite	"Waiting for WAL buffers to be written to disk."
 ControlFile	"Waiting to read or update the <filename>pg_control</filename> file or create a new WAL file."
+Repack	"Waiting to read or update tables in process by concurrent repack."
 MultiXactGen	"Waiting to read or update shared multixact state."
 RelCacheInit	"Waiting to read or update a <filename>pg_internal.init</filename> relation cache initialization file."
 CheckpointerComm	"Waiting to manage fsync requests."
diff --git a/src/include/commands/repack.h b/src/include/commands/repack.h
index fd16e74b179..be7d38b5fae 100644
--- a/src/include/commands/repack.h
+++ b/src/include/commands/repack.h
@@ -42,6 +42,8 @@ extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel);
 
 extern void cluster_rel(RepackCommand command, Relation OldHeap, Oid indexOid,
 						ClusterParams *params, bool isTopLevel);
+extern bool is_table_under_repack(Oid databaseId, Oid relid);
+
 extern void check_index_is_clusterable(Relation OldHeap, Oid indexOid,
 									   LOCKMODE lockmode);
 extern void mark_index_clustered(Relation rel, Oid indexOid, bool is_internal);
diff --git a/src/include/storage/lwlocklist.h b/src/include/storage/lwlocklist.h
index af8553bcb6c..3f08f4a15d4 100644
--- a/src/include/storage/lwlocklist.h
+++ b/src/include/storage/lwlocklist.h
@@ -41,7 +41,7 @@ PG_LWLOCK(6, SInvalWrite)
 PG_LWLOCK(7, WALBufMapping)
 PG_LWLOCK(8, WALWrite)
 PG_LWLOCK(9, ControlFile)
-/* 10 was CheckpointLock */
+PG_LWLOCK(10, Repack)
 /* 11 was XactSLRULock */
 /* 12 was SubtransSLRULock */
 PG_LWLOCK(13, MultiXactGen)
diff --git a/src/include/storage/subsystemlist.h b/src/include/storage/subsystemlist.h
index 9ad619080be..4e683b8b0a8 100644
--- a/src/include/storage/subsystemlist.h
+++ b/src/include/storage/subsystemlist.h
@@ -72,6 +72,7 @@ PG_SHMEM_SUBSYSTEM(WalSummarizerShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(PgArchShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(ApplyLauncherShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(SlotSyncShmemCallbacks)
+PG_SHMEM_SUBSYSTEM(RepackShmemCallbacks)
 
 /* other modules that need some shared memory space */
 PG_SHMEM_SUBSYSTEM(BTreeShmemCallbacks)
diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list
index 637c669a146..d019e03aaf1 100644
--- a/src/tools/pgindent/typedefs.list
+++ b/src/tools/pgindent/typedefs.list
@@ -2639,9 +2639,12 @@ ReorderBufferTupleCidEnt
 ReorderBufferTupleCidKey
 ReorderBufferUpdateProgressTxnCB
 ReorderTuple
+RepackCleanupContext
 RepackCommand
 RepackDecodingState
+RepackShmemStruct
 RepackStmt
+RepackWorkerInfo
 ReparameterizeForeignPathByChild_function
 ReplOriginId
 ReplOriginXactState
-- 
2.47.3


--kdrcpfmkbkc4lqhu--





^ permalink  raw  reply  [nested|flat] 223+ messages in thread

* [PATCH 2/2] Publish list of tables being repacked in shared memory
@ 2026-04-07 20:29 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 223+ messages in thread

From: Álvaro Herrera @ 2026-04-07 20:29 UTC (permalink / raw)

Use it in autovacuum to skip processing tables that are being repacked.
This is mostly to avoid repeated attempts to process such tables, which
would fail due to the special deadlock checker behavior for repack.

Author: Álvaro Herrera <[email protected]>
Discussion: https://postgr.es/m/[email protected]
---
 src/backend/commands/repack.c                 | 195 ++++++++++++++++--
 src/backend/postmaster/autovacuum.c           |  20 ++
 .../utils/activity/wait_event_names.txt       |   1 +
 src/include/commands/repack.h                 |   2 +
 src/include/storage/lwlocklist.h              |   2 +-
 src/include/storage/subsystemlist.h           |   1 +
 src/tools/pgindent/typedefs.list              |   3 +
 7 files changed, 210 insertions(+), 14 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index a5f5df77291..ee7072dce6a 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -63,9 +63,11 @@
 #include "optimizer/optimizer.h"
 #include "pgstat.h"
 #include "storage/bufmgr.h"
+#include "storage/ipc.h"
 #include "storage/lmgr.h"
 #include "storage/predicate.h"
 #include "storage/proc.h"
+#include "storage/subsystems.h"
 #include "utils/acl.h"
 #include "utils/fmgroids.h"
 #include "utils/guc.h"
@@ -79,6 +81,32 @@
 #include "utils/syscache.h"
 #include "utils/wait_event_types.h"
 
+
+/* Shared memory layout for REPACK */
+typedef struct RepackWorkerInfo
+{
+	bool		ri_in_use;
+	pid_t		ri_backendpid;
+	Oid			ri_dbid;
+	Oid			ri_relid;
+	Oid			ri_toastrelid;
+} RepackWorkerInfo;
+
+typedef struct
+{
+	bool		re_useless;
+	RepackWorkerInfo re_workerinfo[FLEXIBLE_ARRAY_MEMBER];
+} RepackShmemStruct;
+
+static RepackShmemStruct *RepackShmem;
+
+typedef struct RepackCleanupContext
+{
+	bool		concurrent;
+	int			workerindex;
+} RepackCleanupContext;
+
+
 /*
  * This struct is used to pass around the information on tables to be
  * clustered. We need this so we can make a list of them when invoked without
@@ -90,6 +118,7 @@ typedef struct
 	Oid			indexOid;
 } RelToCluster;
 
+
 /*
  * The first file exported by the decoding worker must contain a snapshot, the
  * following ones contain the data changes.
@@ -166,6 +195,10 @@ static List *get_tables_to_repack_partitioned(RepackCommand cmd,
 											  MemoryContext permcxt);
 static bool repack_is_permitted_for_relation(RepackCommand cmd,
 											 Oid relid, Oid userid);
+static void RepackCleanup(RepackCleanupContext *context);
+static void RepackCleanupCb(int code, Datum arg);
+static void RepackShmemRequest(void *arg);
+static void RepackShmemInit(void *arg);
 
 static void apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt);
 static void apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
@@ -210,6 +243,11 @@ static void ProcessRepackMessage(StringInfo msg);
 static const char *RepackCommandAsString(RepackCommand cmd);
 
 
+const ShmemCallbacks RepackShmemCallbacks = {
+	.request_fn = RepackShmemRequest,
+	.init_fn = RepackShmemInit,
+};
+
 /*
  * The repack code allows for processing multiple tables at once. Because
  * of this, we cannot just run everything on a single transaction, or we
@@ -514,6 +552,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 	Oid			tableOid = RelationGetRelid(OldHeap);
 	Relation	index;
 	LOCKMODE	lmode;
+	RepackCleanupContext context;
 	Oid			save_userid;
 	int			save_sec_context;
 	int			save_nestlevel;
@@ -660,24 +699,43 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 		TransferPredicateLocksToHeapRelation(OldHeap);
 
 	/* rebuild_relation does all the dirty work */
-	PG_TRY();
-	{
-		rebuild_relation(OldHeap, index, verbose, ident_idx);
-	}
-	PG_FINALLY();
+	context.concurrent = concurrent;
+
+	PG_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
 	{
 		if (concurrent)
 		{
-			/*
-			 * Since during normal operation the worker was already asked to
-			 * exit, stopping it explicitly is especially important on ERROR.
-			 * However it still seems a good practice to make sure that the
-			 * worker never survives the REPACK command.
-			 */
-			stop_repack_decoding_worker();
+			bool		freefound = false;
+
+			LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+			for (int i = 0; i < max_repack_replication_slots; i++)
+			{
+				RepackWorkerInfo *worker;
+
+				if (RepackShmem->re_workerinfo[i].ri_in_use)
+					continue;
+
+				freefound = true;
+				worker = &RepackShmem->re_workerinfo[i];
+				context.workerindex = i;
+
+				worker->ri_in_use = true;
+				worker->ri_backendpid = MyProcPid;
+				worker->ri_dbid = MyDatabaseId;
+				worker->ri_relid = RelationGetRelid(OldHeap);
+				worker->ri_toastrelid = OldHeap->rd_rel->reltoastrelid;
+				break;
+			}
+			if (!freefound)
+				elog(ERROR, "could not find free repack entry");
+			LWLockRelease(RepackLock);
 		}
+
+		rebuild_relation(OldHeap, index, verbose, ident_idx);
 	}
-	PG_END_TRY();
+	PG_END_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
+
+	RepackCleanup(&context);
 
 	/* rebuild_relation closes OldHeap, and index if valid */
 
@@ -691,6 +749,117 @@ out:
 	pgstat_progress_end_command();
 }
 
+/*
+ * Return whether any backend is running concurrent REPACK on the given table
+ * (which could be a toast table).
+ */
+bool
+is_table_under_repack(Oid databaseId, Oid relid)
+{
+	bool		retval = false;
+
+	LWLockAcquire(RepackLock, LW_SHARED);
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		RepackWorkerInfo *rworker;
+
+		if (!RepackShmem->re_workerinfo[i].ri_in_use)
+			continue;
+
+		rworker = &RepackShmem->re_workerinfo[i];
+		if (rworker->ri_dbid == MyDatabaseId &&
+			(rworker->ri_relid == relid ||
+			 rworker->ri_toastrelid == relid))
+			retval = true;
+	}
+	LWLockRelease(RepackLock);
+
+	return retval;
+}
+
+/*
+ * Remove ourselves from the workerinfo array.
+ */
+static void
+RepackCleanup(RepackCleanupContext *context)
+{
+	if (context->concurrent)
+	{
+		RepackWorkerInfo *worker;
+
+		/*
+		 * The worker would normally terminate on its own when the work is
+		 * done, but make sure we signal it just in case.
+		 */
+		stop_repack_decoding_worker();
+
+		/*
+		 * also, make sure we stop advertising the relation we were repacking,
+		 * so that autovacuum reverts to handling it normally.
+		 */
+		LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+
+		worker = &RepackShmem->re_workerinfo[context->workerindex];
+		Assert(worker->ri_backendpid == MyProcPid);
+		worker->ri_in_use = false;
+		worker->ri_backendpid = 0;
+		worker->ri_dbid = InvalidOid;
+		worker->ri_relid = InvalidOid;
+		worker->ri_toastrelid = InvalidOid;
+		LWLockRelease(RepackLock);
+	}
+}
+
+/*
+ * RepackCleanup wrapped as an on_shmem_exit callback function
+ */
+static void
+RepackCleanupCb(int code, Datum arg)
+{
+	RepackCleanup((RepackCleanupContext *) DatumGetPointer(arg));
+}
+
+/*
+ * RepackShmemRequest
+ *		Register shared memory space needed for repack
+ */
+static void
+RepackShmemRequest(void *arg)
+{
+	Size		size;
+
+	/*
+	 * Need the fixed struct and the array of RepackWorkerInfo.
+	 */
+	size = sizeof(RepackShmemStruct);
+	size = MAXALIGN(size);
+	size = add_size(size, mul_size(max_repack_replication_slots,
+								   sizeof(RepackWorkerInfo)));
+
+	ShmemRequestStruct(.name = "Repack Data",
+					   .size = size,
+					   .ptr = (void **) &RepackShmem,
+		);
+}
+
+static void
+RepackShmemInit(void *arg)
+{
+	RepackWorkerInfo *reinfo;
+
+	reinfo = (RepackWorkerInfo *) ((char *) RepackShmem +
+								   MAXALIGN(sizeof(RepackShmemStruct)));
+
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		reinfo[i].ri_in_use = false;
+		reinfo[i].ri_backendpid = 0;
+		reinfo[i].ri_dbid = InvalidOid;
+		reinfo[i].ri_relid = InvalidOid;
+		reinfo[i].ri_toastrelid = InvalidOid;
+	}
+}
+
 /*
  * Check if the table (and its index) still meets the requirements of
  * cluster_rel().
diff --git a/src/backend/postmaster/autovacuum.c b/src/backend/postmaster/autovacuum.c
index bd626a16363..080c64ea3c8 100644
--- a/src/backend/postmaster/autovacuum.c
+++ b/src/backend/postmaster/autovacuum.c
@@ -78,6 +78,7 @@
 #include "catalog/namespace.h"
 #include "catalog/pg_database.h"
 #include "catalog/pg_namespace.h"
+#include "commands/repack.h"
 #include "commands/vacuum.h"
 #include "common/int.h"
 #include "funcapi.h"
@@ -2422,6 +2423,25 @@ do_autovacuum(void)
 			}
 		}
 		LWLockRelease(AutovacuumLock);
+
+		/*
+		 * Similarly, if the table is being processed by concurrent repack,
+		 * skip it (but make a note of that).  We wouldn't be able to acquire
+		 * its lock anyway.
+		 */
+		if (!skipit)
+		{
+			MemoryContextSwitchTo(PortalContext);
+
+			skipit = is_table_under_repack(MyDatabaseId, relid);
+			if (skipit)
+				ereport(LOG,
+						errmsg("skipping table \"%s.%s.%s\" because it's being repacked in concurrent mode",
+							   get_database_name(MyDatabaseId),
+							   get_namespace_name(get_rel_namespace(relid)),
+							   get_rel_name(relid)));
+		}
+
 		if (skipit)
 		{
 			LWLockRelease(AutovacuumScheduleLock);
diff --git a/src/backend/utils/activity/wait_event_names.txt b/src/backend/utils/activity/wait_event_names.txt
index 7bda5298558..e206304f204 100644
--- a/src/backend/utils/activity/wait_event_names.txt
+++ b/src/backend/utils/activity/wait_event_names.txt
@@ -332,6 +332,7 @@ SInvalWrite	"Waiting to add a message to the shared catalog invalidation queue."
 WALBufMapping	"Waiting to replace a page in WAL buffers."
 WALWrite	"Waiting for WAL buffers to be written to disk."
 ControlFile	"Waiting to read or update the <filename>pg_control</filename> file or create a new WAL file."
+Repack	"Waiting to read or update tables in process by concurrent repack."
 MultiXactGen	"Waiting to read or update shared multixact state."
 RelCacheInit	"Waiting to read or update a <filename>pg_internal.init</filename> relation cache initialization file."
 CheckpointerComm	"Waiting to manage fsync requests."
diff --git a/src/include/commands/repack.h b/src/include/commands/repack.h
index fd16e74b179..be7d38b5fae 100644
--- a/src/include/commands/repack.h
+++ b/src/include/commands/repack.h
@@ -42,6 +42,8 @@ extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel);
 
 extern void cluster_rel(RepackCommand command, Relation OldHeap, Oid indexOid,
 						ClusterParams *params, bool isTopLevel);
+extern bool is_table_under_repack(Oid databaseId, Oid relid);
+
 extern void check_index_is_clusterable(Relation OldHeap, Oid indexOid,
 									   LOCKMODE lockmode);
 extern void mark_index_clustered(Relation rel, Oid indexOid, bool is_internal);
diff --git a/src/include/storage/lwlocklist.h b/src/include/storage/lwlocklist.h
index af8553bcb6c..3f08f4a15d4 100644
--- a/src/include/storage/lwlocklist.h
+++ b/src/include/storage/lwlocklist.h
@@ -41,7 +41,7 @@ PG_LWLOCK(6, SInvalWrite)
 PG_LWLOCK(7, WALBufMapping)
 PG_LWLOCK(8, WALWrite)
 PG_LWLOCK(9, ControlFile)
-/* 10 was CheckpointLock */
+PG_LWLOCK(10, Repack)
 /* 11 was XactSLRULock */
 /* 12 was SubtransSLRULock */
 PG_LWLOCK(13, MultiXactGen)
diff --git a/src/include/storage/subsystemlist.h b/src/include/storage/subsystemlist.h
index 9ad619080be..4e683b8b0a8 100644
--- a/src/include/storage/subsystemlist.h
+++ b/src/include/storage/subsystemlist.h
@@ -72,6 +72,7 @@ PG_SHMEM_SUBSYSTEM(WalSummarizerShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(PgArchShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(ApplyLauncherShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(SlotSyncShmemCallbacks)
+PG_SHMEM_SUBSYSTEM(RepackShmemCallbacks)
 
 /* other modules that need some shared memory space */
 PG_SHMEM_SUBSYSTEM(BTreeShmemCallbacks)
diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list
index 637c669a146..d019e03aaf1 100644
--- a/src/tools/pgindent/typedefs.list
+++ b/src/tools/pgindent/typedefs.list
@@ -2639,9 +2639,12 @@ ReorderBufferTupleCidEnt
 ReorderBufferTupleCidKey
 ReorderBufferUpdateProgressTxnCB
 ReorderTuple
+RepackCleanupContext
 RepackCommand
 RepackDecodingState
+RepackShmemStruct
 RepackStmt
+RepackWorkerInfo
 ReparameterizeForeignPathByChild_function
 ReplOriginId
 ReplOriginXactState
-- 
2.47.3


--kdrcpfmkbkc4lqhu--





^ permalink  raw  reply  [nested|flat] 223+ messages in thread

* [PATCH 2/2] Publish list of tables being repacked in shared memory
@ 2026-04-07 20:29 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 223+ messages in thread

From: Álvaro Herrera @ 2026-04-07 20:29 UTC (permalink / raw)

Use it in autovacuum to skip processing tables that are being repacked.
This is mostly to avoid repeated attempts to process such tables, which
would fail due to the special deadlock checker behavior for repack.

Author: Álvaro Herrera <[email protected]>
Discussion: https://postgr.es/m/[email protected]
---
 src/backend/commands/repack.c                 | 195 ++++++++++++++++--
 src/backend/postmaster/autovacuum.c           |  20 ++
 .../utils/activity/wait_event_names.txt       |   1 +
 src/include/commands/repack.h                 |   2 +
 src/include/storage/lwlocklist.h              |   2 +-
 src/include/storage/subsystemlist.h           |   1 +
 src/tools/pgindent/typedefs.list              |   3 +
 7 files changed, 210 insertions(+), 14 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index a5f5df77291..ee7072dce6a 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -63,9 +63,11 @@
 #include "optimizer/optimizer.h"
 #include "pgstat.h"
 #include "storage/bufmgr.h"
+#include "storage/ipc.h"
 #include "storage/lmgr.h"
 #include "storage/predicate.h"
 #include "storage/proc.h"
+#include "storage/subsystems.h"
 #include "utils/acl.h"
 #include "utils/fmgroids.h"
 #include "utils/guc.h"
@@ -79,6 +81,32 @@
 #include "utils/syscache.h"
 #include "utils/wait_event_types.h"
 
+
+/* Shared memory layout for REPACK */
+typedef struct RepackWorkerInfo
+{
+	bool		ri_in_use;
+	pid_t		ri_backendpid;
+	Oid			ri_dbid;
+	Oid			ri_relid;
+	Oid			ri_toastrelid;
+} RepackWorkerInfo;
+
+typedef struct
+{
+	bool		re_useless;
+	RepackWorkerInfo re_workerinfo[FLEXIBLE_ARRAY_MEMBER];
+} RepackShmemStruct;
+
+static RepackShmemStruct *RepackShmem;
+
+typedef struct RepackCleanupContext
+{
+	bool		concurrent;
+	int			workerindex;
+} RepackCleanupContext;
+
+
 /*
  * This struct is used to pass around the information on tables to be
  * clustered. We need this so we can make a list of them when invoked without
@@ -90,6 +118,7 @@ typedef struct
 	Oid			indexOid;
 } RelToCluster;
 
+
 /*
  * The first file exported by the decoding worker must contain a snapshot, the
  * following ones contain the data changes.
@@ -166,6 +195,10 @@ static List *get_tables_to_repack_partitioned(RepackCommand cmd,
 											  MemoryContext permcxt);
 static bool repack_is_permitted_for_relation(RepackCommand cmd,
 											 Oid relid, Oid userid);
+static void RepackCleanup(RepackCleanupContext *context);
+static void RepackCleanupCb(int code, Datum arg);
+static void RepackShmemRequest(void *arg);
+static void RepackShmemInit(void *arg);
 
 static void apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt);
 static void apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
@@ -210,6 +243,11 @@ static void ProcessRepackMessage(StringInfo msg);
 static const char *RepackCommandAsString(RepackCommand cmd);
 
 
+const ShmemCallbacks RepackShmemCallbacks = {
+	.request_fn = RepackShmemRequest,
+	.init_fn = RepackShmemInit,
+};
+
 /*
  * The repack code allows for processing multiple tables at once. Because
  * of this, we cannot just run everything on a single transaction, or we
@@ -514,6 +552,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 	Oid			tableOid = RelationGetRelid(OldHeap);
 	Relation	index;
 	LOCKMODE	lmode;
+	RepackCleanupContext context;
 	Oid			save_userid;
 	int			save_sec_context;
 	int			save_nestlevel;
@@ -660,24 +699,43 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 		TransferPredicateLocksToHeapRelation(OldHeap);
 
 	/* rebuild_relation does all the dirty work */
-	PG_TRY();
-	{
-		rebuild_relation(OldHeap, index, verbose, ident_idx);
-	}
-	PG_FINALLY();
+	context.concurrent = concurrent;
+
+	PG_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
 	{
 		if (concurrent)
 		{
-			/*
-			 * Since during normal operation the worker was already asked to
-			 * exit, stopping it explicitly is especially important on ERROR.
-			 * However it still seems a good practice to make sure that the
-			 * worker never survives the REPACK command.
-			 */
-			stop_repack_decoding_worker();
+			bool		freefound = false;
+
+			LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+			for (int i = 0; i < max_repack_replication_slots; i++)
+			{
+				RepackWorkerInfo *worker;
+
+				if (RepackShmem->re_workerinfo[i].ri_in_use)
+					continue;
+
+				freefound = true;
+				worker = &RepackShmem->re_workerinfo[i];
+				context.workerindex = i;
+
+				worker->ri_in_use = true;
+				worker->ri_backendpid = MyProcPid;
+				worker->ri_dbid = MyDatabaseId;
+				worker->ri_relid = RelationGetRelid(OldHeap);
+				worker->ri_toastrelid = OldHeap->rd_rel->reltoastrelid;
+				break;
+			}
+			if (!freefound)
+				elog(ERROR, "could not find free repack entry");
+			LWLockRelease(RepackLock);
 		}
+
+		rebuild_relation(OldHeap, index, verbose, ident_idx);
 	}
-	PG_END_TRY();
+	PG_END_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
+
+	RepackCleanup(&context);
 
 	/* rebuild_relation closes OldHeap, and index if valid */
 
@@ -691,6 +749,117 @@ out:
 	pgstat_progress_end_command();
 }
 
+/*
+ * Return whether any backend is running concurrent REPACK on the given table
+ * (which could be a toast table).
+ */
+bool
+is_table_under_repack(Oid databaseId, Oid relid)
+{
+	bool		retval = false;
+
+	LWLockAcquire(RepackLock, LW_SHARED);
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		RepackWorkerInfo *rworker;
+
+		if (!RepackShmem->re_workerinfo[i].ri_in_use)
+			continue;
+
+		rworker = &RepackShmem->re_workerinfo[i];
+		if (rworker->ri_dbid == MyDatabaseId &&
+			(rworker->ri_relid == relid ||
+			 rworker->ri_toastrelid == relid))
+			retval = true;
+	}
+	LWLockRelease(RepackLock);
+
+	return retval;
+}
+
+/*
+ * Remove ourselves from the workerinfo array.
+ */
+static void
+RepackCleanup(RepackCleanupContext *context)
+{
+	if (context->concurrent)
+	{
+		RepackWorkerInfo *worker;
+
+		/*
+		 * The worker would normally terminate on its own when the work is
+		 * done, but make sure we signal it just in case.
+		 */
+		stop_repack_decoding_worker();
+
+		/*
+		 * also, make sure we stop advertising the relation we were repacking,
+		 * so that autovacuum reverts to handling it normally.
+		 */
+		LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+
+		worker = &RepackShmem->re_workerinfo[context->workerindex];
+		Assert(worker->ri_backendpid == MyProcPid);
+		worker->ri_in_use = false;
+		worker->ri_backendpid = 0;
+		worker->ri_dbid = InvalidOid;
+		worker->ri_relid = InvalidOid;
+		worker->ri_toastrelid = InvalidOid;
+		LWLockRelease(RepackLock);
+	}
+}
+
+/*
+ * RepackCleanup wrapped as an on_shmem_exit callback function
+ */
+static void
+RepackCleanupCb(int code, Datum arg)
+{
+	RepackCleanup((RepackCleanupContext *) DatumGetPointer(arg));
+}
+
+/*
+ * RepackShmemRequest
+ *		Register shared memory space needed for repack
+ */
+static void
+RepackShmemRequest(void *arg)
+{
+	Size		size;
+
+	/*
+	 * Need the fixed struct and the array of RepackWorkerInfo.
+	 */
+	size = sizeof(RepackShmemStruct);
+	size = MAXALIGN(size);
+	size = add_size(size, mul_size(max_repack_replication_slots,
+								   sizeof(RepackWorkerInfo)));
+
+	ShmemRequestStruct(.name = "Repack Data",
+					   .size = size,
+					   .ptr = (void **) &RepackShmem,
+		);
+}
+
+static void
+RepackShmemInit(void *arg)
+{
+	RepackWorkerInfo *reinfo;
+
+	reinfo = (RepackWorkerInfo *) ((char *) RepackShmem +
+								   MAXALIGN(sizeof(RepackShmemStruct)));
+
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		reinfo[i].ri_in_use = false;
+		reinfo[i].ri_backendpid = 0;
+		reinfo[i].ri_dbid = InvalidOid;
+		reinfo[i].ri_relid = InvalidOid;
+		reinfo[i].ri_toastrelid = InvalidOid;
+	}
+}
+
 /*
  * Check if the table (and its index) still meets the requirements of
  * cluster_rel().
diff --git a/src/backend/postmaster/autovacuum.c b/src/backend/postmaster/autovacuum.c
index bd626a16363..080c64ea3c8 100644
--- a/src/backend/postmaster/autovacuum.c
+++ b/src/backend/postmaster/autovacuum.c
@@ -78,6 +78,7 @@
 #include "catalog/namespace.h"
 #include "catalog/pg_database.h"
 #include "catalog/pg_namespace.h"
+#include "commands/repack.h"
 #include "commands/vacuum.h"
 #include "common/int.h"
 #include "funcapi.h"
@@ -2422,6 +2423,25 @@ do_autovacuum(void)
 			}
 		}
 		LWLockRelease(AutovacuumLock);
+
+		/*
+		 * Similarly, if the table is being processed by concurrent repack,
+		 * skip it (but make a note of that).  We wouldn't be able to acquire
+		 * its lock anyway.
+		 */
+		if (!skipit)
+		{
+			MemoryContextSwitchTo(PortalContext);
+
+			skipit = is_table_under_repack(MyDatabaseId, relid);
+			if (skipit)
+				ereport(LOG,
+						errmsg("skipping table \"%s.%s.%s\" because it's being repacked in concurrent mode",
+							   get_database_name(MyDatabaseId),
+							   get_namespace_name(get_rel_namespace(relid)),
+							   get_rel_name(relid)));
+		}
+
 		if (skipit)
 		{
 			LWLockRelease(AutovacuumScheduleLock);
diff --git a/src/backend/utils/activity/wait_event_names.txt b/src/backend/utils/activity/wait_event_names.txt
index 7bda5298558..e206304f204 100644
--- a/src/backend/utils/activity/wait_event_names.txt
+++ b/src/backend/utils/activity/wait_event_names.txt
@@ -332,6 +332,7 @@ SInvalWrite	"Waiting to add a message to the shared catalog invalidation queue."
 WALBufMapping	"Waiting to replace a page in WAL buffers."
 WALWrite	"Waiting for WAL buffers to be written to disk."
 ControlFile	"Waiting to read or update the <filename>pg_control</filename> file or create a new WAL file."
+Repack	"Waiting to read or update tables in process by concurrent repack."
 MultiXactGen	"Waiting to read or update shared multixact state."
 RelCacheInit	"Waiting to read or update a <filename>pg_internal.init</filename> relation cache initialization file."
 CheckpointerComm	"Waiting to manage fsync requests."
diff --git a/src/include/commands/repack.h b/src/include/commands/repack.h
index fd16e74b179..be7d38b5fae 100644
--- a/src/include/commands/repack.h
+++ b/src/include/commands/repack.h
@@ -42,6 +42,8 @@ extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel);
 
 extern void cluster_rel(RepackCommand command, Relation OldHeap, Oid indexOid,
 						ClusterParams *params, bool isTopLevel);
+extern bool is_table_under_repack(Oid databaseId, Oid relid);
+
 extern void check_index_is_clusterable(Relation OldHeap, Oid indexOid,
 									   LOCKMODE lockmode);
 extern void mark_index_clustered(Relation rel, Oid indexOid, bool is_internal);
diff --git a/src/include/storage/lwlocklist.h b/src/include/storage/lwlocklist.h
index af8553bcb6c..3f08f4a15d4 100644
--- a/src/include/storage/lwlocklist.h
+++ b/src/include/storage/lwlocklist.h
@@ -41,7 +41,7 @@ PG_LWLOCK(6, SInvalWrite)
 PG_LWLOCK(7, WALBufMapping)
 PG_LWLOCK(8, WALWrite)
 PG_LWLOCK(9, ControlFile)
-/* 10 was CheckpointLock */
+PG_LWLOCK(10, Repack)
 /* 11 was XactSLRULock */
 /* 12 was SubtransSLRULock */
 PG_LWLOCK(13, MultiXactGen)
diff --git a/src/include/storage/subsystemlist.h b/src/include/storage/subsystemlist.h
index 9ad619080be..4e683b8b0a8 100644
--- a/src/include/storage/subsystemlist.h
+++ b/src/include/storage/subsystemlist.h
@@ -72,6 +72,7 @@ PG_SHMEM_SUBSYSTEM(WalSummarizerShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(PgArchShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(ApplyLauncherShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(SlotSyncShmemCallbacks)
+PG_SHMEM_SUBSYSTEM(RepackShmemCallbacks)
 
 /* other modules that need some shared memory space */
 PG_SHMEM_SUBSYSTEM(BTreeShmemCallbacks)
diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list
index 637c669a146..d019e03aaf1 100644
--- a/src/tools/pgindent/typedefs.list
+++ b/src/tools/pgindent/typedefs.list
@@ -2639,9 +2639,12 @@ ReorderBufferTupleCidEnt
 ReorderBufferTupleCidKey
 ReorderBufferUpdateProgressTxnCB
 ReorderTuple
+RepackCleanupContext
 RepackCommand
 RepackDecodingState
+RepackShmemStruct
 RepackStmt
+RepackWorkerInfo
 ReparameterizeForeignPathByChild_function
 ReplOriginId
 ReplOriginXactState
-- 
2.47.3


--kdrcpfmkbkc4lqhu--





^ permalink  raw  reply  [nested|flat] 223+ messages in thread

* [PATCH 2/2] Publish list of tables being repacked in shared memory
@ 2026-04-07 20:29 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 223+ messages in thread

From: Álvaro Herrera @ 2026-04-07 20:29 UTC (permalink / raw)

Use it in autovacuum to skip processing tables that are being repacked.
This is mostly to avoid repeated attempts to process such tables, which
would fail due to the special deadlock checker behavior for repack.

Author: Álvaro Herrera <[email protected]>
Discussion: https://postgr.es/m/[email protected]
---
 src/backend/commands/repack.c                 | 195 ++++++++++++++++--
 src/backend/postmaster/autovacuum.c           |  20 ++
 .../utils/activity/wait_event_names.txt       |   1 +
 src/include/commands/repack.h                 |   2 +
 src/include/storage/lwlocklist.h              |   2 +-
 src/include/storage/subsystemlist.h           |   1 +
 src/tools/pgindent/typedefs.list              |   3 +
 7 files changed, 210 insertions(+), 14 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index a5f5df77291..ee7072dce6a 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -63,9 +63,11 @@
 #include "optimizer/optimizer.h"
 #include "pgstat.h"
 #include "storage/bufmgr.h"
+#include "storage/ipc.h"
 #include "storage/lmgr.h"
 #include "storage/predicate.h"
 #include "storage/proc.h"
+#include "storage/subsystems.h"
 #include "utils/acl.h"
 #include "utils/fmgroids.h"
 #include "utils/guc.h"
@@ -79,6 +81,32 @@
 #include "utils/syscache.h"
 #include "utils/wait_event_types.h"
 
+
+/* Shared memory layout for REPACK */
+typedef struct RepackWorkerInfo
+{
+	bool		ri_in_use;
+	pid_t		ri_backendpid;
+	Oid			ri_dbid;
+	Oid			ri_relid;
+	Oid			ri_toastrelid;
+} RepackWorkerInfo;
+
+typedef struct
+{
+	bool		re_useless;
+	RepackWorkerInfo re_workerinfo[FLEXIBLE_ARRAY_MEMBER];
+} RepackShmemStruct;
+
+static RepackShmemStruct *RepackShmem;
+
+typedef struct RepackCleanupContext
+{
+	bool		concurrent;
+	int			workerindex;
+} RepackCleanupContext;
+
+
 /*
  * This struct is used to pass around the information on tables to be
  * clustered. We need this so we can make a list of them when invoked without
@@ -90,6 +118,7 @@ typedef struct
 	Oid			indexOid;
 } RelToCluster;
 
+
 /*
  * The first file exported by the decoding worker must contain a snapshot, the
  * following ones contain the data changes.
@@ -166,6 +195,10 @@ static List *get_tables_to_repack_partitioned(RepackCommand cmd,
 											  MemoryContext permcxt);
 static bool repack_is_permitted_for_relation(RepackCommand cmd,
 											 Oid relid, Oid userid);
+static void RepackCleanup(RepackCleanupContext *context);
+static void RepackCleanupCb(int code, Datum arg);
+static void RepackShmemRequest(void *arg);
+static void RepackShmemInit(void *arg);
 
 static void apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt);
 static void apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
@@ -210,6 +243,11 @@ static void ProcessRepackMessage(StringInfo msg);
 static const char *RepackCommandAsString(RepackCommand cmd);
 
 
+const ShmemCallbacks RepackShmemCallbacks = {
+	.request_fn = RepackShmemRequest,
+	.init_fn = RepackShmemInit,
+};
+
 /*
  * The repack code allows for processing multiple tables at once. Because
  * of this, we cannot just run everything on a single transaction, or we
@@ -514,6 +552,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 	Oid			tableOid = RelationGetRelid(OldHeap);
 	Relation	index;
 	LOCKMODE	lmode;
+	RepackCleanupContext context;
 	Oid			save_userid;
 	int			save_sec_context;
 	int			save_nestlevel;
@@ -660,24 +699,43 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 		TransferPredicateLocksToHeapRelation(OldHeap);
 
 	/* rebuild_relation does all the dirty work */
-	PG_TRY();
-	{
-		rebuild_relation(OldHeap, index, verbose, ident_idx);
-	}
-	PG_FINALLY();
+	context.concurrent = concurrent;
+
+	PG_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
 	{
 		if (concurrent)
 		{
-			/*
-			 * Since during normal operation the worker was already asked to
-			 * exit, stopping it explicitly is especially important on ERROR.
-			 * However it still seems a good practice to make sure that the
-			 * worker never survives the REPACK command.
-			 */
-			stop_repack_decoding_worker();
+			bool		freefound = false;
+
+			LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+			for (int i = 0; i < max_repack_replication_slots; i++)
+			{
+				RepackWorkerInfo *worker;
+
+				if (RepackShmem->re_workerinfo[i].ri_in_use)
+					continue;
+
+				freefound = true;
+				worker = &RepackShmem->re_workerinfo[i];
+				context.workerindex = i;
+
+				worker->ri_in_use = true;
+				worker->ri_backendpid = MyProcPid;
+				worker->ri_dbid = MyDatabaseId;
+				worker->ri_relid = RelationGetRelid(OldHeap);
+				worker->ri_toastrelid = OldHeap->rd_rel->reltoastrelid;
+				break;
+			}
+			if (!freefound)
+				elog(ERROR, "could not find free repack entry");
+			LWLockRelease(RepackLock);
 		}
+
+		rebuild_relation(OldHeap, index, verbose, ident_idx);
 	}
-	PG_END_TRY();
+	PG_END_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
+
+	RepackCleanup(&context);
 
 	/* rebuild_relation closes OldHeap, and index if valid */
 
@@ -691,6 +749,117 @@ out:
 	pgstat_progress_end_command();
 }
 
+/*
+ * Return whether any backend is running concurrent REPACK on the given table
+ * (which could be a toast table).
+ */
+bool
+is_table_under_repack(Oid databaseId, Oid relid)
+{
+	bool		retval = false;
+
+	LWLockAcquire(RepackLock, LW_SHARED);
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		RepackWorkerInfo *rworker;
+
+		if (!RepackShmem->re_workerinfo[i].ri_in_use)
+			continue;
+
+		rworker = &RepackShmem->re_workerinfo[i];
+		if (rworker->ri_dbid == MyDatabaseId &&
+			(rworker->ri_relid == relid ||
+			 rworker->ri_toastrelid == relid))
+			retval = true;
+	}
+	LWLockRelease(RepackLock);
+
+	return retval;
+}
+
+/*
+ * Remove ourselves from the workerinfo array.
+ */
+static void
+RepackCleanup(RepackCleanupContext *context)
+{
+	if (context->concurrent)
+	{
+		RepackWorkerInfo *worker;
+
+		/*
+		 * The worker would normally terminate on its own when the work is
+		 * done, but make sure we signal it just in case.
+		 */
+		stop_repack_decoding_worker();
+
+		/*
+		 * also, make sure we stop advertising the relation we were repacking,
+		 * so that autovacuum reverts to handling it normally.
+		 */
+		LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+
+		worker = &RepackShmem->re_workerinfo[context->workerindex];
+		Assert(worker->ri_backendpid == MyProcPid);
+		worker->ri_in_use = false;
+		worker->ri_backendpid = 0;
+		worker->ri_dbid = InvalidOid;
+		worker->ri_relid = InvalidOid;
+		worker->ri_toastrelid = InvalidOid;
+		LWLockRelease(RepackLock);
+	}
+}
+
+/*
+ * RepackCleanup wrapped as an on_shmem_exit callback function
+ */
+static void
+RepackCleanupCb(int code, Datum arg)
+{
+	RepackCleanup((RepackCleanupContext *) DatumGetPointer(arg));
+}
+
+/*
+ * RepackShmemRequest
+ *		Register shared memory space needed for repack
+ */
+static void
+RepackShmemRequest(void *arg)
+{
+	Size		size;
+
+	/*
+	 * Need the fixed struct and the array of RepackWorkerInfo.
+	 */
+	size = sizeof(RepackShmemStruct);
+	size = MAXALIGN(size);
+	size = add_size(size, mul_size(max_repack_replication_slots,
+								   sizeof(RepackWorkerInfo)));
+
+	ShmemRequestStruct(.name = "Repack Data",
+					   .size = size,
+					   .ptr = (void **) &RepackShmem,
+		);
+}
+
+static void
+RepackShmemInit(void *arg)
+{
+	RepackWorkerInfo *reinfo;
+
+	reinfo = (RepackWorkerInfo *) ((char *) RepackShmem +
+								   MAXALIGN(sizeof(RepackShmemStruct)));
+
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		reinfo[i].ri_in_use = false;
+		reinfo[i].ri_backendpid = 0;
+		reinfo[i].ri_dbid = InvalidOid;
+		reinfo[i].ri_relid = InvalidOid;
+		reinfo[i].ri_toastrelid = InvalidOid;
+	}
+}
+
 /*
  * Check if the table (and its index) still meets the requirements of
  * cluster_rel().
diff --git a/src/backend/postmaster/autovacuum.c b/src/backend/postmaster/autovacuum.c
index bd626a16363..080c64ea3c8 100644
--- a/src/backend/postmaster/autovacuum.c
+++ b/src/backend/postmaster/autovacuum.c
@@ -78,6 +78,7 @@
 #include "catalog/namespace.h"
 #include "catalog/pg_database.h"
 #include "catalog/pg_namespace.h"
+#include "commands/repack.h"
 #include "commands/vacuum.h"
 #include "common/int.h"
 #include "funcapi.h"
@@ -2422,6 +2423,25 @@ do_autovacuum(void)
 			}
 		}
 		LWLockRelease(AutovacuumLock);
+
+		/*
+		 * Similarly, if the table is being processed by concurrent repack,
+		 * skip it (but make a note of that).  We wouldn't be able to acquire
+		 * its lock anyway.
+		 */
+		if (!skipit)
+		{
+			MemoryContextSwitchTo(PortalContext);
+
+			skipit = is_table_under_repack(MyDatabaseId, relid);
+			if (skipit)
+				ereport(LOG,
+						errmsg("skipping table \"%s.%s.%s\" because it's being repacked in concurrent mode",
+							   get_database_name(MyDatabaseId),
+							   get_namespace_name(get_rel_namespace(relid)),
+							   get_rel_name(relid)));
+		}
+
 		if (skipit)
 		{
 			LWLockRelease(AutovacuumScheduleLock);
diff --git a/src/backend/utils/activity/wait_event_names.txt b/src/backend/utils/activity/wait_event_names.txt
index 7bda5298558..e206304f204 100644
--- a/src/backend/utils/activity/wait_event_names.txt
+++ b/src/backend/utils/activity/wait_event_names.txt
@@ -332,6 +332,7 @@ SInvalWrite	"Waiting to add a message to the shared catalog invalidation queue."
 WALBufMapping	"Waiting to replace a page in WAL buffers."
 WALWrite	"Waiting for WAL buffers to be written to disk."
 ControlFile	"Waiting to read or update the <filename>pg_control</filename> file or create a new WAL file."
+Repack	"Waiting to read or update tables in process by concurrent repack."
 MultiXactGen	"Waiting to read or update shared multixact state."
 RelCacheInit	"Waiting to read or update a <filename>pg_internal.init</filename> relation cache initialization file."
 CheckpointerComm	"Waiting to manage fsync requests."
diff --git a/src/include/commands/repack.h b/src/include/commands/repack.h
index fd16e74b179..be7d38b5fae 100644
--- a/src/include/commands/repack.h
+++ b/src/include/commands/repack.h
@@ -42,6 +42,8 @@ extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel);
 
 extern void cluster_rel(RepackCommand command, Relation OldHeap, Oid indexOid,
 						ClusterParams *params, bool isTopLevel);
+extern bool is_table_under_repack(Oid databaseId, Oid relid);
+
 extern void check_index_is_clusterable(Relation OldHeap, Oid indexOid,
 									   LOCKMODE lockmode);
 extern void mark_index_clustered(Relation rel, Oid indexOid, bool is_internal);
diff --git a/src/include/storage/lwlocklist.h b/src/include/storage/lwlocklist.h
index af8553bcb6c..3f08f4a15d4 100644
--- a/src/include/storage/lwlocklist.h
+++ b/src/include/storage/lwlocklist.h
@@ -41,7 +41,7 @@ PG_LWLOCK(6, SInvalWrite)
 PG_LWLOCK(7, WALBufMapping)
 PG_LWLOCK(8, WALWrite)
 PG_LWLOCK(9, ControlFile)
-/* 10 was CheckpointLock */
+PG_LWLOCK(10, Repack)
 /* 11 was XactSLRULock */
 /* 12 was SubtransSLRULock */
 PG_LWLOCK(13, MultiXactGen)
diff --git a/src/include/storage/subsystemlist.h b/src/include/storage/subsystemlist.h
index 9ad619080be..4e683b8b0a8 100644
--- a/src/include/storage/subsystemlist.h
+++ b/src/include/storage/subsystemlist.h
@@ -72,6 +72,7 @@ PG_SHMEM_SUBSYSTEM(WalSummarizerShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(PgArchShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(ApplyLauncherShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(SlotSyncShmemCallbacks)
+PG_SHMEM_SUBSYSTEM(RepackShmemCallbacks)
 
 /* other modules that need some shared memory space */
 PG_SHMEM_SUBSYSTEM(BTreeShmemCallbacks)
diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list
index 637c669a146..d019e03aaf1 100644
--- a/src/tools/pgindent/typedefs.list
+++ b/src/tools/pgindent/typedefs.list
@@ -2639,9 +2639,12 @@ ReorderBufferTupleCidEnt
 ReorderBufferTupleCidKey
 ReorderBufferUpdateProgressTxnCB
 ReorderTuple
+RepackCleanupContext
 RepackCommand
 RepackDecodingState
+RepackShmemStruct
 RepackStmt
+RepackWorkerInfo
 ReparameterizeForeignPathByChild_function
 ReplOriginId
 ReplOriginXactState
-- 
2.47.3


--kdrcpfmkbkc4lqhu--





^ permalink  raw  reply  [nested|flat] 223+ messages in thread

* [PATCH 2/2] Publish list of tables being repacked in shared memory
@ 2026-04-07 20:29 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 223+ messages in thread

From: Álvaro Herrera @ 2026-04-07 20:29 UTC (permalink / raw)

Use it in autovacuum to skip processing tables that are being repacked.
This is mostly to avoid repeated attempts to process such tables, which
would fail due to the special deadlock checker behavior for repack.

Author: Álvaro Herrera <[email protected]>
Discussion: https://postgr.es/m/[email protected]
---
 src/backend/commands/repack.c                 | 195 ++++++++++++++++--
 src/backend/postmaster/autovacuum.c           |  20 ++
 .../utils/activity/wait_event_names.txt       |   1 +
 src/include/commands/repack.h                 |   2 +
 src/include/storage/lwlocklist.h              |   2 +-
 src/include/storage/subsystemlist.h           |   1 +
 src/tools/pgindent/typedefs.list              |   3 +
 7 files changed, 210 insertions(+), 14 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index a5f5df77291..ee7072dce6a 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -63,9 +63,11 @@
 #include "optimizer/optimizer.h"
 #include "pgstat.h"
 #include "storage/bufmgr.h"
+#include "storage/ipc.h"
 #include "storage/lmgr.h"
 #include "storage/predicate.h"
 #include "storage/proc.h"
+#include "storage/subsystems.h"
 #include "utils/acl.h"
 #include "utils/fmgroids.h"
 #include "utils/guc.h"
@@ -79,6 +81,32 @@
 #include "utils/syscache.h"
 #include "utils/wait_event_types.h"
 
+
+/* Shared memory layout for REPACK */
+typedef struct RepackWorkerInfo
+{
+	bool		ri_in_use;
+	pid_t		ri_backendpid;
+	Oid			ri_dbid;
+	Oid			ri_relid;
+	Oid			ri_toastrelid;
+} RepackWorkerInfo;
+
+typedef struct
+{
+	bool		re_useless;
+	RepackWorkerInfo re_workerinfo[FLEXIBLE_ARRAY_MEMBER];
+} RepackShmemStruct;
+
+static RepackShmemStruct *RepackShmem;
+
+typedef struct RepackCleanupContext
+{
+	bool		concurrent;
+	int			workerindex;
+} RepackCleanupContext;
+
+
 /*
  * This struct is used to pass around the information on tables to be
  * clustered. We need this so we can make a list of them when invoked without
@@ -90,6 +118,7 @@ typedef struct
 	Oid			indexOid;
 } RelToCluster;
 
+
 /*
  * The first file exported by the decoding worker must contain a snapshot, the
  * following ones contain the data changes.
@@ -166,6 +195,10 @@ static List *get_tables_to_repack_partitioned(RepackCommand cmd,
 											  MemoryContext permcxt);
 static bool repack_is_permitted_for_relation(RepackCommand cmd,
 											 Oid relid, Oid userid);
+static void RepackCleanup(RepackCleanupContext *context);
+static void RepackCleanupCb(int code, Datum arg);
+static void RepackShmemRequest(void *arg);
+static void RepackShmemInit(void *arg);
 
 static void apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt);
 static void apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
@@ -210,6 +243,11 @@ static void ProcessRepackMessage(StringInfo msg);
 static const char *RepackCommandAsString(RepackCommand cmd);
 
 
+const ShmemCallbacks RepackShmemCallbacks = {
+	.request_fn = RepackShmemRequest,
+	.init_fn = RepackShmemInit,
+};
+
 /*
  * The repack code allows for processing multiple tables at once. Because
  * of this, we cannot just run everything on a single transaction, or we
@@ -514,6 +552,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 	Oid			tableOid = RelationGetRelid(OldHeap);
 	Relation	index;
 	LOCKMODE	lmode;
+	RepackCleanupContext context;
 	Oid			save_userid;
 	int			save_sec_context;
 	int			save_nestlevel;
@@ -660,24 +699,43 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 		TransferPredicateLocksToHeapRelation(OldHeap);
 
 	/* rebuild_relation does all the dirty work */
-	PG_TRY();
-	{
-		rebuild_relation(OldHeap, index, verbose, ident_idx);
-	}
-	PG_FINALLY();
+	context.concurrent = concurrent;
+
+	PG_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
 	{
 		if (concurrent)
 		{
-			/*
-			 * Since during normal operation the worker was already asked to
-			 * exit, stopping it explicitly is especially important on ERROR.
-			 * However it still seems a good practice to make sure that the
-			 * worker never survives the REPACK command.
-			 */
-			stop_repack_decoding_worker();
+			bool		freefound = false;
+
+			LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+			for (int i = 0; i < max_repack_replication_slots; i++)
+			{
+				RepackWorkerInfo *worker;
+
+				if (RepackShmem->re_workerinfo[i].ri_in_use)
+					continue;
+
+				freefound = true;
+				worker = &RepackShmem->re_workerinfo[i];
+				context.workerindex = i;
+
+				worker->ri_in_use = true;
+				worker->ri_backendpid = MyProcPid;
+				worker->ri_dbid = MyDatabaseId;
+				worker->ri_relid = RelationGetRelid(OldHeap);
+				worker->ri_toastrelid = OldHeap->rd_rel->reltoastrelid;
+				break;
+			}
+			if (!freefound)
+				elog(ERROR, "could not find free repack entry");
+			LWLockRelease(RepackLock);
 		}
+
+		rebuild_relation(OldHeap, index, verbose, ident_idx);
 	}
-	PG_END_TRY();
+	PG_END_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
+
+	RepackCleanup(&context);
 
 	/* rebuild_relation closes OldHeap, and index if valid */
 
@@ -691,6 +749,117 @@ out:
 	pgstat_progress_end_command();
 }
 
+/*
+ * Return whether any backend is running concurrent REPACK on the given table
+ * (which could be a toast table).
+ */
+bool
+is_table_under_repack(Oid databaseId, Oid relid)
+{
+	bool		retval = false;
+
+	LWLockAcquire(RepackLock, LW_SHARED);
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		RepackWorkerInfo *rworker;
+
+		if (!RepackShmem->re_workerinfo[i].ri_in_use)
+			continue;
+
+		rworker = &RepackShmem->re_workerinfo[i];
+		if (rworker->ri_dbid == MyDatabaseId &&
+			(rworker->ri_relid == relid ||
+			 rworker->ri_toastrelid == relid))
+			retval = true;
+	}
+	LWLockRelease(RepackLock);
+
+	return retval;
+}
+
+/*
+ * Remove ourselves from the workerinfo array.
+ */
+static void
+RepackCleanup(RepackCleanupContext *context)
+{
+	if (context->concurrent)
+	{
+		RepackWorkerInfo *worker;
+
+		/*
+		 * The worker would normally terminate on its own when the work is
+		 * done, but make sure we signal it just in case.
+		 */
+		stop_repack_decoding_worker();
+
+		/*
+		 * also, make sure we stop advertising the relation we were repacking,
+		 * so that autovacuum reverts to handling it normally.
+		 */
+		LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+
+		worker = &RepackShmem->re_workerinfo[context->workerindex];
+		Assert(worker->ri_backendpid == MyProcPid);
+		worker->ri_in_use = false;
+		worker->ri_backendpid = 0;
+		worker->ri_dbid = InvalidOid;
+		worker->ri_relid = InvalidOid;
+		worker->ri_toastrelid = InvalidOid;
+		LWLockRelease(RepackLock);
+	}
+}
+
+/*
+ * RepackCleanup wrapped as an on_shmem_exit callback function
+ */
+static void
+RepackCleanupCb(int code, Datum arg)
+{
+	RepackCleanup((RepackCleanupContext *) DatumGetPointer(arg));
+}
+
+/*
+ * RepackShmemRequest
+ *		Register shared memory space needed for repack
+ */
+static void
+RepackShmemRequest(void *arg)
+{
+	Size		size;
+
+	/*
+	 * Need the fixed struct and the array of RepackWorkerInfo.
+	 */
+	size = sizeof(RepackShmemStruct);
+	size = MAXALIGN(size);
+	size = add_size(size, mul_size(max_repack_replication_slots,
+								   sizeof(RepackWorkerInfo)));
+
+	ShmemRequestStruct(.name = "Repack Data",
+					   .size = size,
+					   .ptr = (void **) &RepackShmem,
+		);
+}
+
+static void
+RepackShmemInit(void *arg)
+{
+	RepackWorkerInfo *reinfo;
+
+	reinfo = (RepackWorkerInfo *) ((char *) RepackShmem +
+								   MAXALIGN(sizeof(RepackShmemStruct)));
+
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		reinfo[i].ri_in_use = false;
+		reinfo[i].ri_backendpid = 0;
+		reinfo[i].ri_dbid = InvalidOid;
+		reinfo[i].ri_relid = InvalidOid;
+		reinfo[i].ri_toastrelid = InvalidOid;
+	}
+}
+
 /*
  * Check if the table (and its index) still meets the requirements of
  * cluster_rel().
diff --git a/src/backend/postmaster/autovacuum.c b/src/backend/postmaster/autovacuum.c
index bd626a16363..080c64ea3c8 100644
--- a/src/backend/postmaster/autovacuum.c
+++ b/src/backend/postmaster/autovacuum.c
@@ -78,6 +78,7 @@
 #include "catalog/namespace.h"
 #include "catalog/pg_database.h"
 #include "catalog/pg_namespace.h"
+#include "commands/repack.h"
 #include "commands/vacuum.h"
 #include "common/int.h"
 #include "funcapi.h"
@@ -2422,6 +2423,25 @@ do_autovacuum(void)
 			}
 		}
 		LWLockRelease(AutovacuumLock);
+
+		/*
+		 * Similarly, if the table is being processed by concurrent repack,
+		 * skip it (but make a note of that).  We wouldn't be able to acquire
+		 * its lock anyway.
+		 */
+		if (!skipit)
+		{
+			MemoryContextSwitchTo(PortalContext);
+
+			skipit = is_table_under_repack(MyDatabaseId, relid);
+			if (skipit)
+				ereport(LOG,
+						errmsg("skipping table \"%s.%s.%s\" because it's being repacked in concurrent mode",
+							   get_database_name(MyDatabaseId),
+							   get_namespace_name(get_rel_namespace(relid)),
+							   get_rel_name(relid)));
+		}
+
 		if (skipit)
 		{
 			LWLockRelease(AutovacuumScheduleLock);
diff --git a/src/backend/utils/activity/wait_event_names.txt b/src/backend/utils/activity/wait_event_names.txt
index 7bda5298558..e206304f204 100644
--- a/src/backend/utils/activity/wait_event_names.txt
+++ b/src/backend/utils/activity/wait_event_names.txt
@@ -332,6 +332,7 @@ SInvalWrite	"Waiting to add a message to the shared catalog invalidation queue."
 WALBufMapping	"Waiting to replace a page in WAL buffers."
 WALWrite	"Waiting for WAL buffers to be written to disk."
 ControlFile	"Waiting to read or update the <filename>pg_control</filename> file or create a new WAL file."
+Repack	"Waiting to read or update tables in process by concurrent repack."
 MultiXactGen	"Waiting to read or update shared multixact state."
 RelCacheInit	"Waiting to read or update a <filename>pg_internal.init</filename> relation cache initialization file."
 CheckpointerComm	"Waiting to manage fsync requests."
diff --git a/src/include/commands/repack.h b/src/include/commands/repack.h
index fd16e74b179..be7d38b5fae 100644
--- a/src/include/commands/repack.h
+++ b/src/include/commands/repack.h
@@ -42,6 +42,8 @@ extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel);
 
 extern void cluster_rel(RepackCommand command, Relation OldHeap, Oid indexOid,
 						ClusterParams *params, bool isTopLevel);
+extern bool is_table_under_repack(Oid databaseId, Oid relid);
+
 extern void check_index_is_clusterable(Relation OldHeap, Oid indexOid,
 									   LOCKMODE lockmode);
 extern void mark_index_clustered(Relation rel, Oid indexOid, bool is_internal);
diff --git a/src/include/storage/lwlocklist.h b/src/include/storage/lwlocklist.h
index af8553bcb6c..3f08f4a15d4 100644
--- a/src/include/storage/lwlocklist.h
+++ b/src/include/storage/lwlocklist.h
@@ -41,7 +41,7 @@ PG_LWLOCK(6, SInvalWrite)
 PG_LWLOCK(7, WALBufMapping)
 PG_LWLOCK(8, WALWrite)
 PG_LWLOCK(9, ControlFile)
-/* 10 was CheckpointLock */
+PG_LWLOCK(10, Repack)
 /* 11 was XactSLRULock */
 /* 12 was SubtransSLRULock */
 PG_LWLOCK(13, MultiXactGen)
diff --git a/src/include/storage/subsystemlist.h b/src/include/storage/subsystemlist.h
index 9ad619080be..4e683b8b0a8 100644
--- a/src/include/storage/subsystemlist.h
+++ b/src/include/storage/subsystemlist.h
@@ -72,6 +72,7 @@ PG_SHMEM_SUBSYSTEM(WalSummarizerShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(PgArchShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(ApplyLauncherShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(SlotSyncShmemCallbacks)
+PG_SHMEM_SUBSYSTEM(RepackShmemCallbacks)
 
 /* other modules that need some shared memory space */
 PG_SHMEM_SUBSYSTEM(BTreeShmemCallbacks)
diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list
index 637c669a146..d019e03aaf1 100644
--- a/src/tools/pgindent/typedefs.list
+++ b/src/tools/pgindent/typedefs.list
@@ -2639,9 +2639,12 @@ ReorderBufferTupleCidEnt
 ReorderBufferTupleCidKey
 ReorderBufferUpdateProgressTxnCB
 ReorderTuple
+RepackCleanupContext
 RepackCommand
 RepackDecodingState
+RepackShmemStruct
 RepackStmt
+RepackWorkerInfo
 ReparameterizeForeignPathByChild_function
 ReplOriginId
 ReplOriginXactState
-- 
2.47.3


--kdrcpfmkbkc4lqhu--





^ permalink  raw  reply  [nested|flat] 223+ messages in thread

* [PATCH 2/2] Publish list of tables being repacked in shared memory
@ 2026-04-07 20:29 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 223+ messages in thread

From: Álvaro Herrera @ 2026-04-07 20:29 UTC (permalink / raw)

Use it in autovacuum to skip processing tables that are being repacked.
This is mostly to avoid repeated attempts to process such tables, which
would fail due to the special deadlock checker behavior for repack.

Author: Álvaro Herrera <[email protected]>
Discussion: https://postgr.es/m/[email protected]
---
 src/backend/commands/repack.c                 | 195 ++++++++++++++++--
 src/backend/postmaster/autovacuum.c           |  20 ++
 .../utils/activity/wait_event_names.txt       |   1 +
 src/include/commands/repack.h                 |   2 +
 src/include/storage/lwlocklist.h              |   2 +-
 src/include/storage/subsystemlist.h           |   1 +
 src/tools/pgindent/typedefs.list              |   3 +
 7 files changed, 210 insertions(+), 14 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index a5f5df77291..ee7072dce6a 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -63,9 +63,11 @@
 #include "optimizer/optimizer.h"
 #include "pgstat.h"
 #include "storage/bufmgr.h"
+#include "storage/ipc.h"
 #include "storage/lmgr.h"
 #include "storage/predicate.h"
 #include "storage/proc.h"
+#include "storage/subsystems.h"
 #include "utils/acl.h"
 #include "utils/fmgroids.h"
 #include "utils/guc.h"
@@ -79,6 +81,32 @@
 #include "utils/syscache.h"
 #include "utils/wait_event_types.h"
 
+
+/* Shared memory layout for REPACK */
+typedef struct RepackWorkerInfo
+{
+	bool		ri_in_use;
+	pid_t		ri_backendpid;
+	Oid			ri_dbid;
+	Oid			ri_relid;
+	Oid			ri_toastrelid;
+} RepackWorkerInfo;
+
+typedef struct
+{
+	bool		re_useless;
+	RepackWorkerInfo re_workerinfo[FLEXIBLE_ARRAY_MEMBER];
+} RepackShmemStruct;
+
+static RepackShmemStruct *RepackShmem;
+
+typedef struct RepackCleanupContext
+{
+	bool		concurrent;
+	int			workerindex;
+} RepackCleanupContext;
+
+
 /*
  * This struct is used to pass around the information on tables to be
  * clustered. We need this so we can make a list of them when invoked without
@@ -90,6 +118,7 @@ typedef struct
 	Oid			indexOid;
 } RelToCluster;
 
+
 /*
  * The first file exported by the decoding worker must contain a snapshot, the
  * following ones contain the data changes.
@@ -166,6 +195,10 @@ static List *get_tables_to_repack_partitioned(RepackCommand cmd,
 											  MemoryContext permcxt);
 static bool repack_is_permitted_for_relation(RepackCommand cmd,
 											 Oid relid, Oid userid);
+static void RepackCleanup(RepackCleanupContext *context);
+static void RepackCleanupCb(int code, Datum arg);
+static void RepackShmemRequest(void *arg);
+static void RepackShmemInit(void *arg);
 
 static void apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt);
 static void apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
@@ -210,6 +243,11 @@ static void ProcessRepackMessage(StringInfo msg);
 static const char *RepackCommandAsString(RepackCommand cmd);
 
 
+const ShmemCallbacks RepackShmemCallbacks = {
+	.request_fn = RepackShmemRequest,
+	.init_fn = RepackShmemInit,
+};
+
 /*
  * The repack code allows for processing multiple tables at once. Because
  * of this, we cannot just run everything on a single transaction, or we
@@ -514,6 +552,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 	Oid			tableOid = RelationGetRelid(OldHeap);
 	Relation	index;
 	LOCKMODE	lmode;
+	RepackCleanupContext context;
 	Oid			save_userid;
 	int			save_sec_context;
 	int			save_nestlevel;
@@ -660,24 +699,43 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 		TransferPredicateLocksToHeapRelation(OldHeap);
 
 	/* rebuild_relation does all the dirty work */
-	PG_TRY();
-	{
-		rebuild_relation(OldHeap, index, verbose, ident_idx);
-	}
-	PG_FINALLY();
+	context.concurrent = concurrent;
+
+	PG_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
 	{
 		if (concurrent)
 		{
-			/*
-			 * Since during normal operation the worker was already asked to
-			 * exit, stopping it explicitly is especially important on ERROR.
-			 * However it still seems a good practice to make sure that the
-			 * worker never survives the REPACK command.
-			 */
-			stop_repack_decoding_worker();
+			bool		freefound = false;
+
+			LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+			for (int i = 0; i < max_repack_replication_slots; i++)
+			{
+				RepackWorkerInfo *worker;
+
+				if (RepackShmem->re_workerinfo[i].ri_in_use)
+					continue;
+
+				freefound = true;
+				worker = &RepackShmem->re_workerinfo[i];
+				context.workerindex = i;
+
+				worker->ri_in_use = true;
+				worker->ri_backendpid = MyProcPid;
+				worker->ri_dbid = MyDatabaseId;
+				worker->ri_relid = RelationGetRelid(OldHeap);
+				worker->ri_toastrelid = OldHeap->rd_rel->reltoastrelid;
+				break;
+			}
+			if (!freefound)
+				elog(ERROR, "could not find free repack entry");
+			LWLockRelease(RepackLock);
 		}
+
+		rebuild_relation(OldHeap, index, verbose, ident_idx);
 	}
-	PG_END_TRY();
+	PG_END_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
+
+	RepackCleanup(&context);
 
 	/* rebuild_relation closes OldHeap, and index if valid */
 
@@ -691,6 +749,117 @@ out:
 	pgstat_progress_end_command();
 }
 
+/*
+ * Return whether any backend is running concurrent REPACK on the given table
+ * (which could be a toast table).
+ */
+bool
+is_table_under_repack(Oid databaseId, Oid relid)
+{
+	bool		retval = false;
+
+	LWLockAcquire(RepackLock, LW_SHARED);
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		RepackWorkerInfo *rworker;
+
+		if (!RepackShmem->re_workerinfo[i].ri_in_use)
+			continue;
+
+		rworker = &RepackShmem->re_workerinfo[i];
+		if (rworker->ri_dbid == MyDatabaseId &&
+			(rworker->ri_relid == relid ||
+			 rworker->ri_toastrelid == relid))
+			retval = true;
+	}
+	LWLockRelease(RepackLock);
+
+	return retval;
+}
+
+/*
+ * Remove ourselves from the workerinfo array.
+ */
+static void
+RepackCleanup(RepackCleanupContext *context)
+{
+	if (context->concurrent)
+	{
+		RepackWorkerInfo *worker;
+
+		/*
+		 * The worker would normally terminate on its own when the work is
+		 * done, but make sure we signal it just in case.
+		 */
+		stop_repack_decoding_worker();
+
+		/*
+		 * also, make sure we stop advertising the relation we were repacking,
+		 * so that autovacuum reverts to handling it normally.
+		 */
+		LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+
+		worker = &RepackShmem->re_workerinfo[context->workerindex];
+		Assert(worker->ri_backendpid == MyProcPid);
+		worker->ri_in_use = false;
+		worker->ri_backendpid = 0;
+		worker->ri_dbid = InvalidOid;
+		worker->ri_relid = InvalidOid;
+		worker->ri_toastrelid = InvalidOid;
+		LWLockRelease(RepackLock);
+	}
+}
+
+/*
+ * RepackCleanup wrapped as an on_shmem_exit callback function
+ */
+static void
+RepackCleanupCb(int code, Datum arg)
+{
+	RepackCleanup((RepackCleanupContext *) DatumGetPointer(arg));
+}
+
+/*
+ * RepackShmemRequest
+ *		Register shared memory space needed for repack
+ */
+static void
+RepackShmemRequest(void *arg)
+{
+	Size		size;
+
+	/*
+	 * Need the fixed struct and the array of RepackWorkerInfo.
+	 */
+	size = sizeof(RepackShmemStruct);
+	size = MAXALIGN(size);
+	size = add_size(size, mul_size(max_repack_replication_slots,
+								   sizeof(RepackWorkerInfo)));
+
+	ShmemRequestStruct(.name = "Repack Data",
+					   .size = size,
+					   .ptr = (void **) &RepackShmem,
+		);
+}
+
+static void
+RepackShmemInit(void *arg)
+{
+	RepackWorkerInfo *reinfo;
+
+	reinfo = (RepackWorkerInfo *) ((char *) RepackShmem +
+								   MAXALIGN(sizeof(RepackShmemStruct)));
+
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		reinfo[i].ri_in_use = false;
+		reinfo[i].ri_backendpid = 0;
+		reinfo[i].ri_dbid = InvalidOid;
+		reinfo[i].ri_relid = InvalidOid;
+		reinfo[i].ri_toastrelid = InvalidOid;
+	}
+}
+
 /*
  * Check if the table (and its index) still meets the requirements of
  * cluster_rel().
diff --git a/src/backend/postmaster/autovacuum.c b/src/backend/postmaster/autovacuum.c
index bd626a16363..080c64ea3c8 100644
--- a/src/backend/postmaster/autovacuum.c
+++ b/src/backend/postmaster/autovacuum.c
@@ -78,6 +78,7 @@
 #include "catalog/namespace.h"
 #include "catalog/pg_database.h"
 #include "catalog/pg_namespace.h"
+#include "commands/repack.h"
 #include "commands/vacuum.h"
 #include "common/int.h"
 #include "funcapi.h"
@@ -2422,6 +2423,25 @@ do_autovacuum(void)
 			}
 		}
 		LWLockRelease(AutovacuumLock);
+
+		/*
+		 * Similarly, if the table is being processed by concurrent repack,
+		 * skip it (but make a note of that).  We wouldn't be able to acquire
+		 * its lock anyway.
+		 */
+		if (!skipit)
+		{
+			MemoryContextSwitchTo(PortalContext);
+
+			skipit = is_table_under_repack(MyDatabaseId, relid);
+			if (skipit)
+				ereport(LOG,
+						errmsg("skipping table \"%s.%s.%s\" because it's being repacked in concurrent mode",
+							   get_database_name(MyDatabaseId),
+							   get_namespace_name(get_rel_namespace(relid)),
+							   get_rel_name(relid)));
+		}
+
 		if (skipit)
 		{
 			LWLockRelease(AutovacuumScheduleLock);
diff --git a/src/backend/utils/activity/wait_event_names.txt b/src/backend/utils/activity/wait_event_names.txt
index 7bda5298558..e206304f204 100644
--- a/src/backend/utils/activity/wait_event_names.txt
+++ b/src/backend/utils/activity/wait_event_names.txt
@@ -332,6 +332,7 @@ SInvalWrite	"Waiting to add a message to the shared catalog invalidation queue."
 WALBufMapping	"Waiting to replace a page in WAL buffers."
 WALWrite	"Waiting for WAL buffers to be written to disk."
 ControlFile	"Waiting to read or update the <filename>pg_control</filename> file or create a new WAL file."
+Repack	"Waiting to read or update tables in process by concurrent repack."
 MultiXactGen	"Waiting to read or update shared multixact state."
 RelCacheInit	"Waiting to read or update a <filename>pg_internal.init</filename> relation cache initialization file."
 CheckpointerComm	"Waiting to manage fsync requests."
diff --git a/src/include/commands/repack.h b/src/include/commands/repack.h
index fd16e74b179..be7d38b5fae 100644
--- a/src/include/commands/repack.h
+++ b/src/include/commands/repack.h
@@ -42,6 +42,8 @@ extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel);
 
 extern void cluster_rel(RepackCommand command, Relation OldHeap, Oid indexOid,
 						ClusterParams *params, bool isTopLevel);
+extern bool is_table_under_repack(Oid databaseId, Oid relid);
+
 extern void check_index_is_clusterable(Relation OldHeap, Oid indexOid,
 									   LOCKMODE lockmode);
 extern void mark_index_clustered(Relation rel, Oid indexOid, bool is_internal);
diff --git a/src/include/storage/lwlocklist.h b/src/include/storage/lwlocklist.h
index af8553bcb6c..3f08f4a15d4 100644
--- a/src/include/storage/lwlocklist.h
+++ b/src/include/storage/lwlocklist.h
@@ -41,7 +41,7 @@ PG_LWLOCK(6, SInvalWrite)
 PG_LWLOCK(7, WALBufMapping)
 PG_LWLOCK(8, WALWrite)
 PG_LWLOCK(9, ControlFile)
-/* 10 was CheckpointLock */
+PG_LWLOCK(10, Repack)
 /* 11 was XactSLRULock */
 /* 12 was SubtransSLRULock */
 PG_LWLOCK(13, MultiXactGen)
diff --git a/src/include/storage/subsystemlist.h b/src/include/storage/subsystemlist.h
index 9ad619080be..4e683b8b0a8 100644
--- a/src/include/storage/subsystemlist.h
+++ b/src/include/storage/subsystemlist.h
@@ -72,6 +72,7 @@ PG_SHMEM_SUBSYSTEM(WalSummarizerShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(PgArchShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(ApplyLauncherShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(SlotSyncShmemCallbacks)
+PG_SHMEM_SUBSYSTEM(RepackShmemCallbacks)
 
 /* other modules that need some shared memory space */
 PG_SHMEM_SUBSYSTEM(BTreeShmemCallbacks)
diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list
index 637c669a146..d019e03aaf1 100644
--- a/src/tools/pgindent/typedefs.list
+++ b/src/tools/pgindent/typedefs.list
@@ -2639,9 +2639,12 @@ ReorderBufferTupleCidEnt
 ReorderBufferTupleCidKey
 ReorderBufferUpdateProgressTxnCB
 ReorderTuple
+RepackCleanupContext
 RepackCommand
 RepackDecodingState
+RepackShmemStruct
 RepackStmt
+RepackWorkerInfo
 ReparameterizeForeignPathByChild_function
 ReplOriginId
 ReplOriginXactState
-- 
2.47.3


--kdrcpfmkbkc4lqhu--





^ permalink  raw  reply  [nested|flat] 223+ messages in thread

* [PATCH 2/2] Publish list of tables being repacked in shared memory
@ 2026-04-07 20:29 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 223+ messages in thread

From: Álvaro Herrera @ 2026-04-07 20:29 UTC (permalink / raw)

Use it in autovacuum to skip processing tables that are being repacked.
This is mostly to avoid repeated attempts to process such tables, which
would fail due to the special deadlock checker behavior for repack.

Author: Álvaro Herrera <[email protected]>
Discussion: https://postgr.es/m/[email protected]
---
 src/backend/commands/repack.c                 | 195 ++++++++++++++++--
 src/backend/postmaster/autovacuum.c           |  20 ++
 .../utils/activity/wait_event_names.txt       |   1 +
 src/include/commands/repack.h                 |   2 +
 src/include/storage/lwlocklist.h              |   2 +-
 src/include/storage/subsystemlist.h           |   1 +
 src/tools/pgindent/typedefs.list              |   3 +
 7 files changed, 210 insertions(+), 14 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index a5f5df77291..ee7072dce6a 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -63,9 +63,11 @@
 #include "optimizer/optimizer.h"
 #include "pgstat.h"
 #include "storage/bufmgr.h"
+#include "storage/ipc.h"
 #include "storage/lmgr.h"
 #include "storage/predicate.h"
 #include "storage/proc.h"
+#include "storage/subsystems.h"
 #include "utils/acl.h"
 #include "utils/fmgroids.h"
 #include "utils/guc.h"
@@ -79,6 +81,32 @@
 #include "utils/syscache.h"
 #include "utils/wait_event_types.h"
 
+
+/* Shared memory layout for REPACK */
+typedef struct RepackWorkerInfo
+{
+	bool		ri_in_use;
+	pid_t		ri_backendpid;
+	Oid			ri_dbid;
+	Oid			ri_relid;
+	Oid			ri_toastrelid;
+} RepackWorkerInfo;
+
+typedef struct
+{
+	bool		re_useless;
+	RepackWorkerInfo re_workerinfo[FLEXIBLE_ARRAY_MEMBER];
+} RepackShmemStruct;
+
+static RepackShmemStruct *RepackShmem;
+
+typedef struct RepackCleanupContext
+{
+	bool		concurrent;
+	int			workerindex;
+} RepackCleanupContext;
+
+
 /*
  * This struct is used to pass around the information on tables to be
  * clustered. We need this so we can make a list of them when invoked without
@@ -90,6 +118,7 @@ typedef struct
 	Oid			indexOid;
 } RelToCluster;
 
+
 /*
  * The first file exported by the decoding worker must contain a snapshot, the
  * following ones contain the data changes.
@@ -166,6 +195,10 @@ static List *get_tables_to_repack_partitioned(RepackCommand cmd,
 											  MemoryContext permcxt);
 static bool repack_is_permitted_for_relation(RepackCommand cmd,
 											 Oid relid, Oid userid);
+static void RepackCleanup(RepackCleanupContext *context);
+static void RepackCleanupCb(int code, Datum arg);
+static void RepackShmemRequest(void *arg);
+static void RepackShmemInit(void *arg);
 
 static void apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt);
 static void apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
@@ -210,6 +243,11 @@ static void ProcessRepackMessage(StringInfo msg);
 static const char *RepackCommandAsString(RepackCommand cmd);
 
 
+const ShmemCallbacks RepackShmemCallbacks = {
+	.request_fn = RepackShmemRequest,
+	.init_fn = RepackShmemInit,
+};
+
 /*
  * The repack code allows for processing multiple tables at once. Because
  * of this, we cannot just run everything on a single transaction, or we
@@ -514,6 +552,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 	Oid			tableOid = RelationGetRelid(OldHeap);
 	Relation	index;
 	LOCKMODE	lmode;
+	RepackCleanupContext context;
 	Oid			save_userid;
 	int			save_sec_context;
 	int			save_nestlevel;
@@ -660,24 +699,43 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 		TransferPredicateLocksToHeapRelation(OldHeap);
 
 	/* rebuild_relation does all the dirty work */
-	PG_TRY();
-	{
-		rebuild_relation(OldHeap, index, verbose, ident_idx);
-	}
-	PG_FINALLY();
+	context.concurrent = concurrent;
+
+	PG_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
 	{
 		if (concurrent)
 		{
-			/*
-			 * Since during normal operation the worker was already asked to
-			 * exit, stopping it explicitly is especially important on ERROR.
-			 * However it still seems a good practice to make sure that the
-			 * worker never survives the REPACK command.
-			 */
-			stop_repack_decoding_worker();
+			bool		freefound = false;
+
+			LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+			for (int i = 0; i < max_repack_replication_slots; i++)
+			{
+				RepackWorkerInfo *worker;
+
+				if (RepackShmem->re_workerinfo[i].ri_in_use)
+					continue;
+
+				freefound = true;
+				worker = &RepackShmem->re_workerinfo[i];
+				context.workerindex = i;
+
+				worker->ri_in_use = true;
+				worker->ri_backendpid = MyProcPid;
+				worker->ri_dbid = MyDatabaseId;
+				worker->ri_relid = RelationGetRelid(OldHeap);
+				worker->ri_toastrelid = OldHeap->rd_rel->reltoastrelid;
+				break;
+			}
+			if (!freefound)
+				elog(ERROR, "could not find free repack entry");
+			LWLockRelease(RepackLock);
 		}
+
+		rebuild_relation(OldHeap, index, verbose, ident_idx);
 	}
-	PG_END_TRY();
+	PG_END_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
+
+	RepackCleanup(&context);
 
 	/* rebuild_relation closes OldHeap, and index if valid */
 
@@ -691,6 +749,117 @@ out:
 	pgstat_progress_end_command();
 }
 
+/*
+ * Return whether any backend is running concurrent REPACK on the given table
+ * (which could be a toast table).
+ */
+bool
+is_table_under_repack(Oid databaseId, Oid relid)
+{
+	bool		retval = false;
+
+	LWLockAcquire(RepackLock, LW_SHARED);
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		RepackWorkerInfo *rworker;
+
+		if (!RepackShmem->re_workerinfo[i].ri_in_use)
+			continue;
+
+		rworker = &RepackShmem->re_workerinfo[i];
+		if (rworker->ri_dbid == MyDatabaseId &&
+			(rworker->ri_relid == relid ||
+			 rworker->ri_toastrelid == relid))
+			retval = true;
+	}
+	LWLockRelease(RepackLock);
+
+	return retval;
+}
+
+/*
+ * Remove ourselves from the workerinfo array.
+ */
+static void
+RepackCleanup(RepackCleanupContext *context)
+{
+	if (context->concurrent)
+	{
+		RepackWorkerInfo *worker;
+
+		/*
+		 * The worker would normally terminate on its own when the work is
+		 * done, but make sure we signal it just in case.
+		 */
+		stop_repack_decoding_worker();
+
+		/*
+		 * also, make sure we stop advertising the relation we were repacking,
+		 * so that autovacuum reverts to handling it normally.
+		 */
+		LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+
+		worker = &RepackShmem->re_workerinfo[context->workerindex];
+		Assert(worker->ri_backendpid == MyProcPid);
+		worker->ri_in_use = false;
+		worker->ri_backendpid = 0;
+		worker->ri_dbid = InvalidOid;
+		worker->ri_relid = InvalidOid;
+		worker->ri_toastrelid = InvalidOid;
+		LWLockRelease(RepackLock);
+	}
+}
+
+/*
+ * RepackCleanup wrapped as an on_shmem_exit callback function
+ */
+static void
+RepackCleanupCb(int code, Datum arg)
+{
+	RepackCleanup((RepackCleanupContext *) DatumGetPointer(arg));
+}
+
+/*
+ * RepackShmemRequest
+ *		Register shared memory space needed for repack
+ */
+static void
+RepackShmemRequest(void *arg)
+{
+	Size		size;
+
+	/*
+	 * Need the fixed struct and the array of RepackWorkerInfo.
+	 */
+	size = sizeof(RepackShmemStruct);
+	size = MAXALIGN(size);
+	size = add_size(size, mul_size(max_repack_replication_slots,
+								   sizeof(RepackWorkerInfo)));
+
+	ShmemRequestStruct(.name = "Repack Data",
+					   .size = size,
+					   .ptr = (void **) &RepackShmem,
+		);
+}
+
+static void
+RepackShmemInit(void *arg)
+{
+	RepackWorkerInfo *reinfo;
+
+	reinfo = (RepackWorkerInfo *) ((char *) RepackShmem +
+								   MAXALIGN(sizeof(RepackShmemStruct)));
+
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		reinfo[i].ri_in_use = false;
+		reinfo[i].ri_backendpid = 0;
+		reinfo[i].ri_dbid = InvalidOid;
+		reinfo[i].ri_relid = InvalidOid;
+		reinfo[i].ri_toastrelid = InvalidOid;
+	}
+}
+
 /*
  * Check if the table (and its index) still meets the requirements of
  * cluster_rel().
diff --git a/src/backend/postmaster/autovacuum.c b/src/backend/postmaster/autovacuum.c
index bd626a16363..080c64ea3c8 100644
--- a/src/backend/postmaster/autovacuum.c
+++ b/src/backend/postmaster/autovacuum.c
@@ -78,6 +78,7 @@
 #include "catalog/namespace.h"
 #include "catalog/pg_database.h"
 #include "catalog/pg_namespace.h"
+#include "commands/repack.h"
 #include "commands/vacuum.h"
 #include "common/int.h"
 #include "funcapi.h"
@@ -2422,6 +2423,25 @@ do_autovacuum(void)
 			}
 		}
 		LWLockRelease(AutovacuumLock);
+
+		/*
+		 * Similarly, if the table is being processed by concurrent repack,
+		 * skip it (but make a note of that).  We wouldn't be able to acquire
+		 * its lock anyway.
+		 */
+		if (!skipit)
+		{
+			MemoryContextSwitchTo(PortalContext);
+
+			skipit = is_table_under_repack(MyDatabaseId, relid);
+			if (skipit)
+				ereport(LOG,
+						errmsg("skipping table \"%s.%s.%s\" because it's being repacked in concurrent mode",
+							   get_database_name(MyDatabaseId),
+							   get_namespace_name(get_rel_namespace(relid)),
+							   get_rel_name(relid)));
+		}
+
 		if (skipit)
 		{
 			LWLockRelease(AutovacuumScheduleLock);
diff --git a/src/backend/utils/activity/wait_event_names.txt b/src/backend/utils/activity/wait_event_names.txt
index 7bda5298558..e206304f204 100644
--- a/src/backend/utils/activity/wait_event_names.txt
+++ b/src/backend/utils/activity/wait_event_names.txt
@@ -332,6 +332,7 @@ SInvalWrite	"Waiting to add a message to the shared catalog invalidation queue."
 WALBufMapping	"Waiting to replace a page in WAL buffers."
 WALWrite	"Waiting for WAL buffers to be written to disk."
 ControlFile	"Waiting to read or update the <filename>pg_control</filename> file or create a new WAL file."
+Repack	"Waiting to read or update tables in process by concurrent repack."
 MultiXactGen	"Waiting to read or update shared multixact state."
 RelCacheInit	"Waiting to read or update a <filename>pg_internal.init</filename> relation cache initialization file."
 CheckpointerComm	"Waiting to manage fsync requests."
diff --git a/src/include/commands/repack.h b/src/include/commands/repack.h
index fd16e74b179..be7d38b5fae 100644
--- a/src/include/commands/repack.h
+++ b/src/include/commands/repack.h
@@ -42,6 +42,8 @@ extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel);
 
 extern void cluster_rel(RepackCommand command, Relation OldHeap, Oid indexOid,
 						ClusterParams *params, bool isTopLevel);
+extern bool is_table_under_repack(Oid databaseId, Oid relid);
+
 extern void check_index_is_clusterable(Relation OldHeap, Oid indexOid,
 									   LOCKMODE lockmode);
 extern void mark_index_clustered(Relation rel, Oid indexOid, bool is_internal);
diff --git a/src/include/storage/lwlocklist.h b/src/include/storage/lwlocklist.h
index af8553bcb6c..3f08f4a15d4 100644
--- a/src/include/storage/lwlocklist.h
+++ b/src/include/storage/lwlocklist.h
@@ -41,7 +41,7 @@ PG_LWLOCK(6, SInvalWrite)
 PG_LWLOCK(7, WALBufMapping)
 PG_LWLOCK(8, WALWrite)
 PG_LWLOCK(9, ControlFile)
-/* 10 was CheckpointLock */
+PG_LWLOCK(10, Repack)
 /* 11 was XactSLRULock */
 /* 12 was SubtransSLRULock */
 PG_LWLOCK(13, MultiXactGen)
diff --git a/src/include/storage/subsystemlist.h b/src/include/storage/subsystemlist.h
index 9ad619080be..4e683b8b0a8 100644
--- a/src/include/storage/subsystemlist.h
+++ b/src/include/storage/subsystemlist.h
@@ -72,6 +72,7 @@ PG_SHMEM_SUBSYSTEM(WalSummarizerShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(PgArchShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(ApplyLauncherShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(SlotSyncShmemCallbacks)
+PG_SHMEM_SUBSYSTEM(RepackShmemCallbacks)
 
 /* other modules that need some shared memory space */
 PG_SHMEM_SUBSYSTEM(BTreeShmemCallbacks)
diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list
index 637c669a146..d019e03aaf1 100644
--- a/src/tools/pgindent/typedefs.list
+++ b/src/tools/pgindent/typedefs.list
@@ -2639,9 +2639,12 @@ ReorderBufferTupleCidEnt
 ReorderBufferTupleCidKey
 ReorderBufferUpdateProgressTxnCB
 ReorderTuple
+RepackCleanupContext
 RepackCommand
 RepackDecodingState
+RepackShmemStruct
 RepackStmt
+RepackWorkerInfo
 ReparameterizeForeignPathByChild_function
 ReplOriginId
 ReplOriginXactState
-- 
2.47.3


--kdrcpfmkbkc4lqhu--





^ permalink  raw  reply  [nested|flat] 223+ messages in thread

* [PATCH 2/2] Publish list of tables being repacked in shared memory
@ 2026-04-07 20:29 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 223+ messages in thread

From: Álvaro Herrera @ 2026-04-07 20:29 UTC (permalink / raw)

Use it in autovacuum to skip processing tables that are being repacked.
This is mostly to avoid repeated attempts to process such tables, which
would fail due to the special deadlock checker behavior for repack.

Author: Álvaro Herrera <[email protected]>
Discussion: https://postgr.es/m/[email protected]
---
 src/backend/commands/repack.c                 | 195 ++++++++++++++++--
 src/backend/postmaster/autovacuum.c           |  20 ++
 .../utils/activity/wait_event_names.txt       |   1 +
 src/include/commands/repack.h                 |   2 +
 src/include/storage/lwlocklist.h              |   2 +-
 src/include/storage/subsystemlist.h           |   1 +
 src/tools/pgindent/typedefs.list              |   3 +
 7 files changed, 210 insertions(+), 14 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index a5f5df77291..ee7072dce6a 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -63,9 +63,11 @@
 #include "optimizer/optimizer.h"
 #include "pgstat.h"
 #include "storage/bufmgr.h"
+#include "storage/ipc.h"
 #include "storage/lmgr.h"
 #include "storage/predicate.h"
 #include "storage/proc.h"
+#include "storage/subsystems.h"
 #include "utils/acl.h"
 #include "utils/fmgroids.h"
 #include "utils/guc.h"
@@ -79,6 +81,32 @@
 #include "utils/syscache.h"
 #include "utils/wait_event_types.h"
 
+
+/* Shared memory layout for REPACK */
+typedef struct RepackWorkerInfo
+{
+	bool		ri_in_use;
+	pid_t		ri_backendpid;
+	Oid			ri_dbid;
+	Oid			ri_relid;
+	Oid			ri_toastrelid;
+} RepackWorkerInfo;
+
+typedef struct
+{
+	bool		re_useless;
+	RepackWorkerInfo re_workerinfo[FLEXIBLE_ARRAY_MEMBER];
+} RepackShmemStruct;
+
+static RepackShmemStruct *RepackShmem;
+
+typedef struct RepackCleanupContext
+{
+	bool		concurrent;
+	int			workerindex;
+} RepackCleanupContext;
+
+
 /*
  * This struct is used to pass around the information on tables to be
  * clustered. We need this so we can make a list of them when invoked without
@@ -90,6 +118,7 @@ typedef struct
 	Oid			indexOid;
 } RelToCluster;
 
+
 /*
  * The first file exported by the decoding worker must contain a snapshot, the
  * following ones contain the data changes.
@@ -166,6 +195,10 @@ static List *get_tables_to_repack_partitioned(RepackCommand cmd,
 											  MemoryContext permcxt);
 static bool repack_is_permitted_for_relation(RepackCommand cmd,
 											 Oid relid, Oid userid);
+static void RepackCleanup(RepackCleanupContext *context);
+static void RepackCleanupCb(int code, Datum arg);
+static void RepackShmemRequest(void *arg);
+static void RepackShmemInit(void *arg);
 
 static void apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt);
 static void apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
@@ -210,6 +243,11 @@ static void ProcessRepackMessage(StringInfo msg);
 static const char *RepackCommandAsString(RepackCommand cmd);
 
 
+const ShmemCallbacks RepackShmemCallbacks = {
+	.request_fn = RepackShmemRequest,
+	.init_fn = RepackShmemInit,
+};
+
 /*
  * The repack code allows for processing multiple tables at once. Because
  * of this, we cannot just run everything on a single transaction, or we
@@ -514,6 +552,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 	Oid			tableOid = RelationGetRelid(OldHeap);
 	Relation	index;
 	LOCKMODE	lmode;
+	RepackCleanupContext context;
 	Oid			save_userid;
 	int			save_sec_context;
 	int			save_nestlevel;
@@ -660,24 +699,43 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 		TransferPredicateLocksToHeapRelation(OldHeap);
 
 	/* rebuild_relation does all the dirty work */
-	PG_TRY();
-	{
-		rebuild_relation(OldHeap, index, verbose, ident_idx);
-	}
-	PG_FINALLY();
+	context.concurrent = concurrent;
+
+	PG_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
 	{
 		if (concurrent)
 		{
-			/*
-			 * Since during normal operation the worker was already asked to
-			 * exit, stopping it explicitly is especially important on ERROR.
-			 * However it still seems a good practice to make sure that the
-			 * worker never survives the REPACK command.
-			 */
-			stop_repack_decoding_worker();
+			bool		freefound = false;
+
+			LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+			for (int i = 0; i < max_repack_replication_slots; i++)
+			{
+				RepackWorkerInfo *worker;
+
+				if (RepackShmem->re_workerinfo[i].ri_in_use)
+					continue;
+
+				freefound = true;
+				worker = &RepackShmem->re_workerinfo[i];
+				context.workerindex = i;
+
+				worker->ri_in_use = true;
+				worker->ri_backendpid = MyProcPid;
+				worker->ri_dbid = MyDatabaseId;
+				worker->ri_relid = RelationGetRelid(OldHeap);
+				worker->ri_toastrelid = OldHeap->rd_rel->reltoastrelid;
+				break;
+			}
+			if (!freefound)
+				elog(ERROR, "could not find free repack entry");
+			LWLockRelease(RepackLock);
 		}
+
+		rebuild_relation(OldHeap, index, verbose, ident_idx);
 	}
-	PG_END_TRY();
+	PG_END_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
+
+	RepackCleanup(&context);
 
 	/* rebuild_relation closes OldHeap, and index if valid */
 
@@ -691,6 +749,117 @@ out:
 	pgstat_progress_end_command();
 }
 
+/*
+ * Return whether any backend is running concurrent REPACK on the given table
+ * (which could be a toast table).
+ */
+bool
+is_table_under_repack(Oid databaseId, Oid relid)
+{
+	bool		retval = false;
+
+	LWLockAcquire(RepackLock, LW_SHARED);
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		RepackWorkerInfo *rworker;
+
+		if (!RepackShmem->re_workerinfo[i].ri_in_use)
+			continue;
+
+		rworker = &RepackShmem->re_workerinfo[i];
+		if (rworker->ri_dbid == MyDatabaseId &&
+			(rworker->ri_relid == relid ||
+			 rworker->ri_toastrelid == relid))
+			retval = true;
+	}
+	LWLockRelease(RepackLock);
+
+	return retval;
+}
+
+/*
+ * Remove ourselves from the workerinfo array.
+ */
+static void
+RepackCleanup(RepackCleanupContext *context)
+{
+	if (context->concurrent)
+	{
+		RepackWorkerInfo *worker;
+
+		/*
+		 * The worker would normally terminate on its own when the work is
+		 * done, but make sure we signal it just in case.
+		 */
+		stop_repack_decoding_worker();
+
+		/*
+		 * also, make sure we stop advertising the relation we were repacking,
+		 * so that autovacuum reverts to handling it normally.
+		 */
+		LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+
+		worker = &RepackShmem->re_workerinfo[context->workerindex];
+		Assert(worker->ri_backendpid == MyProcPid);
+		worker->ri_in_use = false;
+		worker->ri_backendpid = 0;
+		worker->ri_dbid = InvalidOid;
+		worker->ri_relid = InvalidOid;
+		worker->ri_toastrelid = InvalidOid;
+		LWLockRelease(RepackLock);
+	}
+}
+
+/*
+ * RepackCleanup wrapped as an on_shmem_exit callback function
+ */
+static void
+RepackCleanupCb(int code, Datum arg)
+{
+	RepackCleanup((RepackCleanupContext *) DatumGetPointer(arg));
+}
+
+/*
+ * RepackShmemRequest
+ *		Register shared memory space needed for repack
+ */
+static void
+RepackShmemRequest(void *arg)
+{
+	Size		size;
+
+	/*
+	 * Need the fixed struct and the array of RepackWorkerInfo.
+	 */
+	size = sizeof(RepackShmemStruct);
+	size = MAXALIGN(size);
+	size = add_size(size, mul_size(max_repack_replication_slots,
+								   sizeof(RepackWorkerInfo)));
+
+	ShmemRequestStruct(.name = "Repack Data",
+					   .size = size,
+					   .ptr = (void **) &RepackShmem,
+		);
+}
+
+static void
+RepackShmemInit(void *arg)
+{
+	RepackWorkerInfo *reinfo;
+
+	reinfo = (RepackWorkerInfo *) ((char *) RepackShmem +
+								   MAXALIGN(sizeof(RepackShmemStruct)));
+
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		reinfo[i].ri_in_use = false;
+		reinfo[i].ri_backendpid = 0;
+		reinfo[i].ri_dbid = InvalidOid;
+		reinfo[i].ri_relid = InvalidOid;
+		reinfo[i].ri_toastrelid = InvalidOid;
+	}
+}
+
 /*
  * Check if the table (and its index) still meets the requirements of
  * cluster_rel().
diff --git a/src/backend/postmaster/autovacuum.c b/src/backend/postmaster/autovacuum.c
index bd626a16363..080c64ea3c8 100644
--- a/src/backend/postmaster/autovacuum.c
+++ b/src/backend/postmaster/autovacuum.c
@@ -78,6 +78,7 @@
 #include "catalog/namespace.h"
 #include "catalog/pg_database.h"
 #include "catalog/pg_namespace.h"
+#include "commands/repack.h"
 #include "commands/vacuum.h"
 #include "common/int.h"
 #include "funcapi.h"
@@ -2422,6 +2423,25 @@ do_autovacuum(void)
 			}
 		}
 		LWLockRelease(AutovacuumLock);
+
+		/*
+		 * Similarly, if the table is being processed by concurrent repack,
+		 * skip it (but make a note of that).  We wouldn't be able to acquire
+		 * its lock anyway.
+		 */
+		if (!skipit)
+		{
+			MemoryContextSwitchTo(PortalContext);
+
+			skipit = is_table_under_repack(MyDatabaseId, relid);
+			if (skipit)
+				ereport(LOG,
+						errmsg("skipping table \"%s.%s.%s\" because it's being repacked in concurrent mode",
+							   get_database_name(MyDatabaseId),
+							   get_namespace_name(get_rel_namespace(relid)),
+							   get_rel_name(relid)));
+		}
+
 		if (skipit)
 		{
 			LWLockRelease(AutovacuumScheduleLock);
diff --git a/src/backend/utils/activity/wait_event_names.txt b/src/backend/utils/activity/wait_event_names.txt
index 7bda5298558..e206304f204 100644
--- a/src/backend/utils/activity/wait_event_names.txt
+++ b/src/backend/utils/activity/wait_event_names.txt
@@ -332,6 +332,7 @@ SInvalWrite	"Waiting to add a message to the shared catalog invalidation queue."
 WALBufMapping	"Waiting to replace a page in WAL buffers."
 WALWrite	"Waiting for WAL buffers to be written to disk."
 ControlFile	"Waiting to read or update the <filename>pg_control</filename> file or create a new WAL file."
+Repack	"Waiting to read or update tables in process by concurrent repack."
 MultiXactGen	"Waiting to read or update shared multixact state."
 RelCacheInit	"Waiting to read or update a <filename>pg_internal.init</filename> relation cache initialization file."
 CheckpointerComm	"Waiting to manage fsync requests."
diff --git a/src/include/commands/repack.h b/src/include/commands/repack.h
index fd16e74b179..be7d38b5fae 100644
--- a/src/include/commands/repack.h
+++ b/src/include/commands/repack.h
@@ -42,6 +42,8 @@ extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel);
 
 extern void cluster_rel(RepackCommand command, Relation OldHeap, Oid indexOid,
 						ClusterParams *params, bool isTopLevel);
+extern bool is_table_under_repack(Oid databaseId, Oid relid);
+
 extern void check_index_is_clusterable(Relation OldHeap, Oid indexOid,
 									   LOCKMODE lockmode);
 extern void mark_index_clustered(Relation rel, Oid indexOid, bool is_internal);
diff --git a/src/include/storage/lwlocklist.h b/src/include/storage/lwlocklist.h
index af8553bcb6c..3f08f4a15d4 100644
--- a/src/include/storage/lwlocklist.h
+++ b/src/include/storage/lwlocklist.h
@@ -41,7 +41,7 @@ PG_LWLOCK(6, SInvalWrite)
 PG_LWLOCK(7, WALBufMapping)
 PG_LWLOCK(8, WALWrite)
 PG_LWLOCK(9, ControlFile)
-/* 10 was CheckpointLock */
+PG_LWLOCK(10, Repack)
 /* 11 was XactSLRULock */
 /* 12 was SubtransSLRULock */
 PG_LWLOCK(13, MultiXactGen)
diff --git a/src/include/storage/subsystemlist.h b/src/include/storage/subsystemlist.h
index 9ad619080be..4e683b8b0a8 100644
--- a/src/include/storage/subsystemlist.h
+++ b/src/include/storage/subsystemlist.h
@@ -72,6 +72,7 @@ PG_SHMEM_SUBSYSTEM(WalSummarizerShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(PgArchShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(ApplyLauncherShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(SlotSyncShmemCallbacks)
+PG_SHMEM_SUBSYSTEM(RepackShmemCallbacks)
 
 /* other modules that need some shared memory space */
 PG_SHMEM_SUBSYSTEM(BTreeShmemCallbacks)
diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list
index 637c669a146..d019e03aaf1 100644
--- a/src/tools/pgindent/typedefs.list
+++ b/src/tools/pgindent/typedefs.list
@@ -2639,9 +2639,12 @@ ReorderBufferTupleCidEnt
 ReorderBufferTupleCidKey
 ReorderBufferUpdateProgressTxnCB
 ReorderTuple
+RepackCleanupContext
 RepackCommand
 RepackDecodingState
+RepackShmemStruct
 RepackStmt
+RepackWorkerInfo
 ReparameterizeForeignPathByChild_function
 ReplOriginId
 ReplOriginXactState
-- 
2.47.3


--kdrcpfmkbkc4lqhu--





^ permalink  raw  reply  [nested|flat] 223+ messages in thread

* [PATCH 2/2] Publish list of tables being repacked in shared memory
@ 2026-04-07 20:29 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 223+ messages in thread

From: Álvaro Herrera @ 2026-04-07 20:29 UTC (permalink / raw)

Use it in autovacuum to skip processing tables that are being repacked.
This is mostly to avoid repeated attempts to process such tables, which
would fail due to the special deadlock checker behavior for repack.

Author: Álvaro Herrera <[email protected]>
Discussion: https://postgr.es/m/[email protected]
---
 src/backend/commands/repack.c                 | 195 ++++++++++++++++--
 src/backend/postmaster/autovacuum.c           |  20 ++
 .../utils/activity/wait_event_names.txt       |   1 +
 src/include/commands/repack.h                 |   2 +
 src/include/storage/lwlocklist.h              |   2 +-
 src/include/storage/subsystemlist.h           |   1 +
 src/tools/pgindent/typedefs.list              |   3 +
 7 files changed, 210 insertions(+), 14 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index a5f5df77291..ee7072dce6a 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -63,9 +63,11 @@
 #include "optimizer/optimizer.h"
 #include "pgstat.h"
 #include "storage/bufmgr.h"
+#include "storage/ipc.h"
 #include "storage/lmgr.h"
 #include "storage/predicate.h"
 #include "storage/proc.h"
+#include "storage/subsystems.h"
 #include "utils/acl.h"
 #include "utils/fmgroids.h"
 #include "utils/guc.h"
@@ -79,6 +81,32 @@
 #include "utils/syscache.h"
 #include "utils/wait_event_types.h"
 
+
+/* Shared memory layout for REPACK */
+typedef struct RepackWorkerInfo
+{
+	bool		ri_in_use;
+	pid_t		ri_backendpid;
+	Oid			ri_dbid;
+	Oid			ri_relid;
+	Oid			ri_toastrelid;
+} RepackWorkerInfo;
+
+typedef struct
+{
+	bool		re_useless;
+	RepackWorkerInfo re_workerinfo[FLEXIBLE_ARRAY_MEMBER];
+} RepackShmemStruct;
+
+static RepackShmemStruct *RepackShmem;
+
+typedef struct RepackCleanupContext
+{
+	bool		concurrent;
+	int			workerindex;
+} RepackCleanupContext;
+
+
 /*
  * This struct is used to pass around the information on tables to be
  * clustered. We need this so we can make a list of them when invoked without
@@ -90,6 +118,7 @@ typedef struct
 	Oid			indexOid;
 } RelToCluster;
 
+
 /*
  * The first file exported by the decoding worker must contain a snapshot, the
  * following ones contain the data changes.
@@ -166,6 +195,10 @@ static List *get_tables_to_repack_partitioned(RepackCommand cmd,
 											  MemoryContext permcxt);
 static bool repack_is_permitted_for_relation(RepackCommand cmd,
 											 Oid relid, Oid userid);
+static void RepackCleanup(RepackCleanupContext *context);
+static void RepackCleanupCb(int code, Datum arg);
+static void RepackShmemRequest(void *arg);
+static void RepackShmemInit(void *arg);
 
 static void apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt);
 static void apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
@@ -210,6 +243,11 @@ static void ProcessRepackMessage(StringInfo msg);
 static const char *RepackCommandAsString(RepackCommand cmd);
 
 
+const ShmemCallbacks RepackShmemCallbacks = {
+	.request_fn = RepackShmemRequest,
+	.init_fn = RepackShmemInit,
+};
+
 /*
  * The repack code allows for processing multiple tables at once. Because
  * of this, we cannot just run everything on a single transaction, or we
@@ -514,6 +552,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 	Oid			tableOid = RelationGetRelid(OldHeap);
 	Relation	index;
 	LOCKMODE	lmode;
+	RepackCleanupContext context;
 	Oid			save_userid;
 	int			save_sec_context;
 	int			save_nestlevel;
@@ -660,24 +699,43 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 		TransferPredicateLocksToHeapRelation(OldHeap);
 
 	/* rebuild_relation does all the dirty work */
-	PG_TRY();
-	{
-		rebuild_relation(OldHeap, index, verbose, ident_idx);
-	}
-	PG_FINALLY();
+	context.concurrent = concurrent;
+
+	PG_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
 	{
 		if (concurrent)
 		{
-			/*
-			 * Since during normal operation the worker was already asked to
-			 * exit, stopping it explicitly is especially important on ERROR.
-			 * However it still seems a good practice to make sure that the
-			 * worker never survives the REPACK command.
-			 */
-			stop_repack_decoding_worker();
+			bool		freefound = false;
+
+			LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+			for (int i = 0; i < max_repack_replication_slots; i++)
+			{
+				RepackWorkerInfo *worker;
+
+				if (RepackShmem->re_workerinfo[i].ri_in_use)
+					continue;
+
+				freefound = true;
+				worker = &RepackShmem->re_workerinfo[i];
+				context.workerindex = i;
+
+				worker->ri_in_use = true;
+				worker->ri_backendpid = MyProcPid;
+				worker->ri_dbid = MyDatabaseId;
+				worker->ri_relid = RelationGetRelid(OldHeap);
+				worker->ri_toastrelid = OldHeap->rd_rel->reltoastrelid;
+				break;
+			}
+			if (!freefound)
+				elog(ERROR, "could not find free repack entry");
+			LWLockRelease(RepackLock);
 		}
+
+		rebuild_relation(OldHeap, index, verbose, ident_idx);
 	}
-	PG_END_TRY();
+	PG_END_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
+
+	RepackCleanup(&context);
 
 	/* rebuild_relation closes OldHeap, and index if valid */
 
@@ -691,6 +749,117 @@ out:
 	pgstat_progress_end_command();
 }
 
+/*
+ * Return whether any backend is running concurrent REPACK on the given table
+ * (which could be a toast table).
+ */
+bool
+is_table_under_repack(Oid databaseId, Oid relid)
+{
+	bool		retval = false;
+
+	LWLockAcquire(RepackLock, LW_SHARED);
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		RepackWorkerInfo *rworker;
+
+		if (!RepackShmem->re_workerinfo[i].ri_in_use)
+			continue;
+
+		rworker = &RepackShmem->re_workerinfo[i];
+		if (rworker->ri_dbid == MyDatabaseId &&
+			(rworker->ri_relid == relid ||
+			 rworker->ri_toastrelid == relid))
+			retval = true;
+	}
+	LWLockRelease(RepackLock);
+
+	return retval;
+}
+
+/*
+ * Remove ourselves from the workerinfo array.
+ */
+static void
+RepackCleanup(RepackCleanupContext *context)
+{
+	if (context->concurrent)
+	{
+		RepackWorkerInfo *worker;
+
+		/*
+		 * The worker would normally terminate on its own when the work is
+		 * done, but make sure we signal it just in case.
+		 */
+		stop_repack_decoding_worker();
+
+		/*
+		 * also, make sure we stop advertising the relation we were repacking,
+		 * so that autovacuum reverts to handling it normally.
+		 */
+		LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+
+		worker = &RepackShmem->re_workerinfo[context->workerindex];
+		Assert(worker->ri_backendpid == MyProcPid);
+		worker->ri_in_use = false;
+		worker->ri_backendpid = 0;
+		worker->ri_dbid = InvalidOid;
+		worker->ri_relid = InvalidOid;
+		worker->ri_toastrelid = InvalidOid;
+		LWLockRelease(RepackLock);
+	}
+}
+
+/*
+ * RepackCleanup wrapped as an on_shmem_exit callback function
+ */
+static void
+RepackCleanupCb(int code, Datum arg)
+{
+	RepackCleanup((RepackCleanupContext *) DatumGetPointer(arg));
+}
+
+/*
+ * RepackShmemRequest
+ *		Register shared memory space needed for repack
+ */
+static void
+RepackShmemRequest(void *arg)
+{
+	Size		size;
+
+	/*
+	 * Need the fixed struct and the array of RepackWorkerInfo.
+	 */
+	size = sizeof(RepackShmemStruct);
+	size = MAXALIGN(size);
+	size = add_size(size, mul_size(max_repack_replication_slots,
+								   sizeof(RepackWorkerInfo)));
+
+	ShmemRequestStruct(.name = "Repack Data",
+					   .size = size,
+					   .ptr = (void **) &RepackShmem,
+		);
+}
+
+static void
+RepackShmemInit(void *arg)
+{
+	RepackWorkerInfo *reinfo;
+
+	reinfo = (RepackWorkerInfo *) ((char *) RepackShmem +
+								   MAXALIGN(sizeof(RepackShmemStruct)));
+
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		reinfo[i].ri_in_use = false;
+		reinfo[i].ri_backendpid = 0;
+		reinfo[i].ri_dbid = InvalidOid;
+		reinfo[i].ri_relid = InvalidOid;
+		reinfo[i].ri_toastrelid = InvalidOid;
+	}
+}
+
 /*
  * Check if the table (and its index) still meets the requirements of
  * cluster_rel().
diff --git a/src/backend/postmaster/autovacuum.c b/src/backend/postmaster/autovacuum.c
index bd626a16363..080c64ea3c8 100644
--- a/src/backend/postmaster/autovacuum.c
+++ b/src/backend/postmaster/autovacuum.c
@@ -78,6 +78,7 @@
 #include "catalog/namespace.h"
 #include "catalog/pg_database.h"
 #include "catalog/pg_namespace.h"
+#include "commands/repack.h"
 #include "commands/vacuum.h"
 #include "common/int.h"
 #include "funcapi.h"
@@ -2422,6 +2423,25 @@ do_autovacuum(void)
 			}
 		}
 		LWLockRelease(AutovacuumLock);
+
+		/*
+		 * Similarly, if the table is being processed by concurrent repack,
+		 * skip it (but make a note of that).  We wouldn't be able to acquire
+		 * its lock anyway.
+		 */
+		if (!skipit)
+		{
+			MemoryContextSwitchTo(PortalContext);
+
+			skipit = is_table_under_repack(MyDatabaseId, relid);
+			if (skipit)
+				ereport(LOG,
+						errmsg("skipping table \"%s.%s.%s\" because it's being repacked in concurrent mode",
+							   get_database_name(MyDatabaseId),
+							   get_namespace_name(get_rel_namespace(relid)),
+							   get_rel_name(relid)));
+		}
+
 		if (skipit)
 		{
 			LWLockRelease(AutovacuumScheduleLock);
diff --git a/src/backend/utils/activity/wait_event_names.txt b/src/backend/utils/activity/wait_event_names.txt
index 7bda5298558..e206304f204 100644
--- a/src/backend/utils/activity/wait_event_names.txt
+++ b/src/backend/utils/activity/wait_event_names.txt
@@ -332,6 +332,7 @@ SInvalWrite	"Waiting to add a message to the shared catalog invalidation queue."
 WALBufMapping	"Waiting to replace a page in WAL buffers."
 WALWrite	"Waiting for WAL buffers to be written to disk."
 ControlFile	"Waiting to read or update the <filename>pg_control</filename> file or create a new WAL file."
+Repack	"Waiting to read or update tables in process by concurrent repack."
 MultiXactGen	"Waiting to read or update shared multixact state."
 RelCacheInit	"Waiting to read or update a <filename>pg_internal.init</filename> relation cache initialization file."
 CheckpointerComm	"Waiting to manage fsync requests."
diff --git a/src/include/commands/repack.h b/src/include/commands/repack.h
index fd16e74b179..be7d38b5fae 100644
--- a/src/include/commands/repack.h
+++ b/src/include/commands/repack.h
@@ -42,6 +42,8 @@ extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel);
 
 extern void cluster_rel(RepackCommand command, Relation OldHeap, Oid indexOid,
 						ClusterParams *params, bool isTopLevel);
+extern bool is_table_under_repack(Oid databaseId, Oid relid);
+
 extern void check_index_is_clusterable(Relation OldHeap, Oid indexOid,
 									   LOCKMODE lockmode);
 extern void mark_index_clustered(Relation rel, Oid indexOid, bool is_internal);
diff --git a/src/include/storage/lwlocklist.h b/src/include/storage/lwlocklist.h
index af8553bcb6c..3f08f4a15d4 100644
--- a/src/include/storage/lwlocklist.h
+++ b/src/include/storage/lwlocklist.h
@@ -41,7 +41,7 @@ PG_LWLOCK(6, SInvalWrite)
 PG_LWLOCK(7, WALBufMapping)
 PG_LWLOCK(8, WALWrite)
 PG_LWLOCK(9, ControlFile)
-/* 10 was CheckpointLock */
+PG_LWLOCK(10, Repack)
 /* 11 was XactSLRULock */
 /* 12 was SubtransSLRULock */
 PG_LWLOCK(13, MultiXactGen)
diff --git a/src/include/storage/subsystemlist.h b/src/include/storage/subsystemlist.h
index 9ad619080be..4e683b8b0a8 100644
--- a/src/include/storage/subsystemlist.h
+++ b/src/include/storage/subsystemlist.h
@@ -72,6 +72,7 @@ PG_SHMEM_SUBSYSTEM(WalSummarizerShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(PgArchShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(ApplyLauncherShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(SlotSyncShmemCallbacks)
+PG_SHMEM_SUBSYSTEM(RepackShmemCallbacks)
 
 /* other modules that need some shared memory space */
 PG_SHMEM_SUBSYSTEM(BTreeShmemCallbacks)
diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list
index 637c669a146..d019e03aaf1 100644
--- a/src/tools/pgindent/typedefs.list
+++ b/src/tools/pgindent/typedefs.list
@@ -2639,9 +2639,12 @@ ReorderBufferTupleCidEnt
 ReorderBufferTupleCidKey
 ReorderBufferUpdateProgressTxnCB
 ReorderTuple
+RepackCleanupContext
 RepackCommand
 RepackDecodingState
+RepackShmemStruct
 RepackStmt
+RepackWorkerInfo
 ReparameterizeForeignPathByChild_function
 ReplOriginId
 ReplOriginXactState
-- 
2.47.3


--kdrcpfmkbkc4lqhu--





^ permalink  raw  reply  [nested|flat] 223+ messages in thread

* [PATCH 2/2] Publish list of tables being repacked in shared memory
@ 2026-04-07 20:29 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 223+ messages in thread

From: Álvaro Herrera @ 2026-04-07 20:29 UTC (permalink / raw)

Use it in autovacuum to skip processing tables that are being repacked.
This is mostly to avoid repeated attempts to process such tables, which
would fail due to the special deadlock checker behavior for repack.

Author: Álvaro Herrera <[email protected]>
Discussion: https://postgr.es/m/[email protected]
---
 src/backend/commands/repack.c                 | 195 ++++++++++++++++--
 src/backend/postmaster/autovacuum.c           |  20 ++
 .../utils/activity/wait_event_names.txt       |   1 +
 src/include/commands/repack.h                 |   2 +
 src/include/storage/lwlocklist.h              |   2 +-
 src/include/storage/subsystemlist.h           |   1 +
 src/tools/pgindent/typedefs.list              |   3 +
 7 files changed, 210 insertions(+), 14 deletions(-)

diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index a5f5df77291..ee7072dce6a 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -63,9 +63,11 @@
 #include "optimizer/optimizer.h"
 #include "pgstat.h"
 #include "storage/bufmgr.h"
+#include "storage/ipc.h"
 #include "storage/lmgr.h"
 #include "storage/predicate.h"
 #include "storage/proc.h"
+#include "storage/subsystems.h"
 #include "utils/acl.h"
 #include "utils/fmgroids.h"
 #include "utils/guc.h"
@@ -79,6 +81,32 @@
 #include "utils/syscache.h"
 #include "utils/wait_event_types.h"
 
+
+/* Shared memory layout for REPACK */
+typedef struct RepackWorkerInfo
+{
+	bool		ri_in_use;
+	pid_t		ri_backendpid;
+	Oid			ri_dbid;
+	Oid			ri_relid;
+	Oid			ri_toastrelid;
+} RepackWorkerInfo;
+
+typedef struct
+{
+	bool		re_useless;
+	RepackWorkerInfo re_workerinfo[FLEXIBLE_ARRAY_MEMBER];
+} RepackShmemStruct;
+
+static RepackShmemStruct *RepackShmem;
+
+typedef struct RepackCleanupContext
+{
+	bool		concurrent;
+	int			workerindex;
+} RepackCleanupContext;
+
+
 /*
  * This struct is used to pass around the information on tables to be
  * clustered. We need this so we can make a list of them when invoked without
@@ -90,6 +118,7 @@ typedef struct
 	Oid			indexOid;
 } RelToCluster;
 
+
 /*
  * The first file exported by the decoding worker must contain a snapshot, the
  * following ones contain the data changes.
@@ -166,6 +195,10 @@ static List *get_tables_to_repack_partitioned(RepackCommand cmd,
 											  MemoryContext permcxt);
 static bool repack_is_permitted_for_relation(RepackCommand cmd,
 											 Oid relid, Oid userid);
+static void RepackCleanup(RepackCleanupContext *context);
+static void RepackCleanupCb(int code, Datum arg);
+static void RepackShmemRequest(void *arg);
+static void RepackShmemInit(void *arg);
 
 static void apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt);
 static void apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
@@ -210,6 +243,11 @@ static void ProcessRepackMessage(StringInfo msg);
 static const char *RepackCommandAsString(RepackCommand cmd);
 
 
+const ShmemCallbacks RepackShmemCallbacks = {
+	.request_fn = RepackShmemRequest,
+	.init_fn = RepackShmemInit,
+};
+
 /*
  * The repack code allows for processing multiple tables at once. Because
  * of this, we cannot just run everything on a single transaction, or we
@@ -514,6 +552,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 	Oid			tableOid = RelationGetRelid(OldHeap);
 	Relation	index;
 	LOCKMODE	lmode;
+	RepackCleanupContext context;
 	Oid			save_userid;
 	int			save_sec_context;
 	int			save_nestlevel;
@@ -660,24 +699,43 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid,
 		TransferPredicateLocksToHeapRelation(OldHeap);
 
 	/* rebuild_relation does all the dirty work */
-	PG_TRY();
-	{
-		rebuild_relation(OldHeap, index, verbose, ident_idx);
-	}
-	PG_FINALLY();
+	context.concurrent = concurrent;
+
+	PG_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
 	{
 		if (concurrent)
 		{
-			/*
-			 * Since during normal operation the worker was already asked to
-			 * exit, stopping it explicitly is especially important on ERROR.
-			 * However it still seems a good practice to make sure that the
-			 * worker never survives the REPACK command.
-			 */
-			stop_repack_decoding_worker();
+			bool		freefound = false;
+
+			LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+			for (int i = 0; i < max_repack_replication_slots; i++)
+			{
+				RepackWorkerInfo *worker;
+
+				if (RepackShmem->re_workerinfo[i].ri_in_use)
+					continue;
+
+				freefound = true;
+				worker = &RepackShmem->re_workerinfo[i];
+				context.workerindex = i;
+
+				worker->ri_in_use = true;
+				worker->ri_backendpid = MyProcPid;
+				worker->ri_dbid = MyDatabaseId;
+				worker->ri_relid = RelationGetRelid(OldHeap);
+				worker->ri_toastrelid = OldHeap->rd_rel->reltoastrelid;
+				break;
+			}
+			if (!freefound)
+				elog(ERROR, "could not find free repack entry");
+			LWLockRelease(RepackLock);
 		}
+
+		rebuild_relation(OldHeap, index, verbose, ident_idx);
 	}
-	PG_END_TRY();
+	PG_END_ENSURE_ERROR_CLEANUP(RepackCleanupCb, PointerGetDatum(&context));
+
+	RepackCleanup(&context);
 
 	/* rebuild_relation closes OldHeap, and index if valid */
 
@@ -691,6 +749,117 @@ out:
 	pgstat_progress_end_command();
 }
 
+/*
+ * Return whether any backend is running concurrent REPACK on the given table
+ * (which could be a toast table).
+ */
+bool
+is_table_under_repack(Oid databaseId, Oid relid)
+{
+	bool		retval = false;
+
+	LWLockAcquire(RepackLock, LW_SHARED);
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		RepackWorkerInfo *rworker;
+
+		if (!RepackShmem->re_workerinfo[i].ri_in_use)
+			continue;
+
+		rworker = &RepackShmem->re_workerinfo[i];
+		if (rworker->ri_dbid == MyDatabaseId &&
+			(rworker->ri_relid == relid ||
+			 rworker->ri_toastrelid == relid))
+			retval = true;
+	}
+	LWLockRelease(RepackLock);
+
+	return retval;
+}
+
+/*
+ * Remove ourselves from the workerinfo array.
+ */
+static void
+RepackCleanup(RepackCleanupContext *context)
+{
+	if (context->concurrent)
+	{
+		RepackWorkerInfo *worker;
+
+		/*
+		 * The worker would normally terminate on its own when the work is
+		 * done, but make sure we signal it just in case.
+		 */
+		stop_repack_decoding_worker();
+
+		/*
+		 * also, make sure we stop advertising the relation we were repacking,
+		 * so that autovacuum reverts to handling it normally.
+		 */
+		LWLockAcquire(RepackLock, LW_EXCLUSIVE);
+
+		worker = &RepackShmem->re_workerinfo[context->workerindex];
+		Assert(worker->ri_backendpid == MyProcPid);
+		worker->ri_in_use = false;
+		worker->ri_backendpid = 0;
+		worker->ri_dbid = InvalidOid;
+		worker->ri_relid = InvalidOid;
+		worker->ri_toastrelid = InvalidOid;
+		LWLockRelease(RepackLock);
+	}
+}
+
+/*
+ * RepackCleanup wrapped as an on_shmem_exit callback function
+ */
+static void
+RepackCleanupCb(int code, Datum arg)
+{
+	RepackCleanup((RepackCleanupContext *) DatumGetPointer(arg));
+}
+
+/*
+ * RepackShmemRequest
+ *		Register shared memory space needed for repack
+ */
+static void
+RepackShmemRequest(void *arg)
+{
+	Size		size;
+
+	/*
+	 * Need the fixed struct and the array of RepackWorkerInfo.
+	 */
+	size = sizeof(RepackShmemStruct);
+	size = MAXALIGN(size);
+	size = add_size(size, mul_size(max_repack_replication_slots,
+								   sizeof(RepackWorkerInfo)));
+
+	ShmemRequestStruct(.name = "Repack Data",
+					   .size = size,
+					   .ptr = (void **) &RepackShmem,
+		);
+}
+
+static void
+RepackShmemInit(void *arg)
+{
+	RepackWorkerInfo *reinfo;
+
+	reinfo = (RepackWorkerInfo *) ((char *) RepackShmem +
+								   MAXALIGN(sizeof(RepackShmemStruct)));
+
+	for (int i = 0; i < max_repack_replication_slots; i++)
+	{
+		reinfo[i].ri_in_use = false;
+		reinfo[i].ri_backendpid = 0;
+		reinfo[i].ri_dbid = InvalidOid;
+		reinfo[i].ri_relid = InvalidOid;
+		reinfo[i].ri_toastrelid = InvalidOid;
+	}
+}
+
 /*
  * Check if the table (and its index) still meets the requirements of
  * cluster_rel().
diff --git a/src/backend/postmaster/autovacuum.c b/src/backend/postmaster/autovacuum.c
index bd626a16363..080c64ea3c8 100644
--- a/src/backend/postmaster/autovacuum.c
+++ b/src/backend/postmaster/autovacuum.c
@@ -78,6 +78,7 @@
 #include "catalog/namespace.h"
 #include "catalog/pg_database.h"
 #include "catalog/pg_namespace.h"
+#include "commands/repack.h"
 #include "commands/vacuum.h"
 #include "common/int.h"
 #include "funcapi.h"
@@ -2422,6 +2423,25 @@ do_autovacuum(void)
 			}
 		}
 		LWLockRelease(AutovacuumLock);
+
+		/*
+		 * Similarly, if the table is being processed by concurrent repack,
+		 * skip it (but make a note of that).  We wouldn't be able to acquire
+		 * its lock anyway.
+		 */
+		if (!skipit)
+		{
+			MemoryContextSwitchTo(PortalContext);
+
+			skipit = is_table_under_repack(MyDatabaseId, relid);
+			if (skipit)
+				ereport(LOG,
+						errmsg("skipping table \"%s.%s.%s\" because it's being repacked in concurrent mode",
+							   get_database_name(MyDatabaseId),
+							   get_namespace_name(get_rel_namespace(relid)),
+							   get_rel_name(relid)));
+		}
+
 		if (skipit)
 		{
 			LWLockRelease(AutovacuumScheduleLock);
diff --git a/src/backend/utils/activity/wait_event_names.txt b/src/backend/utils/activity/wait_event_names.txt
index 7bda5298558..e206304f204 100644
--- a/src/backend/utils/activity/wait_event_names.txt
+++ b/src/backend/utils/activity/wait_event_names.txt
@@ -332,6 +332,7 @@ SInvalWrite	"Waiting to add a message to the shared catalog invalidation queue."
 WALBufMapping	"Waiting to replace a page in WAL buffers."
 WALWrite	"Waiting for WAL buffers to be written to disk."
 ControlFile	"Waiting to read or update the <filename>pg_control</filename> file or create a new WAL file."
+Repack	"Waiting to read or update tables in process by concurrent repack."
 MultiXactGen	"Waiting to read or update shared multixact state."
 RelCacheInit	"Waiting to read or update a <filename>pg_internal.init</filename> relation cache initialization file."
 CheckpointerComm	"Waiting to manage fsync requests."
diff --git a/src/include/commands/repack.h b/src/include/commands/repack.h
index fd16e74b179..be7d38b5fae 100644
--- a/src/include/commands/repack.h
+++ b/src/include/commands/repack.h
@@ -42,6 +42,8 @@ extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel);
 
 extern void cluster_rel(RepackCommand command, Relation OldHeap, Oid indexOid,
 						ClusterParams *params, bool isTopLevel);
+extern bool is_table_under_repack(Oid databaseId, Oid relid);
+
 extern void check_index_is_clusterable(Relation OldHeap, Oid indexOid,
 									   LOCKMODE lockmode);
 extern void mark_index_clustered(Relation rel, Oid indexOid, bool is_internal);
diff --git a/src/include/storage/lwlocklist.h b/src/include/storage/lwlocklist.h
index af8553bcb6c..3f08f4a15d4 100644
--- a/src/include/storage/lwlocklist.h
+++ b/src/include/storage/lwlocklist.h
@@ -41,7 +41,7 @@ PG_LWLOCK(6, SInvalWrite)
 PG_LWLOCK(7, WALBufMapping)
 PG_LWLOCK(8, WALWrite)
 PG_LWLOCK(9, ControlFile)
-/* 10 was CheckpointLock */
+PG_LWLOCK(10, Repack)
 /* 11 was XactSLRULock */
 /* 12 was SubtransSLRULock */
 PG_LWLOCK(13, MultiXactGen)
diff --git a/src/include/storage/subsystemlist.h b/src/include/storage/subsystemlist.h
index 9ad619080be..4e683b8b0a8 100644
--- a/src/include/storage/subsystemlist.h
+++ b/src/include/storage/subsystemlist.h
@@ -72,6 +72,7 @@ PG_SHMEM_SUBSYSTEM(WalSummarizerShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(PgArchShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(ApplyLauncherShmemCallbacks)
 PG_SHMEM_SUBSYSTEM(SlotSyncShmemCallbacks)
+PG_SHMEM_SUBSYSTEM(RepackShmemCallbacks)
 
 /* other modules that need some shared memory space */
 PG_SHMEM_SUBSYSTEM(BTreeShmemCallbacks)
diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list
index 637c669a146..d019e03aaf1 100644
--- a/src/tools/pgindent/typedefs.list
+++ b/src/tools/pgindent/typedefs.list
@@ -2639,9 +2639,12 @@ ReorderBufferTupleCidEnt
 ReorderBufferTupleCidKey
 ReorderBufferUpdateProgressTxnCB
 ReorderTuple
+RepackCleanupContext
 RepackCommand
 RepackDecodingState
+RepackShmemStruct
 RepackStmt
+RepackWorkerInfo
 ReparameterizeForeignPathByChild_function
 ReplOriginId
 ReplOriginXactState
-- 
2.47.3


--kdrcpfmkbkc4lqhu--





^ permalink  raw  reply  [nested|flat] 223+ messages in thread