public inbox for [email protected]
help / color / mirror / Atom feedFrom: Antonin Houska <[email protected]>
To: Alvaro Herrera <[email protected]>
Cc: Junwang Zhao <[email protected]>
Cc: Kirill Reshke <[email protected]>
Cc: Pavel Stehule <[email protected]>
Cc: Michael Paquier <[email protected]>
Cc: PostgreSQL Hackers <[email protected]>
Subject: Re: why there is not VACUUM FULL CONCURRENTLY?
Date: Mon, 13 Jan 2025 14:48:31 +0100
Message-ID: <6250.1736776111@antos> (raw)
In-Reply-To: <[email protected]>
References: <[email protected]>
Alvaro Herrera <[email protected]> wrote:
> On 2025-Jan-09, Antonin Houska wrote:
>
> > It seems you accidentally fixed another problem :-) I was referring to the
> > 'lockmode' argument of make_new_heap(). I can try to write a patch for that
> > but ...
> >
> > > Meanwhile the patch 0004 has some seemingly trivial conflicts. If you
> > > want to rebase, I'd appreciate that. In the meantime I'll give a look
> > > at the next two other API changes.
This is the patch series rebased on top of the commit cc811f92ba.
I haven't addressed the problem of a new command yet - for that I'd like to
see some sort of consensus, so that I do not have to do all the related
changes many times.
--
Antonin Houska
Web: https://www.cybertec-postgresql.com
From bf2ec8c5d753de340140839f1b061044ec4c1149 Mon Sep 17 00:00:00 2001
From: Antonin Houska <[email protected]>
Date: Mon, 13 Jan 2025 14:29:54 +0100
Subject: [PATCH 4/8] Add CONCURRENTLY option to both VACUUM FULL and CLUSTER
commands.
Both VACUUM FULL and CLUSTER commands copy the relation data into a new file,
create new indexes and eventually swap the files. To make sure that the old
file does not change during the copying, the relation is locked in an
exclusive mode, which prevents applications from both reading and writing. (To
keep the data consistent, we'd only need to prevent the applications from
writing, but even reading needs to be blocked before we can swap the files -
otherwise some applications could continue using the old file. Since we cannot
get stronger lock without releasing the weaker one first, we acquire the
exclusive lock in the beginning and keep it till the end of the processing.)
This patch introduces an alternative workflow, which only requires the
exclusive lock when the relation (and index) files are being swapped.
(Supposedly, the swapping should be pretty fast.) On the other hand, when we
copy the data to the new file, we allow applications to read from the relation
and even write into it.
First, we scan the relation using a "historic snapshot", and insert all the
tuples satisfying this snapshot into the new file. Note that, before creating
that snapshot, we need to make sure that all the other backends treat the
relation as a system catalog: in particular, they must log information on new
command IDs (CIDs). We achieve that by adding the relation ID into a shared
hash table and waiting until all the transactions currently writing into the
table (i.e. transactions possibly not aware of the new entry) have finished.
Second, logical decoding is used to capture the data changes done by
applications during the copying (i.e. changes that do not satisfy the historic
snapshot mentioned above), and those are applied to the new file before we
acquire the exclusive lock we need to swap the files. (Of course, more data
changes can take place while we are waiting for the lock - these will be
applied to the new file after we have acquired the lock, before we swap the
files.)
While copying the data into the new file, we hold a lock that prevents
applications from changing the relation tuple descriptor (tuples inserted into
the old file must fit into the new file). However, as we have to release that
lock before getting the exclusive one, it's possible that someone adds or
drops a column, or changes the data type of an existing one. Therefore we have
to check the tuple descriptor before we swap the files. If we find out that
the tuple descriptor changed, ERROR is raised and all the changes are rolled
back. Since a lot of effort can be wasted in such a case, the ALTER TABLE
command also tries to check if VACUUM FULL / CLUSTER with the CONCURRENTLY
option is running on the same relation, and raises an ERROR if it is.
Like the existing implementation of both VACUUM FULL and CLUSTER commands, the
variant with the CONCURRENTLY option also requires an extra space for the new
relation and index files (which coexist with the old files for some time). In
addition, the CONCURRENTLY option might introduce a lag in releasing WAL
segments for archiving / recycling. This is due to the decoding of the data
changes done by application concurrently. However, this lag should not be more
than a single WAL segment.
---
doc/src/sgml/monitoring.sgml | 36 +-
doc/src/sgml/ref/cluster.sgml | 116 +-
doc/src/sgml/ref/vacuum.sgml | 22 +-
src/Makefile | 1 +
src/backend/access/heap/heapam.c | 8 +-
src/backend/access/heap/heapam_handler.c | 145 +-
src/backend/access/heap/heapam_visibility.c | 30 +-
src/backend/catalog/index.c | 43 +-
src/backend/catalog/system_views.sql | 17 +-
src/backend/commands/cluster.c | 2585 ++++++++++++++++-
src/backend/commands/matview.c | 2 +-
src/backend/commands/tablecmds.c | 11 +
src/backend/commands/vacuum.c | 126 +-
src/backend/meson.build | 1 +
src/backend/replication/logical/decode.c | 24 +
src/backend/replication/logical/snapbuild.c | 20 +
.../replication/pgoutput_cluster/Makefile | 32 +
.../replication/pgoutput_cluster/meson.build | 18 +
.../pgoutput_cluster/pgoutput_cluster.c | 288 ++
src/backend/storage/ipc/ipci.c | 3 +
src/backend/tcop/utility.c | 11 +
src/backend/utils/activity/backend_progress.c | 16 +
.../utils/activity/wait_event_names.txt | 1 +
src/backend/utils/cache/inval.c | 22 +
src/backend/utils/cache/relcache.c | 5 +
src/backend/utils/time/snapmgr.c | 3 +-
src/bin/psql/tab-complete.in.c | 5 +-
src/include/access/heapam.h | 4 +
src/include/access/tableam.h | 10 +
src/include/catalog/index.h | 3 +
src/include/commands/cluster.h | 94 +-
src/include/commands/progress.h | 17 +-
src/include/commands/vacuum.h | 17 +-
src/include/replication/snapbuild.h | 1 +
src/include/storage/lockdefs.h | 5 +-
src/include/storage/lwlocklist.h | 1 +
src/include/utils/backend_progress.h | 3 +-
src/include/utils/inval.h | 2 +
src/include/utils/rel.h | 7 +-
src/include/utils/snapmgr.h | 2 +
src/test/regress/expected/rules.out | 17 +-
41 files changed, 3572 insertions(+), 202 deletions(-)
create mode 100644 src/backend/replication/pgoutput_cluster/Makefile
create mode 100644 src/backend/replication/pgoutput_cluster/meson.build
create mode 100644 src/backend/replication/pgoutput_cluster/pgoutput_cluster.c
diff --git a/doc/src/sgml/monitoring.sgml b/doc/src/sgml/monitoring.sgml
index d0d176cc54..985b20a81e 100644
--- a/doc/src/sgml/monitoring.sgml
+++ b/doc/src/sgml/monitoring.sgml
@@ -5727,14 +5727,35 @@ FROM pg_stat_get_backend_idset() AS backendid;
<row>
<entry role="catalog_table_entry"><para role="column_definition">
- <structfield>heap_tuples_written</structfield> <type>bigint</type>
+ <structfield>heap_tuples_inserted</structfield> <type>bigint</type>
</para>
<para>
- Number of heap tuples written.
+ Number of heap tuples inserted.
This counter only advances when the phase is
<literal>seq scanning heap</literal>,
- <literal>index scanning heap</literal>
- or <literal>writing new heap</literal>.
+ <literal>index scanning heap</literal>,
+ <literal>writing new heap</literal>
+ or <literal>catch-up</literal>.
+ </para></entry>
+ </row>
+
+ <row>
+ <entry role="catalog_table_entry"><para role="column_definition">
+ <structfield>heap_tuples_updated</structfield> <type>bigint</type>
+ </para>
+ <para>
+ Number of heap tuples updated.
+ This counter only advances when the phase is <literal>catch-up</literal>.
+ </para></entry>
+ </row>
+
+ <row>
+ <entry role="catalog_table_entry"><para role="column_definition">
+ <structfield>heap_tuples_deleted</structfield> <type>bigint</type>
+ </para>
+ <para>
+ Number of heap tuples deleted.
+ This counter only advances when the phase is <literal>catch-up</literal>.
</para></entry>
</row>
@@ -5815,6 +5836,13 @@ FROM pg_stat_get_backend_idset() AS backendid;
<command>CLUSTER</command> is currently writing the new heap.
</entry>
</row>
+ <row>
+ <entry><literal>catch-up</literal></entry>
+ <entry>
+ <command>CLUSTER</command> is currently processing the DML commands
+ that other transactions executed during any of the preceding phase.
+ </entry>
+ </row>
<row>
<entry><literal>swapping relation files</literal></entry>
<entry>
diff --git a/doc/src/sgml/ref/cluster.sgml b/doc/src/sgml/ref/cluster.sgml
index 8811f169ea..356b40e3fe 100644
--- a/doc/src/sgml/ref/cluster.sgml
+++ b/doc/src/sgml/ref/cluster.sgml
@@ -26,6 +26,7 @@ CLUSTER [ ( <replaceable class="parameter">option</replaceable> [, ...] ) ] [ <r
<phrase>where <replaceable class="parameter">option</replaceable> can be one of:</phrase>
VERBOSE [ <replaceable class="parameter">boolean</replaceable> ]
+ CONCURRENTLY [ <replaceable class="parameter">boolean</replaceable> ]
</synopsis>
</refsynopsisdiv>
@@ -69,14 +70,17 @@ CLUSTER [ ( <replaceable class="parameter">option</replaceable> [, ...] ) ] [ <r
<replaceable class="parameter">table_name</replaceable> reclusters all the
previously-clustered tables in the current database that the calling user
has privileges for. This form of <command>CLUSTER</command> cannot be
- executed inside a transaction block.
+ executed inside a transaction block. Also, this form is not allowed if
+ the <literal>CONCURRENTLY</literal> option is used.
</para>
<para>
- When a table is being clustered, an <literal>ACCESS
- EXCLUSIVE</literal> lock is acquired on it. This prevents any other
- database operations (both reads and writes) from operating on the
- table until the <command>CLUSTER</command> is finished.
+ When a table is being clustered, an <literal>ACCESS EXCLUSIVE</literal>
+ lock is acquired on it. This prevents any other database operations (both
+ reads and writes) from operating on the table until
+ the <command>CLUSTER</command> is finished. If you want to keep the table
+ accessible during the clustering, consider using
+ the <literal>CONCURRENTLY</literal> option.
</para>
</refsect1>
@@ -112,6 +116,108 @@ CLUSTER [ ( <replaceable class="parameter">option</replaceable> [, ...] ) ] [ <r
</listitem>
</varlistentry>
+ <varlistentry>
+ <term><literal>CONCURRENTLY</literal></term>
+ <listitem>
+ <para>
+ Allow other transactions to use the table while it is being clustered.
+ </para>
+
+ <para>
+ Internally, <command>CLUSTER</command> copies the contents of the table
+ (ignoring dead tuples) into a new file, sorted by the specified index,
+ and also creates a new file for each index. Then it swaps the old and
+ new files for the table and all the indexes, and deletes the old
+ files. The <literal>ACCESS EXCLUSIVE</literal> lock is needed to make
+ sure that the old files do not change during the processing because the
+ changes would get lost due to the swap.
+ </para>
+
+ <para>
+ With the <literal>CONCURRENTLY</literal> option, the <literal>ACCESS
+ EXCLUSIVE</literal> lock is only acquired to swap the table and index
+ files. The data changes that took place during the creation of the new
+ table and index files are captured using logical decoding
+ (<xref linkend="logicaldecoding"/>) and applied before
+ the <literal>ACCESS EXCLUSIVE</literal> lock is requested. Thus the lock
+ is typically held only for the time needed to swap the files, which
+ should be pretty short.
+ </para>
+
+ <para>
+ Note that <command>CLUSTER</command> with the
+ the <literal>CONCURRENTLY</literal> option does not try to order the
+ rows inserted into the table after the clustering started. Also
+ note <command>CLUSTER</command> might fail to complete due to DDL
+ commands executed on the table by other transactions during the
+ clustering.
+ </para>
+
+ <note>
+ <para>
+ In addition to the temporary space requirements explained below,
+ the <literal>CONCURRENTLY</literal> option can add to the usage of
+ temporary space a bit more. The reason is that other transactions can
+ perform DML operations which cannot be applied to the new file until
+ <command>CLUSTER</command> has copied all the tuples from the old
+ file. Thus the tuples inserted into the old file during the copying are
+ also stored in separately in a temporary file, so they can eventually
+ be applied to the new file.
+ </para>
+
+ <para>
+ Furthermore, the data changes performed during the copying are
+ extracted from <link linkend="wal">write-ahead log</link> (WAL), and
+ this extraction (decoding) only takes place when certain amount of WAL
+ has been written. Therefore, WAL removal can be delayed by this
+ threshold. Currently the threshold is equal to the value of
+ the <link linkend="guc-wal-segment-size"><varname>wal_segment_size</varname></link>
+ configuration parameter.
+ </para>
+ </note>
+
+ <para>
+ The <literal>CONCURRENTLY</literal> option cannot be used in the
+ following cases:
+
+ <itemizedlist>
+ <listitem>
+ <para>
+ The table is partitioned.
+ </para>
+ </listitem>
+
+ <listitem>
+ <para>
+ The table is a system catalog or a <acronym>TOAST</acronym> table.
+ </para>
+ </listitem>
+
+ <listitem>
+ <para>
+ <command>CLUSTER</command> is executed inside a transaction block.
+ </para>
+ </listitem>
+
+ <listitem>
+ <para>
+ The <link linkend="guc-wal-level"><varname>wal_level</varname></link>
+ configuration parameter is less than <literal>logical</literal>.
+ </para>
+ </listitem>
+
+ <listitem>
+ <para>
+ The <link linkend="guc-max-replication-slots"><varname>max_replication_slots</varname></link>
+ configuration parameter does not allow for creation of an additional
+ replication slot.
+ </para>
+ </listitem>
+ </itemizedlist>
+ </para>
+ </listitem>
+ </varlistentry>
+
<varlistentry>
<term><replaceable class="parameter">boolean</replaceable></term>
<listitem>
diff --git a/doc/src/sgml/ref/vacuum.sgml b/doc/src/sgml/ref/vacuum.sgml
index 971b1237d4..ba8669026a 100644
--- a/doc/src/sgml/ref/vacuum.sgml
+++ b/doc/src/sgml/ref/vacuum.sgml
@@ -39,6 +39,7 @@ VACUUM [ ( <replaceable class="parameter">option</replaceable> [, ...] ) ] [ <re
SKIP_DATABASE_STATS [ <replaceable class="parameter">boolean</replaceable> ]
ONLY_DATABASE_STATS [ <replaceable class="parameter">boolean</replaceable> ]
BUFFER_USAGE_LIMIT <replaceable class="parameter">size</replaceable>
+ CONCURRENTLY [ <replaceable class="parameter">boolean</replaceable> ]
<phrase>and <replaceable class="parameter">table_and_columns</replaceable> is:</phrase>
@@ -62,7 +63,8 @@ VACUUM [ ( <replaceable class="parameter">option</replaceable> [, ...] ) ] [ <re
Without a <replaceable class="parameter">table_and_columns</replaceable>
list, <command>VACUUM</command> processes every table and materialized view
in the current database that the current user has permission to vacuum.
- With a list, <command>VACUUM</command> processes only those table(s).
+ With a list, <command>VACUUM</command> processes only those table(s). The
+ list is required if the <literal>CONCURRENTLY</literal> option is used.
</para>
<para>
@@ -361,6 +363,24 @@ VACUUM [ ( <replaceable class="parameter">option</replaceable> [, ...] ) ] [ <re
</listitem>
</varlistentry>
+ <varlistentry>
+ <term><literal>CONCURRENTLY</literal></term>
+ <listitem>
+ <para>
+ Allow other transactions to use the table while it is being vacuumed. If
+ this option is specified, <command>VACUUM</command> can only process
+ tables which have already been clustered. For more information, see the
+ description of the <literal>CONCURRENTLY</literal> option of the
+ <xref linkend="sql-cluster"/> command.
+ </para>
+
+ <para>
+ The <literal>CONCURRENTLY</literal> option can only be used
+ if <literal>FULL</literal> is used at the same time.
+ </para>
+ </listitem>
+ </varlistentry>
+
<varlistentry>
<term><replaceable class="parameter">boolean</replaceable></term>
<listitem>
diff --git a/src/Makefile b/src/Makefile
index 2f31a2f20a..8b9d30ff72 100644
--- a/src/Makefile
+++ b/src/Makefile
@@ -23,6 +23,7 @@ SUBDIRS = \
interfaces \
backend/replication/libpqwalreceiver \
backend/replication/pgoutput \
+ backend/replication/pgoutput_cluster \
fe_utils \
bin \
pl \
diff --git a/src/backend/access/heap/heapam.c b/src/backend/access/heap/heapam.c
index 485525f4d6..552993d4ef 100644
--- a/src/backend/access/heap/heapam.c
+++ b/src/backend/access/heap/heapam.c
@@ -2073,8 +2073,14 @@ heap_insert(Relation relation, HeapTuple tup, CommandId cid,
/*
* If this is a catalog, we need to transmit combo CIDs to properly
* decode, so log that as well.
+ *
+ * For the main heap (as opposed to TOAST), we only receive
+ * HEAP_INSERT_NO_LOGICAL when doing VACUUM FULL / CLUSTER, in which
+ * case the visibility information does not change. Therefore, there's
+ * no need to update the decoding snapshot.
*/
- if (RelationIsAccessibleInLogicalDecoding(relation))
+ if ((options & HEAP_INSERT_NO_LOGICAL) == 0 &&
+ RelationIsAccessibleInLogicalDecoding(relation))
log_heap_new_cid(relation, heaptup);
/*
diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c
index e817f8f8f8..c5ec21ca2f 100644
--- a/src/backend/access/heap/heapam_handler.c
+++ b/src/backend/access/heap/heapam_handler.c
@@ -33,6 +33,7 @@
#include "catalog/index.h"
#include "catalog/storage.h"
#include "catalog/storage_xlog.h"
+#include "commands/cluster.h"
#include "commands/progress.h"
#include "executor/executor.h"
#include "miscadmin.h"
@@ -53,6 +54,9 @@ static void reform_and_rewrite_tuple(HeapTuple tuple,
static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer,
HeapTuple tuple,
OffsetNumber tupoffset);
+static HeapTuple accept_tuple_for_concurrent_copy(HeapTuple tuple,
+ Snapshot snapshot,
+ Buffer buffer);
static BlockNumber heapam_scan_get_blocks_done(HeapScanDesc hscan);
@@ -681,6 +685,8 @@ static void
heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap,
Relation OldIndex, bool use_sort,
TransactionId OldestXmin,
+ Snapshot snapshot,
+ LogicalDecodingContext *decoding_ctx,
TransactionId *xid_cutoff,
MultiXactId *multi_cutoff,
double *num_tuples,
@@ -701,6 +707,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap,
bool *isnull;
BufferHeapTupleTableSlot *hslot;
BlockNumber prev_cblock = InvalidBlockNumber;
+ bool concurrent = snapshot != NULL;
+ XLogRecPtr end_of_wal_prev = GetFlushRecPtr(NULL);
/* Remember if it's a system catalog */
is_system_catalog = IsSystemRelation(OldHeap);
@@ -779,8 +787,10 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap,
for (;;)
{
HeapTuple tuple;
+ bool tuple_copied = false;
Buffer buf;
bool isdead;
+ HTSV_Result vis;
CHECK_FOR_INTERRUPTS();
@@ -835,7 +845,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap,
LockBuffer(buf, BUFFER_LOCK_SHARE);
- switch (HeapTupleSatisfiesVacuum(tuple, OldestXmin, buf))
+ switch ((vis = HeapTupleSatisfiesVacuum(tuple, OldestXmin, buf)))
{
case HEAPTUPLE_DEAD:
/* Definitely dead */
@@ -851,14 +861,15 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap,
case HEAPTUPLE_INSERT_IN_PROGRESS:
/*
- * Since we hold exclusive lock on the relation, normally the
- * only way to see this is if it was inserted earlier in our
- * own transaction. However, it can happen in system
+ * As long as we hold exclusive lock on the relation, normally
+ * the only way to see this is if it was inserted earlier in
+ * our own transaction. However, it can happen in system
* catalogs, since we tend to release write lock before commit
- * there. Give a warning if neither case applies; but in any
- * case we had better copy it.
+ * there. Also, there's no exclusive lock during concurrent
+ * processing. Give a warning if neither case applies; but in
+ * any case we had better copy it.
*/
- if (!is_system_catalog &&
+ if (!is_system_catalog && !concurrent &&
!TransactionIdIsCurrentTransactionId(HeapTupleHeaderGetXmin(tuple->t_data)))
elog(WARNING, "concurrent insert in progress within table \"%s\"",
RelationGetRelationName(OldHeap));
@@ -870,7 +881,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap,
/*
* Similar situation to INSERT_IN_PROGRESS case.
*/
- if (!is_system_catalog &&
+ if (!is_system_catalog && !concurrent &&
!TransactionIdIsCurrentTransactionId(HeapTupleHeaderGetUpdateXid(tuple->t_data)))
elog(WARNING, "concurrent delete in progress within table \"%s\"",
RelationGetRelationName(OldHeap));
@@ -884,8 +895,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap,
break;
}
- LockBuffer(buf, BUFFER_LOCK_UNLOCK);
-
if (isdead)
{
*tups_vacuumed += 1;
@@ -896,9 +905,47 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap,
*tups_vacuumed += 1;
*tups_recently_dead -= 1;
}
+
+ LockBuffer(buf, BUFFER_LOCK_UNLOCK);
continue;
}
+ if (concurrent)
+ {
+ /*
+ * Ignore concurrent changes now, they'll be processed later via
+ * logical decoding.
+ *
+ * INSERT_IN_PROGRESS is rejected right away because our snapshot
+ * should represent a point in time which should precede (or be
+ * equal to) the state of transactions as it was when the
+ * "SatisfiesVacuum" test was performed. Thus
+ * accept_tuple_for_concurrent_copy() should not consider the
+ * tuple inserted.
+ */
+ if (vis == HEAPTUPLE_INSERT_IN_PROGRESS)
+ tuple = NULL;
+ else
+ tuple = accept_tuple_for_concurrent_copy(tuple, snapshot,
+ buf);
+ /* Tuple not suitable for the new heap? */
+ if (tuple == NULL)
+ {
+ LockBuffer(buf, BUFFER_LOCK_UNLOCK);
+ continue;
+ }
+
+ /* Remember that we have to free the tuple eventually. */
+ tuple_copied = true;
+ }
+
+ /*
+ * In the concurrent case, we have a copy of the tuple, so we don't
+ * worry whether the source tuple will be deleted / updated after we
+ * release the lock.
+ */
+ LockBuffer(buf, BUFFER_LOCK_UNLOCK);
+
*num_tuples += 1;
if (tuplesort != NULL)
{
@@ -915,7 +962,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap,
{
const int ct_index[] = {
PROGRESS_CLUSTER_HEAP_TUPLES_SCANNED,
- PROGRESS_CLUSTER_HEAP_TUPLES_WRITTEN
+ PROGRESS_CLUSTER_HEAP_TUPLES_INSERTED
};
int64 ct_val[2];
@@ -930,6 +977,33 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap,
ct_val[1] = *num_tuples;
pgstat_progress_update_multi_param(2, ct_index, ct_val);
}
+ if (tuple_copied)
+ heap_freetuple(tuple);
+
+ /*
+ * Process the WAL produced by the load, as well as by other
+ * transactions, so that the replication slot can advance and WAL does
+ * not pile up. Use wal_segment_size as a threshold so that we do not
+ * introduce the decoding overhead too often.
+ *
+ * Of course, we must not apply the changes until the initial load has
+ * completed.
+ *
+ * Note that our insertions into the new table should not be decoded
+ * as we (intentionally) do not write the logical decoding specific
+ * information to WAL.
+ */
+ if (concurrent)
+ {
+ XLogRecPtr end_of_wal;
+
+ end_of_wal = GetFlushRecPtr(NULL);
+ if ((end_of_wal - end_of_wal_prev) > wal_segment_size)
+ {
+ cluster_decode_concurrent_changes(decoding_ctx, end_of_wal);
+ end_of_wal_prev = end_of_wal;
+ }
+ }
}
if (indexScan != NULL)
@@ -973,7 +1047,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap,
values, isnull,
rwstate);
/* Report n_tuples */
- pgstat_progress_update_param(PROGRESS_CLUSTER_HEAP_TUPLES_WRITTEN,
+ pgstat_progress_update_param(PROGRESS_CLUSTER_HEAP_TUPLES_INSERTED,
n_tuples);
}
@@ -2609,6 +2683,53 @@ SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer,
}
}
+/*
+ * Return copy of 'tuple' if it has been inserted according to 'snapshot', or
+ * NULL if the insertion took place in the future. If the tuple is already
+ * marked as deleted or updated by a transaction that 'snapshot' still
+ * considers running, clear the deletion / update XID in the header of the
+ * copied tuple. This way the returned tuple is suitable for insertion into
+ * the new heap.
+ */
+static HeapTuple
+accept_tuple_for_concurrent_copy(HeapTuple tuple, Snapshot snapshot,
+ Buffer buffer)
+{
+ HeapTuple result;
+
+ Assert(snapshot->snapshot_type == SNAPSHOT_MVCC);
+
+ /*
+ * First, check if the tuple insertion is visible by our snapshot.
+ */
+ if (!HeapTupleMVCCInserted(tuple, snapshot, buffer))
+ return NULL;
+
+ result = heap_copytuple(tuple);
+
+ /*
+ * If the tuple was deleted / updated but our snapshot still sees it, we
+ * need to keep it. In that case, clear the information that indicates the
+ * deletion / update. Otherwise the tuple chain would stay incomplete (as
+ * we will reject the new tuple above), and the delete / update would fail
+ * if executed later during logical decoding.
+ */
+ if (TransactionIdIsNormal(HeapTupleHeaderGetRawXmax(result->t_data)) &&
+ HeapTupleMVCCNotDeleted(result, snapshot, buffer))
+ {
+ /* TODO More work needed here?*/
+ result->t_data->t_infomask |= HEAP_XMAX_INVALID;
+ HeapTupleHeaderSetXmax(result->t_data, 0);
+ }
+
+ /*
+ * Accept the tuple even if our snapshot considers it deleted - older
+ * snapshots can still see the tuple, while the decoded transactions
+ * should not try to update / delete it again.
+ */
+ return result;
+}
+
/* ------------------------------------------------------------------------
* Definition of the heap table access method.
diff --git a/src/backend/access/heap/heapam_visibility.c b/src/backend/access/heap/heapam_visibility.c
index e146605bd5..d9be93aadc 100644
--- a/src/backend/access/heap/heapam_visibility.c
+++ b/src/backend/access/heap/heapam_visibility.c
@@ -955,16 +955,31 @@ HeapTupleSatisfiesDirty(HeapTuple htup, Snapshot snapshot,
* did TransactionIdIsInProgress in each call --- to no avail, as long as the
* inserting/deleting transaction was still running --- which was more cycles
* and more contention on ProcArrayLock.
+ *
+ * The checks are split into two functions, HeapTupleMVCCInserted() and
+ * HeapTupleMVCCNotDeleted(), because they are also useful separately.
*/
static bool
HeapTupleSatisfiesMVCC(HeapTuple htup, Snapshot snapshot,
Buffer buffer)
{
- HeapTupleHeader tuple = htup->t_data;
-
Assert(ItemPointerIsValid(&htup->t_self));
Assert(htup->t_tableOid != InvalidOid);
+ return HeapTupleMVCCInserted(htup, snapshot, buffer) &&
+ HeapTupleMVCCNotDeleted(htup, snapshot, buffer);
+}
+
+/*
+ * HeapTupleMVCCInserted
+ * True iff heap tuple was successfully inserted for the given MVCC
+ * snapshot.
+ */
+bool
+HeapTupleMVCCInserted(HeapTuple htup, Snapshot snapshot, Buffer buffer)
+{
+ HeapTupleHeader tuple = htup->t_data;
+
if (!HeapTupleHeaderXminCommitted(tuple))
{
if (HeapTupleHeaderXminInvalid(tuple))
@@ -1073,6 +1088,17 @@ HeapTupleSatisfiesMVCC(HeapTuple htup, Snapshot snapshot,
}
/* by here, the inserting transaction has committed */
+ return true;
+}
+
+/*
+ * HeapTupleMVCCNotDeleted
+ * True iff heap tuple was not deleted for the given MVCC snapshot.
+ */
+bool
+HeapTupleMVCCNotDeleted(HeapTuple htup, Snapshot snapshot, Buffer buffer)
+{
+ HeapTupleHeader tuple = htup->t_data;
if (tuple->t_infomask & HEAP_XMAX_INVALID) /* xid invalid or aborted */
return true;
diff --git a/src/backend/catalog/index.c b/src/backend/catalog/index.c
index 7377912b41..1f9aa5bf8f 100644
--- a/src/backend/catalog/index.c
+++ b/src/backend/catalog/index.c
@@ -1417,22 +1417,7 @@ index_concurrently_create_copy(Relation heapRelation, Oid oldIndexId,
for (int i = 0; i < newInfo->ii_NumIndexAttrs; i++)
opclassOptions[i] = get_attoptions(oldIndexId, i + 1);
- /* Extract statistic targets for each attribute */
- stattargets = palloc0_array(NullableDatum, newInfo->ii_NumIndexAttrs);
- for (int i = 0; i < newInfo->ii_NumIndexAttrs; i++)
- {
- HeapTuple tp;
- Datum dat;
-
- tp = SearchSysCache2(ATTNUM, ObjectIdGetDatum(oldIndexId), Int16GetDatum(i + 1));
- if (!HeapTupleIsValid(tp))
- elog(ERROR, "cache lookup failed for attribute %d of relation %u",
- i + 1, oldIndexId);
- dat = SysCacheGetAttr(ATTNUM, tp, Anum_pg_attribute_attstattarget, &isnull);
- ReleaseSysCache(tp);
- stattargets[i].value = dat;
- stattargets[i].isnull = isnull;
- }
+ stattargets = get_index_stattargets(oldIndexId, newInfo);
/*
* Now create the new index.
@@ -1471,6 +1456,32 @@ index_concurrently_create_copy(Relation heapRelation, Oid oldIndexId,
return newIndexId;
}
+NullableDatum *
+get_index_stattargets(Oid indexid, IndexInfo *indInfo)
+{
+ NullableDatum *stattargets;
+
+ /* Extract statistic targets for each attribute */
+ stattargets = palloc0_array(NullableDatum, indInfo->ii_NumIndexAttrs);
+ for (int i = 0; i < indInfo->ii_NumIndexAttrs; i++)
+ {
+ HeapTuple tp;
+ Datum dat;
+ bool isnull;
+
+ tp = SearchSysCache2(ATTNUM, ObjectIdGetDatum(indexid), Int16GetDatum(i + 1));
+ if (!HeapTupleIsValid(tp))
+ elog(ERROR, "cache lookup failed for attribute %d of relation %u",
+ i + 1, indexid);
+ dat = SysCacheGetAttr(ATTNUM, tp, Anum_pg_attribute_attstattarget, &isnull);
+ ReleaseSysCache(tp);
+ stattargets[i].value = dat;
+ stattargets[i].isnull = isnull;
+ }
+
+ return stattargets;
+}
+
/*
* index_concurrently_build
*
diff --git a/src/backend/catalog/system_views.sql b/src/backend/catalog/system_views.sql
index 7a595c84db..d1e75b5f44 100644
--- a/src/backend/catalog/system_views.sql
+++ b/src/backend/catalog/system_views.sql
@@ -1240,16 +1240,19 @@ CREATE VIEW pg_stat_progress_cluster AS
WHEN 2 THEN 'index scanning heap'
WHEN 3 THEN 'sorting tuples'
WHEN 4 THEN 'writing new heap'
- WHEN 5 THEN 'swapping relation files'
- WHEN 6 THEN 'rebuilding index'
- WHEN 7 THEN 'performing final cleanup'
+ WHEN 5 THEN 'catch-up'
+ WHEN 6 THEN 'swapping relation files'
+ WHEN 7 THEN 'rebuilding index'
+ WHEN 8 THEN 'performing final cleanup'
END AS phase,
CAST(S.param3 AS oid) AS cluster_index_relid,
S.param4 AS heap_tuples_scanned,
- S.param5 AS heap_tuples_written,
- S.param6 AS heap_blks_total,
- S.param7 AS heap_blks_scanned,
- S.param8 AS index_rebuild_count
+ S.param5 AS heap_tuples_inserted,
+ S.param6 AS heap_tuples_updated,
+ S.param7 AS heap_tuples_deleted,
+ S.param8 AS heap_blks_total,
+ S.param9 AS heap_blks_scanned,
+ S.param10 AS index_rebuild_count
FROM pg_stat_get_progress_info('CLUSTER') AS S
LEFT JOIN pg_database D ON S.datid = D.oid;
diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index 99193f5c88..c9cc061c45 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -25,6 +25,10 @@
#include "access/toast_internals.h"
#include "access/transam.h"
#include "access/xact.h"
+#include "access/xlog.h"
+#include "access/xlog_internal.h"
+#include "access/xloginsert.h"
+#include "access/xlogutils.h"
#include "catalog/catalog.h"
#include "catalog/dependency.h"
#include "catalog/heap.h"
@@ -32,6 +36,7 @@
#include "catalog/namespace.h"
#include "catalog/objectaccess.h"
#include "catalog/pg_am.h"
+#include "catalog/pg_control.h"
#include "catalog/pg_inherits.h"
#include "catalog/toasting.h"
#include "commands/cluster.h"
@@ -39,10 +44,15 @@
#include "commands/progress.h"
#include "commands/tablecmds.h"
#include "commands/vacuum.h"
+#include "executor/executor.h"
#include "miscadmin.h"
#include "optimizer/optimizer.h"
#include "pgstat.h"
+#include "replication/decode.h"
+#include "replication/logical.h"
+#include "replication/snapbuild.h"
#include "storage/bufmgr.h"
+#include "storage/ipc.h"
#include "storage/lmgr.h"
#include "storage/predicate.h"
#include "utils/acl.h"
@@ -56,6 +66,8 @@
#include "utils/snapmgr.h"
#include "utils/syscache.h"
+typedef struct RewriteStateData *RewriteState;
+
/*
* 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
@@ -67,17 +79,183 @@ typedef struct
Oid indexOid;
} RelToCluster;
+/*
+ * The following definitions are used for concurrent processing.
+ */
+
+/*
+ * OID of the table being processed by this backend.
+ */
+static Oid clustered_rel = InvalidOid;
+/* The same for its TOAST relation. */
+static Oid clustered_rel_toast = InvalidOid;
+
+/*
+ * The locators are used to avoid logical decoding of data that we do not need
+ * for our table.
+ */
+RelFileLocator clustered_rel_locator = {.relNumber = InvalidOid};
+RelFileLocator clustered_rel_toast_locator = {.relNumber = InvalidOid};
+
+/* XXX Do we also need to mention VACUUM FULL CONCURRENTLY? */
+#define CLUSTER_IN_PROGRESS_MESSAGE \
+ "relation \"%s\" is already being processed by CLUSTER CONCURRENTLY"
+
+/*
+ * Everything we need to call ExecInsertIndexTuples().
+ */
+typedef struct IndexInsertState
+{
+ ResultRelInfo *rri;
+ EState *estate;
+ ExprContext *econtext;
+
+ Relation ident_index;
+} IndexInsertState;
-static void cluster_multiple_rels(List *rtcs, ClusterParams *params);
-static void rebuild_relation(Relation OldHeap, Relation index, bool verbose);
+/*
+ * Catalog information to check if another backend changed the relation in
+ * such a way that makes CLUSTER CONCURRENTLY unable to continue. Such changes
+ * are possible because cluster_rel() has to release its lock on the relation
+ * in order to acquire AccessExclusiveLock that it needs to swap the relation
+ * files.
+ *
+ * The most obvious problem is that the tuple descriptor has changed, since
+ * then the tuples we try to insert into the new storage are not guaranteed to
+ * fit into the storage.
+ *
+ * Another problem is relfilenode changed by another backend. It's not
+ * necessarily a correctness issue (e.g. when the other backend ran
+ * cluster_rel()), but it's safer for us to terminate the table processing in
+ * such cases. However, this information is also needs to be checked during
+ * logical decoding, so we store it in global variables clustered_rel_locator
+ * and clustered_rel_toast_locator above.
+ *
+ * Where possible, commands which might change the relation in an incompatible
+ * way should check if CLUSTER CONCURRENTLY is running, before they start to
+ * do the actual changes (see is_concurrent_cluster_in_progress()). Anything
+ * else must be caught by check_catalog_changes(), which uses this structure.
+ */
+typedef struct CatalogState
+{
+ /* Tuple descriptor of the relation. */
+ TupleDesc tupdesc;
+
+ /* The number of indexes tracked. */
+ int ninds;
+ /* The index OIDs. */
+ Oid *ind_oids;
+ /* The index tuple descriptors. */
+ TupleDesc *ind_tupdescs;
+
+ /* The following are copies of the corresponding fields of pg_class. */
+ char relpersistence;
+ char replident;
+
+ /* rd_replidindex */
+ Oid replidindex;
+} CatalogState;
+
+/* The WAL segment being decoded. */
+static XLogSegNo cluster_current_segment = 0;
+
+static void cluster_multiple_rels(List *rtcs, ClusterParams *params,
+ LOCKMODE lockmode, bool isTopLevel);
+static void rebuild_relation(Relation OldHeap, Relation index, bool verbose,
+ bool concurrent, bool is_vacuum);
static void copy_table_data(Relation NewHeap, Relation OldHeap, Relation OldIndex,
+ Snapshot snapshot, LogicalDecodingContext *decoding_ctx,
bool verbose, bool *pSwapToastByContent,
TransactionId *pFreezeXid, MultiXactId *pCutoffMulti);
static List *get_tables_to_cluster(MemoryContext cluster_context);
static List *get_tables_to_cluster_partitioned(MemoryContext cluster_context,
Oid indexOid);
static bool cluster_is_permitted_for_relation(Oid relid, Oid userid);
+static void check_concurrent_cluster_requirements(Relation rel,
+ bool isTopLevel,
+ bool isCluster);
+static void begin_concurrent_cluster(Relation *rel_p, Relation *index_p,
+ bool *entered_p);
+static void end_concurrent_cluster(bool error);
+static void cluster_before_shmem_exit_callback(int code, Datum arg);
+static CatalogState *get_catalog_state(Relation rel, bool is_vacuum);
+static void free_catalog_state(CatalogState *state);
+static void check_catalog_changes(Relation rel, CatalogState *cat_state);
+static LogicalDecodingContext *setup_logical_decoding(Oid relid,
+ const char *slotname,
+ TupleDesc tupdesc);
+static HeapTuple get_changed_tuple(char *change);
+static void apply_concurrent_changes(ClusterDecodingState *dstate,
+ Relation rel, ScanKey key, int nkeys,
+ IndexInsertState *iistate);
+static void apply_concurrent_insert(Relation rel, ConcurrentChange *change,
+ HeapTuple tup, IndexInsertState *iistate,
+ TupleTableSlot *index_slot);
+static void apply_concurrent_update(Relation rel, HeapTuple tup,
+ HeapTuple tup_target,
+ ConcurrentChange *change,
+ IndexInsertState *iistate,
+ TupleTableSlot *index_slot);
+static void apply_concurrent_delete(Relation rel, HeapTuple tup_target,
+ ConcurrentChange *change);
+static HeapTuple find_target_tuple(Relation rel, ScanKey key, int nkeys,
+ HeapTuple tup_key,
+ IndexInsertState *iistate,
+ TupleTableSlot *ident_slot,
+ IndexScanDesc *scan_p);
+static void process_concurrent_changes(LogicalDecodingContext *ctx,
+ XLogRecPtr end_of_wal,
+ Relation rel_dst,
+ Relation rel_src,
+ ScanKey ident_key,
+ int ident_key_nentries,
+ IndexInsertState *iistate);
+static IndexInsertState *get_index_insert_state(Relation relation,
+ Oid ident_index_id);
+static ScanKey build_identity_key(Oid ident_idx_oid, Relation rel_src,
+ int *nentries);
+static void free_index_insert_state(IndexInsertState *iistate);
+static void cleanup_logical_decoding(LogicalDecodingContext *ctx);
+static void rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
+ Relation cl_index,
+ CatalogState *cat_state,
+ LogicalDecodingContext *ctx,
+ bool swap_toast_by_content,
+ TransactionId frozenXid,
+ MultiXactId cutoffMulti);
+static List *build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes);
+
+/*
+ * Use this API when relation needs to be unlocked, closed and re-opened. If
+ * the relation got dropped while being unlocked, raise ERROR that mentions
+ * the relation name rather than OID.
+ */
+typedef struct RelReopenInfo
+{
+ /*
+ * The relation to be closed. Pointer to the value is stored here so that
+ * the user gets his reference updated automatically on re-opening.
+ *
+ * When calling unlock_and_close_relations(), 'relid' can be passed
+ * instead of 'rel_p' when the caller only needs to gather information for
+ * subsequent opening.
+ */
+ Relation *rel_p;
+ Oid relid;
+ char relkind;
+ LOCKMODE lockmode_orig; /* The existing lock mode */
+ LOCKMODE lockmode_new; /* The lock mode after the relation is
+ * re-opened */
+
+ char *relname; /* Relation name, initialized automatically. */
+} RelReopenInfo;
+
+static void init_rel_reopen_info(RelReopenInfo *rri, Relation *rel_p,
+ Oid relid, LOCKMODE lockmode_orig,
+ LOCKMODE lockmode_new);
+static void unlock_and_close_relations(RelReopenInfo *rels, int nrel);
+static void reopen_relations(RelReopenInfo *rels, int nrel);
/*---------------------------------------------------------------------------
* This cluster code allows for clustering multiple tables at once. Because
@@ -109,10 +287,12 @@ cluster(ParseState *pstate, ClusterStmt *stmt, bool isTopLevel)
ListCell *lc;
ClusterParams params = {0};
bool verbose = false;
+ bool concurrent = false;
Relation rel = NULL;
Oid indexOid = InvalidOid;
MemoryContext cluster_context;
List *rtcs;
+ LOCKMODE lockmode;
/* Parse option list */
foreach(lc, stmt->params)
@@ -121,6 +301,8 @@ cluster(ParseState *pstate, ClusterStmt *stmt, bool isTopLevel)
if (strcmp(opt->defname, "verbose") == 0)
verbose = defGetBoolean(opt);
+ else if (strcmp(opt->defname, "concurrently") == 0)
+ concurrent = defGetBoolean(opt);
else
ereport(ERROR,
(errcode(ERRCODE_SYNTAX_ERROR),
@@ -129,20 +311,30 @@ cluster(ParseState *pstate, ClusterStmt *stmt, bool isTopLevel)
parser_errposition(pstate, opt->location)));
}
- params.options = (verbose ? CLUOPT_VERBOSE : 0);
+ params.options =
+ (verbose ? CLUOPT_VERBOSE : 0) |
+ (concurrent ? CLUOPT_CONCURRENT : 0);
+
+ /*
+ * Determine the lock mode expected by cluster_rel().
+ *
+ * In the exclusive case, we obtain AccessExclusiveLock right away to
+ * avoid lock-upgrade hazard in the single-transaction case. In the
+ * CONCURRENT case, the AccessExclusiveLock will only be used at the end
+ * of processing, supposedly for very short time. Until then, we'll have
+ * to unlock the relation temporarily, so there's no lock-upgrade hazard.
+ */
+ lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ?
+ AccessExclusiveLock : ShareUpdateExclusiveLock;
if (stmt->relation != NULL)
{
/* This is the single-relation case. */
Oid tableOid;
- /*
- * Find, lock, and check permissions on the table. We obtain
- * AccessExclusiveLock right away to avoid lock-upgrade hazard in the
- * single-transaction case.
- */
+ /* Find, lock, and check permissions on the table. */
tableOid = RangeVarGetRelidExtended(stmt->relation,
- AccessExclusiveLock,
+ lockmode,
0,
RangeVarCallbackMaintainsTable,
NULL);
@@ -194,7 +386,7 @@ cluster(ParseState *pstate, ClusterStmt *stmt, bool isTopLevel)
/* For non-partitioned tables, do what we came here to do. */
if (rel->rd_rel->relkind != RELKIND_PARTITIONED_TABLE)
{
- cluster_rel(rel, indexOid, ¶ms);
+ cluster_rel(rel, indexOid, ¶ms, isTopLevel, false);
/* cluster_rel closes the relation, but keeps lock */
return;
@@ -202,10 +394,29 @@ cluster(ParseState *pstate, ClusterStmt *stmt, bool isTopLevel)
}
/*
- * By here, we know we are in a multi-table situation. In order to avoid
- * holding locks for too long, we want to process each table in its own
- * transaction. This forces us to disallow running inside a user
- * transaction block.
+ * By here, we know we are in a multi-table situation.
+ *
+ * Concurrent processing is currently considered rather special (e.g. in
+ * terms of resources consumed) so it is not performed in bulk.
+ */
+ if (params.options & CLUOPT_CONCURRENT)
+ {
+ if (rel != NULL)
+ {
+ Assert(rel->rd_rel->relkind == RELKIND_PARTITIONED_TABLE);
+ ereport(ERROR,
+ (errmsg("CLUSTER (CONCURRENTLY) not supported for partitioned tables"),
+ errhint("Consider running the command for individual partitions.")));
+ }
+ else
+ ereport(ERROR,
+ (errmsg("CLUSTER (CONCURRENTLY) requires explicit table name")));
+ }
+
+ /*
+ * In order to avoid holding locks for too long, we want to process each
+ * table in its own transaction. This forces us to disallow running
+ * inside a user transaction block.
*/
PreventInTransactionBlock(isTopLevel, "CLUSTER");
@@ -230,11 +441,14 @@ cluster(ParseState *pstate, ClusterStmt *stmt, bool isTopLevel)
if (rel != NULL)
{
Assert(rel->rd_rel->relkind == RELKIND_PARTITIONED_TABLE);
+ /* See the ereport() above. */
+ Assert((params.options & CLUOPT_CONCURRENT) == 0);
+
check_index_is_clusterable(rel, indexOid, AccessShareLock);
rtcs = get_tables_to_cluster_partitioned(cluster_context, indexOid);
/* close relation, releasing lock on parent table */
- table_close(rel, AccessExclusiveLock);
+ table_close(rel, lockmode);
}
else
{
@@ -243,7 +457,7 @@ cluster(ParseState *pstate, ClusterStmt *stmt, bool isTopLevel)
}
/* Do the job. */
- cluster_multiple_rels(rtcs, ¶ms);
+ cluster_multiple_rels(rtcs, ¶ms, lockmode, isTopLevel);
/* Start a new transaction for the cleanup work. */
StartTransactionCommand();
@@ -260,7 +474,8 @@ cluster(ParseState *pstate, ClusterStmt *stmt, bool isTopLevel)
* return.
*/
static void
-cluster_multiple_rels(List *rtcs, ClusterParams *params)
+cluster_multiple_rels(List *rtcs, ClusterParams *params, LOCKMODE lockmode,
+ bool isTopLevel)
{
ListCell *lc;
@@ -280,10 +495,10 @@ cluster_multiple_rels(List *rtcs, ClusterParams *params)
/* functions in indexes may want a snapshot set */
PushActiveSnapshot(GetTransactionSnapshot());
- rel = table_open(rtc->tableOid, AccessExclusiveLock);
+ rel = table_open(rtc->tableOid, lockmode);
/* Process this table */
- cluster_rel(rel, rtc->indexOid, params);
+ cluster_rel(rel, rtc->indexOid, params, isTopLevel, false);
/* cluster_rel closes the relation, but keeps lock */
PopActiveSnapshot();
@@ -306,9 +521,16 @@ cluster_multiple_rels(List *rtcs, ClusterParams *params)
* If indexOid is InvalidOid, the table will be rewritten in physical order
* instead of index order. This is the new implementation of VACUUM FULL,
* and error messages should refer to the operation as VACUUM not CLUSTER.
+ *
+ * Note that, in the concurrent case, the function releases the lock at some
+ * point, in order to get AccessExclusiveLock for the final steps (i.e. to
+ * swap the relation files). To make things simpler, the caller should expect
+ * OldHeap to be closed on return, regardless CLUOPT_CONCURRENT. (The
+ * AccessExclusiveLock is kept till the end of the transaction.)
*/
void
-cluster_rel(Relation OldHeap, Oid indexOid, ClusterParams *params)
+cluster_rel(Relation OldHeap, Oid indexOid, ClusterParams *params,
+ bool isTopLevel, bool isVacuum)
{
Oid tableOid = RelationGetRelid(OldHeap);
Oid save_userid;
@@ -317,8 +539,46 @@ cluster_rel(Relation OldHeap, Oid indexOid, ClusterParams *params)
bool verbose = ((params->options & CLUOPT_VERBOSE) != 0);
bool recheck = ((params->options & CLUOPT_RECHECK) != 0);
Relation index;
+ bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0);
+ LOCKMODE lmode;
+ bool entered, success;
+
+ /*
+ * Check that the correct lock is held. The lock mode is
+ * AccessExclusiveLock for normal processing and ShareUpdateExclusiveLock
+ * for concurrent processing (so that SELECT, INSERT, UPDATE and DELETE
+ * commands work, but cluster_rel() cannot be called concurrently for the
+ * same relation).
+ */
+ lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock;
+
+ /*
+ * Skip the relation if it's being processed concurrently. In such a case,
+ * we cannot rely on a lock because the other backend needs to release it
+ * temporarily at some point.
+ *
+ * This check should not take place until we have a lock that prevents
+ * another backend from starting VACUUM FULL / CLUSTER CONCURRENTLY after
+ * our check.
+ */
+ Assert(CheckRelationLockedByMe(OldHeap, lmode, false));
+ if (is_concurrent_cluster_in_progress(tableOid))
+ {
+ ereport(NOTICE,
+ (errmsg(CLUSTER_IN_PROGRESS_MESSAGE,
+ RelationGetRelationName(OldHeap))));
+ table_close(OldHeap, lmode);
+ return;
+ }
+
+ /* There are specific requirements on concurrent processing. */
+ if (concurrent)
+ {
+ check_concurrent_cluster_requirements(OldHeap, isTopLevel,
+ OidIsValid(indexOid));
- Assert(CheckRelationLockedByMe(OldHeap, AccessExclusiveLock, false));
+ check_relation_is_clusterable_concurrently(OldHeap, isVacuum);
+ }
/* Check for user-requested abort. */
CHECK_FOR_INTERRUPTS();
@@ -355,7 +615,7 @@ cluster_rel(Relation OldHeap, Oid indexOid, ClusterParams *params)
/* Check that the user still has privileges for the relation */
if (!cluster_is_permitted_for_relation(tableOid, save_userid))
{
- relation_close(OldHeap, AccessExclusiveLock);
+ relation_close(OldHeap, lmode);
goto out;
}
@@ -370,7 +630,7 @@ cluster_rel(Relation OldHeap, Oid indexOid, ClusterParams *params)
*/
if (RELATION_IS_OTHER_TEMP(OldHeap))
{
- relation_close(OldHeap, AccessExclusiveLock);
+ relation_close(OldHeap, lmode);
goto out;
}
@@ -381,7 +641,7 @@ cluster_rel(Relation OldHeap, Oid indexOid, ClusterParams *params)
*/
if (!SearchSysCacheExists1(RELOID, ObjectIdGetDatum(indexOid)))
{
- relation_close(OldHeap, AccessExclusiveLock);
+ relation_close(OldHeap, lmode);
goto out;
}
@@ -392,7 +652,7 @@ cluster_rel(Relation OldHeap, Oid indexOid, ClusterParams *params)
if ((params->options & CLUOPT_RECHECK_ISCLUSTERED) != 0 &&
!get_index_isclustered(indexOid))
{
- relation_close(OldHeap, AccessExclusiveLock);
+ relation_close(OldHeap, lmode);
goto out;
}
}
@@ -408,6 +668,11 @@ cluster_rel(Relation OldHeap, Oid indexOid, ClusterParams *params)
ereport(ERROR,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("cannot cluster a shared catalog")));
+ /*
+ * The CONCURRENT case should have been rejected earlier because it does
+ * not support system catalogs.
+ */
+ Assert(!(OldHeap->rd_rel->relisshared && concurrent));
/*
* Don't process temp tables of other backends ... their local buffer
@@ -435,7 +700,7 @@ cluster_rel(Relation OldHeap, Oid indexOid, ClusterParams *params)
if (OidIsValid(indexOid))
{
/* verify the index is good and lock it */
- check_index_is_clusterable(OldHeap, indexOid, AccessExclusiveLock);
+ check_index_is_clusterable(OldHeap, indexOid, lmode);
/* also open it */
index = index_open(indexOid, NoLock);
}
@@ -452,7 +717,8 @@ cluster_rel(Relation OldHeap, Oid indexOid, ClusterParams *params)
if (OldHeap->rd_rel->relkind == RELKIND_MATVIEW &&
!RelationIsPopulated(OldHeap))
{
- relation_close(OldHeap, AccessExclusiveLock);
+ index_close(index, lmode);
+ relation_close(OldHeap, lmode);
goto out;
}
@@ -465,11 +731,42 @@ cluster_rel(Relation OldHeap, Oid indexOid, ClusterParams *params)
* invalid, because we move tuples around. Promote them to relation
* locks. Predicate locks on indexes will be promoted when they are
* reindexed.
+ *
+ * During concurrent processing, the heap as well as its indexes stay in
+ * operation, so we postpone this step until they are locked using
+ * AccessExclusiveLock near the end of the processing.
*/
- TransferPredicateLocksToHeapRelation(OldHeap);
+ if (!concurrent)
+ TransferPredicateLocksToHeapRelation(OldHeap);
/* rebuild_relation does all the dirty work */
- rebuild_relation(OldHeap, index, verbose);
+ entered = false;
+ success = false;
+ PG_TRY();
+ {
+ /*
+ * For concurrent processing, make sure other transactions treat this
+ * table as if it was a system / user catalog, and WAL the relevant
+ * additional information. ERROR is raised if another backend is
+ * processing the same table.
+ */
+ if (concurrent)
+ {
+ Relation *index_p = index ? &index : NULL;
+
+ begin_concurrent_cluster(&OldHeap, index_p, &entered);
+ }
+
+ rebuild_relation(OldHeap, index, verbose, concurrent, isVacuum);
+ success = true;
+ }
+ PG_FINALLY();
+ {
+ if (concurrent && entered)
+ end_concurrent_cluster(!success);
+ }
+ PG_END_TRY();
+
/* rebuild_relation closes OldHeap, and index if valid */
out:
@@ -615,18 +912,86 @@ mark_index_clustered(Relation rel, Oid indexOid, bool is_internal)
table_close(pg_index, RowExclusiveLock);
}
+/*
+ * Check if the CONCURRENTLY option is legal for the relation.
+ */
+void
+check_relation_is_clusterable_concurrently(Relation rel, bool is_vacuum)
+{
+ char relpersistence, replident;
+ Oid ident_idx;
+ const char *cmd = is_vacuum ? "VACUUM" : "CLUSTER";
+
+ /* Data changes in system relations are not logically decoded. */
+ if (IsCatalogRelation(rel))
+ ereport(ERROR,
+ (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
+ errmsg("cannot process relation \"%s\"",
+ RelationGetRelationName(rel)),
+ errhint("%s (CONCURRENTLY) is not supported for catalog relations.",
+ cmd)));
+
+ if (IsToastRelation(rel))
+ ereport(ERROR,
+ (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
+ errmsg("cannot process relation \"%s\"",
+ RelationGetRelationName(rel)),
+ errhint("%s (CONCURRENTLY) is not supported for TOAST relations, unless the main relation is processed too.",
+ cmd)));
+
+ relpersistence = rel->rd_rel->relpersistence;
+ if (relpersistence != RELPERSISTENCE_PERMANENT)
+ ereport(ERROR,
+ (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
+ errmsg("cannot process relation \"%s\"",
+ RelationGetRelationName(rel)),
+ errhint("%s (CONCURRENTLY) is only allowed for permanent relations.",
+ cmd)));
+
+ /* With NOTHING, WAL does not contain the old tuple. */
+ replident = rel->rd_rel->relreplident;
+ if (replident == REPLICA_IDENTITY_NOTHING)
+ ereport(ERROR,
+ (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
+ errmsg("cannot process relation \"%s\"",
+ RelationGetRelationName(rel)),
+ errhint("Relation \"%s\" has insufficient replication identity.",
+ RelationGetRelationName(rel))));
+
+ /*
+ * Identity index is not set if the replica identity is FULL, but PK might
+ * exist in such a case.
+ */
+ ident_idx = RelationGetReplicaIndex(rel);
+ if (!OidIsValid(ident_idx) && OidIsValid(rel->rd_pkindex))
+ ident_idx = rel->rd_pkindex;
+ if (!OidIsValid(ident_idx))
+ ereport(ERROR,
+ (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
+ errmsg("cannot process relation \"%s\"",
+ RelationGetRelationName(rel)),
+ (errhint("Relation \"%s\" has no identity index.",
+ RelationGetRelationName(rel)))));
+}
+
/*
* rebuild_relation: rebuild an existing relation in index or physical order
*
- * OldHeap: table to rebuild.
+ * OldHeap: table to rebuild. See cluster_rel() for comments on the required
+ * lock strength.
+ *
* index: index to cluster by, or NULL to rewrite in physical order.
*
- * On entry, heap and index (if one is given) must be open, and
- * AccessExclusiveLock held on them.
- * On exit, they are closed, but locks on them are not released.
+ * On entry, heap and index (if one is given) must be open, and the
+ * appropriate lock held on them (AccessExclusiveLock for exclusive processing
+ * and ShareUpdateExclusiveLock for concurrent processing)..
+ *
+ * On exit, they are closed, but still locked with AccessExclusiveLock (The
+ * function handles the lock upgrade if 'concurrent' is true.)
*/
static void
-rebuild_relation(Relation OldHeap, Relation index, bool verbose)
+rebuild_relation(Relation OldHeap, Relation index, bool verbose,
+ bool concurrent, bool is_vacuum)
{
Oid tableOid = RelationGetRelid(OldHeap);
Oid accessMethod = OldHeap->rd_rel->relam;
@@ -634,13 +999,81 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose)
Oid OIDNewHeap;
Relation NewHeap;
char relpersistence;
- bool is_system_catalog;
bool swap_toast_by_content;
TransactionId frozenXid;
MultiXactId cutoffMulti;
+ NameData slotname;
+ LogicalDecodingContext *ctx = NULL;
+ Snapshot snapshot = NULL;
+ CatalogState *cat_state = NULL;
+ LOCKMODE lmode;
+
+ lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock;
+
+ Assert(CheckRelationLockedByMe(OldHeap, lmode, false) &&
+ (index == NULL || CheckRelationLockedByMe(index, lmode, false)));
+
+ if (concurrent)
+ {
+ TupleDesc tupdesc;
+ RelReopenInfo rri[2];
+ int nrel;
+
+ /*
+ * CLUSTER CONCURRENTLY is not allowed in a transaction block, so this
+ * should never fire.
+ */
+ Assert(GetTopTransactionIdIfAny() == InvalidTransactionId);
+
+ /*
+ * A single backend should not execute multiple CLUSTER commands at a
+ * time, so use PID to make the slot unique.
+ */
+ snprintf(NameStr(slotname), NAMEDATALEN, "cluster_%d", MyProcPid);
+
+ /*
+ * Gather catalog information so that we can check later if the old
+ * relation has not changed while unlocked.
+ *
+ * Since this function also checks if the relation can be processed,
+ * it's important to call it before we spend notable amount of time to
+ * setup the logical decoding. Not sure though if it's necessary to do
+ * it even earlier.
+ */
+ cat_state = get_catalog_state(OldHeap, is_vacuum);
+
+ tupdesc = CreateTupleDescCopy(RelationGetDescr(OldHeap));
+
+ /*
+ * Unlock the relation (and possibly the clustering index) to avoid
+ * deadlock because setup_logical_decoding() will wait for all the
+ * running transactions (with XID assigned) to finish. Some of those
+ * transactions might be waiting for a lock on our relation.
+ */
+ nrel = 0;
+ init_rel_reopen_info(&rri[nrel++], &OldHeap, InvalidOid,
+ ShareUpdateExclusiveLock,
+ ShareUpdateExclusiveLock);
+ if (index)
+ init_rel_reopen_info(&rri[nrel++], &index, InvalidOid,
+ ShareUpdateExclusiveLock,
+ ShareUpdateExclusiveLock);
+ unlock_and_close_relations(rri, nrel);
+
+ /* Prepare to capture the concurrent data changes. */
+ ctx = setup_logical_decoding(tableOid, NameStr(slotname), tupdesc);
+
+ /* Lock the table (and index) again. */
+ reopen_relations(rri, nrel);
+
+ /*
+ * Check if a 'tupdesc' could have changed while the relation was
+ * unlocked.
+ */
+ check_catalog_changes(OldHeap, cat_state);
- Assert(CheckRelationLockedByMe(OldHeap, AccessExclusiveLock, false) &&
- (index == NULL || CheckRelationLockedByMe(index, AccessExclusiveLock, false)));
+ snapshot = SnapBuildInitialSnapshotForCluster(ctx->snapshot_builder);
+ }
if (index)
/* Mark the correct index as clustered */
@@ -648,7 +1081,6 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose)
/* Remember info about rel before closing OldHeap */
relpersistence = OldHeap->rd_rel->relpersistence;
- is_system_catalog = IsSystemRelation(OldHeap);
/*
* Create the transient table that will receive the re-ordered data.
@@ -664,30 +1096,51 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose)
NewHeap = table_open(OIDNewHeap, NoLock);
/* Copy the heap data into the new table in the desired order */
- copy_table_data(NewHeap, OldHeap, index, verbose,
+ copy_table_data(NewHeap, OldHeap, index, snapshot, ctx, verbose,
&swap_toast_by_content, &frozenXid, &cutoffMulti);
+ if (concurrent)
+ {
+ rebuild_relation_finish_concurrent(NewHeap, OldHeap, index,
+ cat_state, ctx,
+ swap_toast_by_content,
+ frozenXid, cutoffMulti);
+
+ pgstat_progress_update_param(PROGRESS_CLUSTER_PHASE,
+ PROGRESS_CLUSTER_PHASE_FINAL_CLEANUP);
+
+ /* Done with decoding. */
+ FreeSnapshot(snapshot);
+ free_catalog_state(cat_state);
+ cleanup_logical_decoding(ctx);
+ ReplicationSlotRelease();
+ ReplicationSlotDrop(NameStr(slotname), false);
+ }
+ else
+ {
+ bool is_system_catalog = IsSystemRelation(OldHeap);
- /* Close relcache entries, but keep lock until transaction commit */
- table_close(OldHeap, NoLock);
- if (index)
- index_close(index, NoLock);
+ /* Close relcache entries, but keep lock until transaction commit */
+ table_close(OldHeap, NoLock);
+ if (index)
+ index_close(index, NoLock);
- /*
- * Close the new relation so it can be dropped as soon as the storage is
- * swapped. The relation is not visible to others, so no need to unlock it
- * explicitly.
- */
- table_close(NewHeap, NoLock);
+ /*
+ * Close the new relation so it can be dropped as soon as the storage
+ * is swapped. The relation is not visible to others, so no need to
+ * unlock it explicitly.
+ */
+ table_close(NewHeap, NoLock);
- /*
- * Swap the physical files of the target and transient tables, then
- * rebuild the target's indexes and throw away the transient table.
- */
- finish_heap_swap(tableOid, OIDNewHeap, is_system_catalog,
- swap_toast_by_content, false, true,
- frozenXid, cutoffMulti,
- relpersistence);
+ /*
+ * Swap the physical files of the target and transient tables, then
+ * rebuild the target's indexes and throw away the transient table.
+ */
+ finish_heap_swap(tableOid, OIDNewHeap, is_system_catalog,
+ swap_toast_by_content, false, true, true,
+ frozenXid, cutoffMulti,
+ relpersistence);
+ }
}
@@ -822,15 +1275,19 @@ make_new_heap(Oid OIDOldHeap, Oid NewTableSpace, Oid NewAccessMethod,
/*
* Do the physical copying of table data.
*
+ * 'snapshot' and 'decoding_ctx': see table_relation_copy_for_cluster(). Pass
+ * iff concurrent processing is required.
+ *
* There are three output parameters:
* *pSwapToastByContent is set true if toast tables must be swapped by content.
* *pFreezeXid receives the TransactionId used as freeze cutoff point.
* *pCutoffMulti receives the MultiXactId used as a cutoff point.
*/
static void
-copy_table_data(Relation NewHeap, Relation OldHeap, Relation OldIndex, bool verbose,
- bool *pSwapToastByContent, TransactionId *pFreezeXid,
- MultiXactId *pCutoffMulti)
+copy_table_data(Relation NewHeap, Relation OldHeap, Relation OldIndex,
+ Snapshot snapshot, LogicalDecodingContext *decoding_ctx,
+ bool verbose, bool *pSwapToastByContent,
+ TransactionId *pFreezeXid, MultiXactId *pCutoffMulti)
{
Relation relRelation;
HeapTuple reltup;
@@ -847,6 +1304,7 @@ copy_table_data(Relation NewHeap, Relation OldHeap, Relation OldIndex, bool verb
int elevel = verbose ? INFO : DEBUG2;
PGRUsage ru0;
char *nspname;
+ bool concurrent = snapshot != NULL;
pg_rusage_init(&ru0);
@@ -873,8 +1331,12 @@ copy_table_data(Relation NewHeap, Relation OldHeap, Relation OldIndex, bool verb
*
* We don't need to open the toast relation here, just lock it. The lock
* will be held till end of transaction.
+ *
+ * In the CONCURRENT case, the lock does not help because we need to
+ * release it temporarily at some point. Instead, we expect VACUUM /
+ * CLUSTER to skip tables which are present in ClusteredRelsHash.
*/
- if (OldHeap->rd_rel->reltoastrelid)
+ if (OldHeap->rd_rel->reltoastrelid && !concurrent)
LockRelationOid(OldHeap->rd_rel->reltoastrelid, AccessExclusiveLock);
/*
@@ -950,8 +1412,46 @@ copy_table_data(Relation NewHeap, Relation OldHeap, Relation OldIndex, bool verb
* provided, else plain seqscan.
*/
if (OldIndex != NULL && OldIndex->rd_rel->relam == BTREE_AM_OID)
+ {
+ ResourceOwner oldowner = CurrentResourceOwner;
+
+ /*
+ * In the CONCURRENT case, do the planning in a subtransaction so that
+ * we don't leave any additional locks behind us that we cannot
+ * release easily.
+ */
+ if (concurrent)
+ {
+ Assert(CheckRelationLockedByMe(OldHeap, ShareUpdateExclusiveLock,
+ false));
+ Assert(CheckRelationLockedByMe(OldIndex, ShareUpdateExclusiveLock,
+ false));
+ BeginInternalSubTransaction("plan_cluster_use_sort");
+ }
+
use_sort = plan_cluster_use_sort(RelationGetRelid(OldHeap),
RelationGetRelid(OldIndex));
+
+ if (concurrent)
+ {
+ PgBackendProgress progress;
+
+ /*
+ * Command progress reporting gets terminated at subtransaction
+ * end. Save the status so it can be eventually restored.
+ */
+ memcpy(&progress, &MyBEEntry->st_progress,
+ sizeof(PgBackendProgress));
+
+ /* Release the locks by aborting the subtransaction. */
+ RollbackAndReleaseCurrentSubTransaction();
+
+ /* Restore the progress reporting status. */
+ pgstat_progress_restore_state(&progress);
+
+ CurrentResourceOwner = oldowner;
+ }
+ }
else
use_sort = false;
@@ -980,7 +1480,9 @@ copy_table_data(Relation NewHeap, Relation OldHeap, Relation OldIndex, bool verb
* values (e.g. because the AM doesn't use freezing).
*/
table_relation_copy_for_cluster(OldHeap, NewHeap, OldIndex, use_sort,
- cutoffs.OldestXmin, &cutoffs.FreezeLimit,
+ cutoffs.OldestXmin, snapshot,
+ decoding_ctx,
+ &cutoffs.FreezeLimit,
&cutoffs.MultiXactCutoff,
&num_tuples, &tups_vacuumed,
&tups_recently_dead);
@@ -989,7 +1491,11 @@ copy_table_data(Relation NewHeap, Relation OldHeap, Relation OldIndex, bool verb
*pFreezeXid = cutoffs.FreezeLimit;
*pCutoffMulti = cutoffs.MultiXactCutoff;
- /* Reset rd_toastoid just to be tidy --- it shouldn't be looked at again */
+ /*
+ * Reset rd_toastoid just to be tidy --- it shouldn't be looked at
+ * again. In the CONCURRENT case, we need to set it again before applying
+ * the concurrent changes.
+ */
NewHeap->rd_toastoid = InvalidOid;
num_pages = RelationGetNumberOfBlocks(NewHeap);
@@ -1442,14 +1948,13 @@ finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap,
bool swap_toast_by_content,
bool check_constraints,
bool is_internal,
+ bool reindex,
TransactionId frozenXid,
MultiXactId cutoffMulti,
char newrelpersistence)
{
ObjectAddress object;
Oid mapped_tables[4];
- int reindex_flags;
- ReindexParams reindex_params = {0};
int i;
/* Report that we are now swapping relation files */
@@ -1475,39 +1980,46 @@ finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap,
if (is_system_catalog)
CacheInvalidateCatalog(OIDOldHeap);
- /*
- * Rebuild each index on the relation (but not the toast table, which is
- * all-new at this point). It is important to do this before the DROP
- * step because if we are processing a system catalog that will be used
- * during DROP, we want to have its indexes available. There is no
- * advantage to the other order anyway because this is all transactional,
- * so no chance to reclaim disk space before commit. We do not need a
- * final CommandCounterIncrement() because reindex_relation does it.
- *
- * Note: because index_build is called via reindex_relation, it will never
- * set indcheckxmin true for the indexes. This is OK even though in some
- * sense we are building new indexes rather than rebuilding existing ones,
- * because the new heap won't contain any HOT chains at all, let alone
- * broken ones, so it can't be necessary to set indcheckxmin.
- */
- reindex_flags = REINDEX_REL_SUPPRESS_INDEX_USE;
- if (check_constraints)
- reindex_flags |= REINDEX_REL_CHECK_CONSTRAINTS;
+ if (reindex)
+ {
+ int reindex_flags;
+ ReindexParams reindex_params = {0};
- /*
- * Ensure that the indexes have the same persistence as the parent
- * relation.
- */
- if (newrelpersistence == RELPERSISTENCE_UNLOGGED)
- reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED;
- else if (newrelpersistence == RELPERSISTENCE_PERMANENT)
- reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT;
+ /*
+ * Rebuild each index on the relation (but not the toast table, which
+ * is all-new at this point). It is important to do this before the
+ * DROP step because if we are processing a system catalog that will
+ * be used during DROP, we want to have its indexes available. There
+ * is no advantage to the other order anyway because this is all
+ * transactional, so no chance to reclaim disk space before commit.
+ * We do not need a final CommandCounterIncrement() because
+ * reindex_relation does it.
+ *
+ * Note: because index_build is called via reindex_relation, it will never
+ * set indcheckxmin true for the indexes. This is OK even though in some
+ * sense we are building new indexes rather than rebuilding existing ones,
+ * because the new heap won't contain any HOT chains at all, let alone
+ * broken ones, so it can't be necessary to set indcheckxmin.
+ */
+ reindex_flags = REINDEX_REL_SUPPRESS_INDEX_USE;
+ if (check_constraints)
+ reindex_flags |= REINDEX_REL_CHECK_CONSTRAINTS;
- /* Report that we are now reindexing relations */
- pgstat_progress_update_param(PROGRESS_CLUSTER_PHASE,
- PROGRESS_CLUSTER_PHASE_REBUILD_INDEX);
+ /*
+ * Ensure that the indexes have the same persistence as the parent
+ * relation.
+ */
+ if (newrelpersistence == RELPERSISTENCE_UNLOGGED)
+ reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED;
+ else if (newrelpersistence == RELPERSISTENCE_PERMANENT)
+ reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT;
- reindex_relation(NULL, OIDOldHeap, reindex_flags, &reindex_params);
+ /* Report that we are now reindexing relations */
+ pgstat_progress_update_param(PROGRESS_CLUSTER_PHASE,
+ PROGRESS_CLUSTER_PHASE_REBUILD_INDEX);
+
+ reindex_relation(NULL, OIDOldHeap, reindex_flags, &reindex_params);
+ }
/* Report that we are now doing clean up */
pgstat_progress_update_param(PROGRESS_CLUSTER_PHASE,
@@ -1747,3 +2259,1886 @@ cluster_is_permitted_for_relation(Oid relid, Oid userid)
get_rel_name(relid))));
return false;
}
+
+#define REPL_PLUGIN_NAME "pgoutput_cluster"
+
+/*
+ * Each relation being processed by CLUSTER CONCURRENTLY must be in the
+ * clusteredRels hashtable.
+ */
+typedef struct ClusteredRel
+{
+ Oid relid;
+ Oid dbid;
+} ClusteredRel;
+
+static HTAB *ClusteredRelsHash = NULL;
+
+/* Maximum number of entries in the hashtable. */
+static int maxClusteredRels = 0;
+
+Size
+ClusterShmemSize(void)
+{
+ /*
+ * A replication slot is needed for the processing, so use this GUC to
+ * allocate memory for the hashtable. Reserve also space for TOAST
+ * relations.
+ */
+ maxClusteredRels = max_replication_slots * 2;
+
+ return hash_estimate_size(maxClusteredRels, sizeof(ClusteredRel));
+}
+
+void
+ClusterShmemInit(void)
+{
+ HASHCTL info;
+
+ info.keysize = sizeof(ClusteredRel);
+ info.entrysize = info.keysize;
+
+ ClusteredRelsHash = ShmemInitHash("Clustered Relations",
+ maxClusteredRels,
+ maxClusteredRels,
+ &info,
+ HASH_ELEM | HASH_BLOBS);
+}
+
+/*
+ * Perform a preliminary check whether CLUSTER / VACUUM FULL CONCURRENTLY is
+ * possible. Note that here we only check things that should not change if we
+ * release the relation lock temporarily. The information that can change due
+ * to unlocking is checked in get_catalog_state().
+ */
+static void
+check_concurrent_cluster_requirements(Relation rel, bool isTopLevel,
+ bool isCluster)
+{
+ const char *stmt;
+
+ if (isCluster)
+ stmt = "CLUSTER (CONCURRENTLY)";
+ else
+ stmt = "VACUUM (FULL, CONCURRENTLY)";
+
+ /*
+ * Make sure we have no XID assigned, otherwise call of
+ * setup_logical_decoding() can cause a deadlock.
+ */
+ PreventInTransactionBlock(isTopLevel, stmt);
+
+ CheckSlotPermissions();
+
+ /*
+ * Use an existing function to check if we can use logical
+ * decoding. However note that RecoveryInProgress() should already have
+ * caused error, as it does for the non-concurrent VACUUM FULL / CLUSTER.
+ */
+ CheckLogicalDecodingRequirements();
+
+ /* See ClusterShmemSize() */
+ if (max_replication_slots < 2)
+ ereport(ERROR,
+ errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
+ (errmsg("%s requires \"max_replication_slots\" to be at least 2",
+ stmt)));
+}
+
+/*
+ * Call this function before CLUSTER CONCURRENTLY starts to setup logical
+ * decoding. It makes sure that other users of the table put enough
+ * information into WAL.
+ *
+ * The point is that on various places we expect that the table we're
+ * processing is treated like a system catalog. For example, we need to be
+ * able to scan it using a "historic snapshot" anytime during the processing
+ * (as opposed to scanning only at the start point of the decoding, logical
+ * replication does during initial table synchronization), in order to apply
+ * concurrent UPDATE / DELETE commands.
+ *
+ * Since we need to close and reopen the relation here, the 'rel_p' and
+ * 'index_p' arguments are in/out.
+ *
+ * 'enter_p' receives a bool value telling whether relation OID was entered
+ * into the hashtable or not.
+ */
+static void
+begin_concurrent_cluster(Relation *rel_p, Relation *index_p,
+ bool *entered_p)
+{
+ Relation rel = *rel_p;
+ Oid relid, toastrelid;
+ ClusteredRel key, *entry;
+ bool found;
+ RelReopenInfo rri[2];
+ int nrel;
+ static bool before_shmem_exit_callback_setup = false;
+
+ relid = RelationGetRelid(rel);
+
+ /*
+ * Make sure that we do not leave an entry in ClusteredRelsHash if exiting
+ * due to FATAL.
+ */
+ if (!before_shmem_exit_callback_setup)
+ {
+ before_shmem_exit(cluster_before_shmem_exit_callback, 0);
+ before_shmem_exit_callback_setup = true;
+ }
+
+ memset(&key, 0, sizeof(key));
+ key.relid = relid;
+ key.dbid = MyDatabaseId;
+
+ *entered_p = false;
+ LWLockAcquire(ClusteredRelsLock, LW_EXCLUSIVE);
+ entry = (ClusteredRel *)
+ hash_search(ClusteredRelsHash, &key, HASH_ENTER_NULL, &found);
+ if (found)
+ {
+ /*
+ * Since CLUSTER CONCURRENTLY takes ShareRowExclusiveLock, a conflict
+ * should occur much earlier. However that lock may be released
+ * temporarily, see below. Anyway, we should complain whatever the
+ * reason of the conflict might be.
+ */
+ ereport(ERROR,
+ (errmsg(CLUSTER_IN_PROGRESS_MESSAGE,
+ RelationGetRelationName(rel))));
+ }
+ if (entry == NULL)
+ ereport(ERROR,
+ (errmsg("too many requests for CLUSTER CONCURRENTLY at a time")),
+ (errhint("Please consider increasing the \"max_replication_slots\" configuration parameter.")));
+
+ /*
+ * Even if the insertion of TOAST relid should fail below, the caller has
+ * to do cleanup.
+ */
+ *entered_p = true;
+
+ /*
+ * Enable the callback to remove the entry in case of exit. We should not
+ * do this earlier, otherwise an attempt to insert already existing entry
+ * could make us remove that entry (inserted by another backend) during
+ * ERROR handling.
+ */
+ Assert(!OidIsValid(clustered_rel));
+ clustered_rel = relid;
+
+ /*
+ * TOAST relation is not accessed using historic snapshot, but we enter it
+ * here to protect it from being VACUUMed by another backend. (Lock does
+ * not help in the CONCURRENT case because cannot hold it continuously
+ * till the end of the transaction.) See the comments on locking TOAST
+ * relation in copy_table_data().
+ */
+ toastrelid = rel->rd_rel->reltoastrelid;
+ if (OidIsValid(toastrelid))
+ {
+ key.relid = toastrelid;
+ entry = (ClusteredRel *)
+ hash_search(ClusteredRelsHash, &key, HASH_ENTER_NULL, &found);
+ if (found)
+ /*
+ * If we could enter the main fork the TOAST should succeed
+ * too. Nevertheless, check.
+ */
+ ereport(ERROR,
+ (errmsg("TOAST relation of \"%s\" is already being processed by CLUSTER CONCURRENTLY",
+ RelationGetRelationName(rel))));
+ if (entry == NULL)
+ ereport(ERROR,
+ (errmsg("too many requests for CLUSTER CONCURRENT at a time")),
+ (errhint("Please consider increasing the \"max_replication_slots\" configuration parameter.")));
+
+ Assert(!OidIsValid(clustered_rel_toast));
+ clustered_rel_toast = toastrelid;
+ }
+ LWLockRelease(ClusteredRelsLock);
+
+ /*
+ * Make sure that other backends are aware of the new hash entry.
+ *
+ * Besides sending the invalidation message, we need to force re-opening
+ * of the relation, which includes the actual invalidation (and thus
+ * checking of our hashtable on the next access).
+ */
+ CacheInvalidateRelcacheImmediate(rel);
+ /*
+ * Since the hashtable only needs to be checked by write transactions,
+ * lock the relation in a mode that conflicts with any DML command. (The
+ * reading transactions are supposed to close the relation before opening
+ * it with higher lock.) Once we have the relation (and its index) locked,
+ * we unlock it immediately and then re-lock using the original mode.
+ */
+ nrel = 0;
+ init_rel_reopen_info(&rri[nrel++], rel_p, InvalidOid,
+ ShareUpdateExclusiveLock, ShareLock);
+ if (index_p)
+ {
+ /*
+ * Another transaction might want to open both the relation and the
+ * index. If it already has the relation lock and is waiting for the
+ * index lock, we should release the index lock, otherwise our request
+ * for ShareLock on the relation can end up in a deadlock.
+ */
+ init_rel_reopen_info(&rri[nrel++], index_p, InvalidOid,
+ ShareUpdateExclusiveLock, ShareLock);
+ }
+ unlock_and_close_relations(rri, nrel);
+ /*
+ * XXX It's not strictly necessary to lock the index here, but it's
+ * probably not worth teaching the "reopen API" about this special case.
+ */
+ reopen_relations(rri, nrel);
+
+ /* Switch back to the original lock. */
+ nrel = 0;
+ init_rel_reopen_info(&rri[nrel++], rel_p, InvalidOid,
+ ShareLock, ShareUpdateExclusiveLock);
+ if (index_p)
+ init_rel_reopen_info(&rri[nrel++], index_p, InvalidOid,
+ ShareLock, ShareUpdateExclusiveLock);
+ unlock_and_close_relations(rri, nrel);
+ reopen_relations(rri, nrel);
+ /* Make sure the reopened relcache entry is used, not the old one. */
+ rel = *rel_p;
+
+ /* Avoid logical decoding of other relations by this backend. */
+ clustered_rel_locator = rel->rd_locator;
+ if (OidIsValid(toastrelid))
+ {
+ Relation toastrel;
+
+ /* Avoid logical decoding of other TOAST relations. */
+ toastrel = table_open(toastrelid, AccessShareLock);
+ clustered_rel_toast_locator = toastrel->rd_locator;
+ table_close(toastrel, AccessShareLock);
+ }
+}
+
+/*
+ * Call this when done with CLUSTER CONCURRENTLY.
+ *
+ * 'error' tells whether the function is being called in order to handle
+ * error.
+ */
+static void
+end_concurrent_cluster(bool error)
+{
+ ClusteredRel key;
+ ClusteredRel *entry = NULL, *entry_toast = NULL;
+ Oid relid = clustered_rel;
+ Oid toastrelid = clustered_rel_toast;
+
+ /* Remove the relation from the hash if we managed to insert one. */
+ if (OidIsValid(clustered_rel))
+ {
+ memset(&key, 0, sizeof(key));
+ key.relid = clustered_rel;
+ key.dbid = MyDatabaseId;
+ LWLockAcquire(ClusteredRelsLock, LW_EXCLUSIVE);
+ entry = hash_search(ClusteredRelsHash, &key, HASH_REMOVE, NULL);
+
+ /*
+ * By clearing this variable we also disable
+ * cluster_before_shmem_exit_callback().
+ */
+ clustered_rel = InvalidOid;
+ }
+
+ /* Remove the TOAST relation if there is one. */
+ if (OidIsValid(clustered_rel_toast))
+ {
+ key.relid = clustered_rel_toast;
+ entry_toast = hash_search(ClusteredRelsHash, &key, HASH_REMOVE,
+ NULL);
+
+ clustered_rel_toast = InvalidOid;
+ }
+ LWLockRelease(ClusteredRelsLock);
+
+ /* Restore normal function of logical decoding. */
+ clustered_rel_locator.relNumber = InvalidOid;
+ clustered_rel_toast_locator.relNumber = InvalidOid;
+
+ /*
+ * On normal completion (!error), we should not really fail to remove the
+ * entry. But if it wasn't there for any reason, raise ERROR to make sure
+ * the transaction is aborted: if other transactions, while changing the
+ * contents of the relation, didn't know that CLUSTER CONCURRENTLY was in
+ * progress, they could have missed to WAL enough information, and thus we
+ * could have produced an inconsistent table contents.
+ *
+ * On the other hand, if we are already handling an error, there's no
+ * reason to worry about inconsistent contents of the new storage because
+ * the transaction is going to be rolled back anyway. Furthermore, by
+ * raising ERROR here we'd shadow the original error.
+ */
+ if (!error)
+ {
+ char *relname;
+
+ if (OidIsValid(relid) && entry == NULL)
+ {
+ relname = get_rel_name(relid);
+ if (!relname)
+ ereport(ERROR,
+ (errmsg("cache lookup failed for relation %u",
+ relid)));
+
+ ereport(ERROR,
+ (errmsg("relation \"%s\" not found among clustered relations",
+ relname)));
+ }
+
+ /*
+ * Likewise, the TOAST relation should not have disappeared.
+ */
+ if (OidIsValid(toastrelid) && entry_toast == NULL)
+ {
+ relname = get_rel_name(key.relid);
+ if (!relname)
+ ereport(ERROR,
+ (errmsg("cache lookup failed for relation %u",
+ key.relid)));
+
+ ereport(ERROR,
+ (errmsg("relation \"%s\" not found among clustered relations",
+ relname)));
+ }
+ }
+
+ /*
+ * Note: unlike begin_concurrent_cluster(), here we do not lock/unlock the
+ * relation: 1) On normal completion, the caller is already holding
+ * AccessExclusiveLock (till the end of the transaction), 2) on ERROR /
+ * FATAL, we try to do the cleanup asap, but the worst case is that other
+ * backends will write unnecessary information to WAL until they close the
+ * relation.
+ */
+}
+
+/*
+ * A wrapper to call end_concurrent_cluster() as a before_shmem_exit callback.
+ */
+static void
+cluster_before_shmem_exit_callback(int code, Datum arg)
+{
+ if (OidIsValid(clustered_rel) || OidIsValid(clustered_rel_toast))
+ end_concurrent_cluster(true);
+}
+
+/*
+ * Check if relation is currently being processed by CLUSTER CONCURRENTLY.
+ */
+bool
+is_concurrent_cluster_in_progress(Oid relid)
+{
+ ClusteredRel key, *entry;
+
+ memset(&key, 0, sizeof(key));
+ key.relid = relid;
+ key.dbid = MyDatabaseId;
+
+ LWLockAcquire(ClusteredRelsLock, LW_SHARED);
+ entry = (ClusteredRel *)
+ hash_search(ClusteredRelsHash, &key, HASH_FIND, NULL);
+ LWLockRelease(ClusteredRelsLock);
+
+ return entry != NULL;
+}
+
+/*
+ * Check if VACUUM FULL / CLUSTER CONCURRENTLY is already running for given
+ * relation, and if so, raise ERROR. The problem is that cluster_rel() needs
+ * to release its lock on the relation temporarily at some point, so our lock
+ * alone does not help. Commands that might break what cluster_rel() is doing
+ * should call this function first.
+ *
+ * Return without checking if lockmode allows for race conditions which would
+ * make the result meaningless. In that case, cluster_rel() itself should
+ * throw ERROR if the relation was changed by us in an incompatible
+ * way. However, if it managed to do most of its work by then, a lot of CPU
+ * time might be wasted.
+ */
+void
+check_for_concurrent_cluster(Oid relid, LOCKMODE lockmode)
+{
+ /*
+ * If the caller does not have a lock that conflicts with
+ * ShareUpdateExclusiveLock, the check makes little sense because the
+ * VACUUM FULL / CLUSTER CONCURRENTLY can start anytime after the check.
+ */
+ if (lockmode < ShareUpdateExclusiveLock)
+ return;
+
+ if (is_concurrent_cluster_in_progress(relid))
+ ereport(ERROR,
+ (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
+ errmsg(CLUSTER_IN_PROGRESS_MESSAGE,
+ get_rel_name(relid))));
+
+}
+
+/*
+ * Check if relation is eligible for CLUSTER CONCURRENTLY and retrieve the
+ * catalog state to be passed later to check_catalog_changes.
+ *
+ * Caller is supposed to hold (at least) ShareUpdateExclusiveLock on the
+ * relation.
+ */
+static CatalogState *
+get_catalog_state(Relation rel, bool is_vacuum)
+{
+ CatalogState *result = palloc_object(CatalogState);
+ List *ind_oids;
+ ListCell *lc;
+ int ninds, i;
+ char relpersistence = rel->rd_rel->relpersistence;
+ char replident = rel->rd_rel->relreplident;
+ Oid ident_idx = RelationGetReplicaIndex(rel);
+ TupleDesc td_src = RelationGetDescr(rel);
+
+ /*
+ * While gathering the catalog information, check if there is a reason not
+ * to proceed.
+ *
+ * This function was already called, but the relation was unlocked since
+ * (see begin_concurrent_cluster()). check_catalog_changes() should catch
+ * any "disruptive" changes in the future.
+ */
+ check_relation_is_clusterable_concurrently(rel, is_vacuum);
+
+ /* No index should be dropped while we are checking it. */
+ Assert(CheckRelationLockedByMe(rel, ShareUpdateExclusiveLock, true));
+
+ ind_oids = RelationGetIndexList(rel);
+ result->ninds = ninds = list_length(ind_oids);
+ result->ind_oids = palloc_array(Oid, ninds);
+ result->ind_tupdescs = palloc_array(TupleDesc, ninds);
+ i = 0;
+ foreach(lc, ind_oids)
+ {
+ Oid ind_oid = lfirst_oid(lc);
+ Relation index;
+ TupleDesc td_ind_src, td_ind_dst;
+
+ /*
+ * Weaker lock should be o.k. for the index, but this one should not
+ * break anything either.
+ */
+ index = index_open(ind_oid, ShareUpdateExclusiveLock);
+
+ result->ind_oids[i] = RelationGetRelid(index);
+ td_ind_src = RelationGetDescr(index);
+ td_ind_dst = palloc(TupleDescSize(td_ind_src));
+ TupleDescCopy(td_ind_dst, td_ind_src);
+ result->ind_tupdescs[i] = td_ind_dst;
+ i++;
+
+ index_close(index, ShareUpdateExclusiveLock);
+ }
+
+ /* Fill-in the relation info. */
+ result->tupdesc = palloc(TupleDescSize(td_src));
+ TupleDescCopy(result->tupdesc, td_src);
+ result->relpersistence = relpersistence;
+ result->replident = replident;
+ result->replidindex = ident_idx;
+
+ return result;
+}
+
+static void
+free_catalog_state(CatalogState *state)
+{
+ /* We are only interested in indexes. */
+ if (state->ninds == 0)
+ return;
+
+ for (int i = 0; i < state->ninds; i++)
+ FreeTupleDesc(state->ind_tupdescs[i]);
+
+ FreeTupleDesc(state->tupdesc);
+ pfree(state->ind_oids);
+ pfree(state->ind_tupdescs);
+ pfree(state);
+}
+
+/*
+ * Raise ERROR if 'rel' changed in a way that does not allow further
+ * processing of CLUSTER CONCURRENTLY.
+ *
+ * Besides the relation's tuple descriptor, it's important to check indexes:
+ * concurrent change of index definition (can it happen in other way than
+ * dropping and re-creating the index, accidentally with the same OID?) can be
+ * a problem because we may already have the new index built. If an index was
+ * created or dropped concurrently, we'd fail to swap the index storage. In
+ * any case, we prefer to check the indexes early to get an explicit error
+ * message about the mismatch. Furthermore, the earlier we detect the change,
+ * the fewer CPU cycles we waste.
+ *
+ * Note that we do not check constraints because the transaction which changed
+ * them must have ensured that the existing tuples satisfy the new
+ * constraints. If any DML commands were necessary for that, we will simply
+ * decode them from WAL and apply them to the new storage.
+ *
+ * Caller is supposed to hold (at least) ShareUpdateExclusiveLock on the
+ * relation.
+ */
+static void
+check_catalog_changes(Relation rel, CatalogState *cat_state)
+{
+ Oid reltoastrelid = rel->rd_rel->reltoastrelid;
+ List *ind_oids;
+ ListCell *lc;
+ LOCKMODE lockmode;
+ Oid ident_idx;
+ TupleDesc td, td_cp;
+
+ /* First, check the relation info. */
+
+ /* TOAST is not easy to change, but check. */
+ if (reltoastrelid != clustered_rel_toast)
+ ereport(ERROR,
+ errmsg("TOAST relation of relation \"%s\" changed by another transaction",
+ RelationGetRelationName(rel)));
+
+ /*
+ * Likewise, check_for_concurrent_cluster() should prevent others from
+ * changing the relation file concurrently, but it's our responsibility to
+ * avoid data loss. (The original locators are stored outside cat_state,
+ * but the check belongs to this function.)
+ */
+ if (!RelFileLocatorEquals(rel->rd_locator, clustered_rel_locator))
+ ereport(ERROR,
+ (errmsg("file of relation \"%s\" changed by another transaction",
+ RelationGetRelationName(rel))));
+ if (OidIsValid(reltoastrelid))
+ {
+ Relation toastrel;
+
+ toastrel = table_open(reltoastrelid, AccessShareLock);
+ if (!RelFileLocatorEquals(toastrel->rd_locator,
+ clustered_rel_toast_locator))
+ ereport(ERROR,
+ (errmsg("file of relation \"%s\" changed by another transaction",
+ RelationGetRelationName(toastrel))));
+ table_close(toastrel, AccessShareLock);
+ }
+
+ if (rel->rd_rel->relpersistence != cat_state->relpersistence)
+ ereport(ERROR,
+ errmsg("persistence of relation \"%s\" changed by another transaction",
+ RelationGetRelationName(rel)));
+
+ if (cat_state->replident != rel->rd_rel->relreplident)
+ ereport(ERROR,
+ errmsg("replica identity of relation \"%s\" changed by another transaction",
+ RelationGetRelationName(rel)));
+
+ ident_idx = RelationGetReplicaIndex(rel);
+ if (ident_idx == InvalidOid && rel->rd_pkindex != InvalidOid)
+ ident_idx = rel->rd_pkindex;
+ if (cat_state->replidindex != ident_idx)
+ ereport(ERROR,
+ errmsg("identity index of relation \"%s\" changed by another transaction",
+ RelationGetRelationName(rel)));
+
+ /*
+ * As cat_state contains a copy (which has the constraint info cleared),
+ * create a temporary copy for the comparison.
+ */
+ td = RelationGetDescr(rel);
+ td_cp = palloc(TupleDescSize(td));
+ TupleDescCopy(td_cp, td);
+ if (!equalTupleDescs(cat_state->tupdesc, td_cp))
+ ereport(ERROR,
+ errmsg("definition of relation \"%s\" changed by another transaction",
+ RelationGetRelationName(rel)));
+ FreeTupleDesc(td_cp);
+
+ /* Now we are only interested in indexes. */
+ if (cat_state->ninds == 0)
+ return;
+
+ /* No index should be dropped while we are checking the relation. */
+ lockmode = ShareUpdateExclusiveLock;
+ Assert(CheckRelationLockedByMe(rel, lockmode, true));
+
+ ind_oids = RelationGetIndexList(rel);
+ if (list_length(ind_oids) != cat_state->ninds)
+ goto failed_index;
+
+ foreach(lc, ind_oids)
+ {
+ Oid ind_oid = lfirst_oid(lc);
+ int i;
+ TupleDesc tupdesc;
+ Relation index;
+
+ /* Find the index in cat_state. */
+ for (i = 0; i < cat_state->ninds; i++)
+ {
+ if (cat_state->ind_oids[i] == ind_oid)
+ break;
+ }
+ /*
+ * OID not found, i.e. the index was replaced by another one. XXX
+ * Should we yet try to find if an index having the desired tuple
+ * descriptor exists? Or should we always look for the tuple
+ * descriptor and not use OIDs at all?
+ */
+ if (i == cat_state->ninds)
+ goto failed_index;
+
+ /* Check the tuple descriptor. */
+ index = try_index_open(ind_oid, lockmode);
+ if (index == NULL)
+ goto failed_index;
+ tupdesc = RelationGetDescr(index);
+ if (!equalTupleDescs(cat_state->ind_tupdescs[i], tupdesc))
+ goto failed_index;
+ index_close(index, lockmode);
+ }
+
+ return;
+
+failed_index:
+ ereport(ERROR,
+ (errmsg("index(es) of relation \"%s\" changed by another transaction",
+ RelationGetRelationName(rel))));
+}
+
+/*
+ * This function is much like pg_create_logical_replication_slot() except that
+ * the new slot is neither released (if anyone else could read changes from
+ * our slot, we could miss changes other backends do while we copy the
+ * existing data into temporary table), nor persisted (it's easier to handle
+ * crash by restarting all the work from scratch).
+ *
+ * XXX Even though CreateInitDecodingContext() does not set state to
+ * RS_PERSISTENT, it does write the slot to disk. We rely on
+ * RestoreSlotFromDisk() to delete ephemeral slots during startup. (Both ERROR
+ * and FATAL should lead to cleanup even before the cluster goes down.)
+ */
+static LogicalDecodingContext *
+setup_logical_decoding(Oid relid, const char *slotname, TupleDesc tupdesc)
+{
+ LogicalDecodingContext *ctx;
+ ClusterDecodingState *dstate;
+
+ /* RS_TEMPORARY so that the slot gets cleaned up on ERROR. */
+ ReplicationSlotCreate(slotname, true, RS_TEMPORARY, false, false, false);
+
+ /*
+ * Neither prepare_write nor do_write callback nor update_progress is
+ * useful for us.
+ *
+ * Regarding the value of need_full_snapshot, we pass false because the
+ * table we are processing is present in ClusteredRelsHash and therefore,
+ * regarding logical decoding, treated like a catalog.
+ */
+ ctx = CreateInitDecodingContext(REPL_PLUGIN_NAME,
+ NIL,
+ false,
+ InvalidXLogRecPtr,
+ XL_ROUTINE(.page_read = read_local_xlog_page,
+ .segment_open = wal_segment_open,
+ .segment_close = wal_segment_close),
+ NULL, NULL, NULL);
+
+ /*
+ * We don't have control on setting fast_forward, so at least check it.
+ */
+ Assert(!ctx->fast_forward);
+
+ DecodingContextFindStartpoint(ctx);
+
+ /* Some WAL records should have been read. */
+ Assert(ctx->reader->EndRecPtr != InvalidXLogRecPtr);
+
+ XLByteToSeg(ctx->reader->EndRecPtr, cluster_current_segment,
+ wal_segment_size);
+
+ /*
+ * Setup structures to store decoded changes.
+ */
+ dstate = palloc0(sizeof(ClusterDecodingState));
+ dstate->relid = relid;
+ dstate->tstore = tuplestore_begin_heap(false, false,
+ maintenance_work_mem);
+ dstate->tupdesc = tupdesc;
+
+ /* Initialize the descriptor to store the changes ... */
+ dstate->tupdesc_change = CreateTemplateTupleDesc(1);
+
+ TupleDescInitEntry(dstate->tupdesc_change, 1, NULL, BYTEAOID, -1, 0);
+ /* ... as well as the corresponding slot. */
+ dstate->tsslot = MakeSingleTupleTableSlot(dstate->tupdesc_change,
+ &TTSOpsMinimalTuple);
+
+ dstate->resowner = ResourceOwnerCreate(CurrentResourceOwner,
+ "logical decoding");
+
+ ctx->output_writer_private = dstate;
+ return ctx;
+}
+
+/*
+ * Retrieve tuple from ConcurrentChange structure.
+ *
+ * The input data starts with the structure but it might not be appropriately
+ * aligned.
+ */
+static HeapTuple
+get_changed_tuple(char *change)
+{
+ HeapTupleData tup_data;
+ HeapTuple result;
+ char *src;
+
+ /*
+ * Ensure alignment before accessing the fields. (This is why we can't use
+ * heap_copytuple() instead of this function.)
+ */
+ src = change + offsetof(ConcurrentChange, tup_data);
+ memcpy(&tup_data, src, sizeof(HeapTupleData));
+
+ result = (HeapTuple) palloc(HEAPTUPLESIZE + tup_data.t_len);
+ memcpy(result, &tup_data, sizeof(HeapTupleData));
+ result->t_data = (HeapTupleHeader) ((char *) result + HEAPTUPLESIZE);
+ src = change + SizeOfConcurrentChange;
+ memcpy(result->t_data, src, result->t_len);
+
+ return result;
+}
+
+/*
+ * Decode logical changes from the WAL sequence up to end_of_wal.
+ */
+void
+cluster_decode_concurrent_changes(LogicalDecodingContext *ctx,
+ XLogRecPtr end_of_wal)
+{
+ ClusterDecodingState *dstate;
+ ResourceOwner resowner_old;
+ PgBackendProgress progress;
+
+ /*
+ * Invalidate the "present" cache before moving to "(recent) history".
+ */
+ InvalidateSystemCaches();
+
+ dstate = (ClusterDecodingState *) ctx->output_writer_private;
+ resowner_old = CurrentResourceOwner;
+ CurrentResourceOwner = dstate->resowner;
+
+ /*
+ * reorderbuffer.c uses internal subtransaction, whose abort ends the
+ * command progress reporting. Save the status here so we can restore when
+ * done with the decoding.
+ */
+ memcpy(&progress, &MyBEEntry->st_progress, sizeof(PgBackendProgress));
+
+ PG_TRY();
+ {
+ while (ctx->reader->EndRecPtr < end_of_wal)
+ {
+ XLogRecord *record;
+ XLogSegNo segno_new;
+ char *errm = NULL;
+ XLogRecPtr end_lsn;
+
+ record = XLogReadRecord(ctx->reader, &errm);
+ if (errm)
+ elog(ERROR, "%s", errm);
+
+ if (record != NULL)
+ LogicalDecodingProcessRecord(ctx, ctx->reader);
+
+ /*
+ * If WAL segment boundary has been crossed, inform the decoding
+ * system that the catalog_xmin can advance. (We can confirm more
+ * often, but a filling a single WAL segment should not take much
+ * time.)
+ */
+ end_lsn = ctx->reader->EndRecPtr;
+ XLByteToSeg(end_lsn, segno_new, wal_segment_size);
+ if (segno_new != cluster_current_segment)
+ {
+ LogicalConfirmReceivedLocation(end_lsn);
+ elog(DEBUG1, "cluster: confirmed receive location %X/%X",
+ (uint32) (end_lsn >> 32), (uint32) end_lsn);
+ cluster_current_segment = segno_new;
+ }
+
+ CHECK_FOR_INTERRUPTS();
+ }
+ InvalidateSystemCaches();
+ CurrentResourceOwner = resowner_old;
+ }
+ PG_CATCH();
+ {
+ InvalidateSystemCaches();
+ CurrentResourceOwner = resowner_old;
+ PG_RE_THROW();
+ }
+ PG_END_TRY();
+
+ /* Restore the progress reporting status. */
+ pgstat_progress_restore_state(&progress);
+}
+
+/*
+ * Apply changes that happened during the initial load.
+ *
+ * Scan key is passed by caller, so it does not have to be constructed
+ * multiple times. Key entries have all fields initialized, except for
+ * sk_argument.
+ */
+static void
+apply_concurrent_changes(ClusterDecodingState *dstate, Relation rel,
+ ScanKey key, int nkeys, IndexInsertState *iistate)
+{
+ TupleTableSlot *index_slot, *ident_slot;
+ HeapTuple tup_old = NULL;
+
+ if (dstate->nchanges == 0)
+ return;
+
+ /* TupleTableSlot is needed to pass the tuple to ExecInsertIndexTuples(). */
+ index_slot = MakeSingleTupleTableSlot(dstate->tupdesc, &TTSOpsHeapTuple);
+ iistate->econtext->ecxt_scantuple = index_slot;
+
+ /* A slot to fetch tuples from identity index. */
+ ident_slot = table_slot_create(rel, NULL);
+
+ while (tuplestore_gettupleslot(dstate->tstore, true, false,
+ dstate->tsslot))
+ {
+ bool shouldFree;
+ HeapTuple tup_change,
+ tup,
+ tup_exist;
+ char *change_raw, *src;
+ ConcurrentChange change;
+ bool isnull[1];
+ Datum values[1];
+
+ CHECK_FOR_INTERRUPTS();
+
+ /* Get the change from the single-column tuple. */
+ tup_change = ExecFetchSlotHeapTuple(dstate->tsslot, false, &shouldFree);
+ heap_deform_tuple(tup_change, dstate->tupdesc_change, values, isnull);
+ Assert(!isnull[0]);
+
+ /* Make sure we access aligned data. */
+ change_raw = (char *) DatumGetByteaP(values[0]);
+ src = (char *) VARDATA(change_raw);
+ memcpy(&change, src, SizeOfConcurrentChange);
+
+ /* TRUNCATE change contains no tuple, so process it separately. */
+ if (change.kind == CHANGE_TRUNCATE)
+ {
+ /*
+ * All the things that ExecuteTruncateGuts() does (such as firing
+ * triggers or handling the DROP_CASCADE behavior) should have
+ * taken place on the source relation. Thus we only do the actual
+ * truncation of the new relation (and its indexes).
+ */
+ heap_truncate_one_rel(rel);
+
+ pfree(tup_change);
+ continue;
+ }
+
+ /*
+ * Extract the tuple from the change. The tuple is copied here because
+ * it might be assigned to 'tup_old', in which case it needs to
+ * survive into the next iteration.
+ */
+ tup = get_changed_tuple(src);
+
+ if (change.kind == CHANGE_UPDATE_OLD)
+ {
+ Assert(tup_old == NULL);
+ tup_old = tup;
+ }
+ else if (change.kind == CHANGE_INSERT)
+ {
+ Assert(tup_old == NULL);
+
+ apply_concurrent_insert(rel, &change, tup, iistate, index_slot);
+
+ pfree(tup);
+ }
+ else if (change.kind == CHANGE_UPDATE_NEW ||
+ change.kind == CHANGE_DELETE)
+ {
+ IndexScanDesc ind_scan = NULL;
+ HeapTuple tup_key;
+
+ if (change.kind == CHANGE_UPDATE_NEW)
+ {
+ tup_key = tup_old != NULL ? tup_old : tup;
+ }
+ else
+ {
+ Assert(tup_old == NULL);
+ tup_key = tup;
+ }
+
+ /*
+ * Find the tuple to be updated or deleted.
+ */
+ tup_exist = find_target_tuple(rel, key, nkeys, tup_key,
+ iistate, ident_slot, &ind_scan);
+ if (tup_exist == NULL)
+ elog(ERROR, "Failed to find target tuple");
+
+ if (change.kind == CHANGE_UPDATE_NEW)
+ apply_concurrent_update(rel, tup, tup_exist, &change, iistate,
+ index_slot);
+ else
+ apply_concurrent_delete(rel, tup_exist, &change);
+
+ if (tup_old != NULL)
+ {
+ pfree(tup_old);
+ tup_old = NULL;
+ }
+
+ pfree(tup);
+ index_endscan(ind_scan);
+ }
+ else
+ elog(ERROR, "Unrecognized kind of change: %d", change.kind);
+
+ /* If there's any change, make it visible to the next iteration. */
+ if (change.kind != CHANGE_UPDATE_OLD)
+ {
+ CommandCounterIncrement();
+ UpdateActiveSnapshotCommandId();
+ }
+
+ /* TTSOpsMinimalTuple has .get_heap_tuple==NULL. */
+ Assert(shouldFree);
+ pfree(tup_change);
+ }
+
+ tuplestore_clear(dstate->tstore);
+ dstate->nchanges = 0;
+
+ /* Cleanup. */
+ ExecDropSingleTupleTableSlot(index_slot);
+ ExecDropSingleTupleTableSlot(ident_slot);
+}
+
+static void
+apply_concurrent_insert(Relation rel, ConcurrentChange *change, HeapTuple tup,
+ IndexInsertState *iistate, TupleTableSlot *index_slot)
+{
+ List *recheck;
+
+
+ heap_insert(rel, tup, GetCurrentCommandId(true), HEAP_INSERT_NO_LOGICAL, NULL);
+
+ /*
+ * Update indexes.
+ *
+ * In case functions in the index need the active snapshot and caller
+ * hasn't set one.
+ */
+ ExecStoreHeapTuple(tup, index_slot, false);
+ recheck = ExecInsertIndexTuples(iistate->rri,
+ index_slot,
+ iistate->estate,
+ false, /* update */
+ false, /* noDupErr */
+ NULL, /* specConflict */
+ NIL, /* arbiterIndexes */
+ false /* onlySummarizing */
+ );
+
+ /*
+ * If recheck is required, it must have been preformed on the source
+ * relation by now. (All the logical changes we process here are already
+ * committed.)
+ */
+ list_free(recheck);
+
+ pgstat_progress_incr_param(PROGRESS_CLUSTER_HEAP_TUPLES_INSERTED, 1);
+}
+
+static void
+apply_concurrent_update(Relation rel, HeapTuple tup, HeapTuple tup_target,
+ ConcurrentChange *change, IndexInsertState *iistate,
+ TupleTableSlot *index_slot)
+{
+ List *recheck;
+ TU_UpdateIndexes update_indexes;
+
+ /*
+ * Write the new tuple into the new heap. ('tup' gets the TID assigned
+ * here.)
+ */
+ simple_heap_update(rel, &tup_target->t_self, tup, &update_indexes);
+
+ ExecStoreHeapTuple(tup, index_slot, false);
+
+ if (update_indexes != TU_None)
+ {
+ recheck = ExecInsertIndexTuples(iistate->rri,
+ index_slot,
+ iistate->estate,
+ true, /* update */
+ false, /* noDupErr */
+ NULL, /* specConflict */
+ NIL, /* arbiterIndexes */
+ /* onlySummarizing */
+ update_indexes == TU_Summarizing);
+ list_free(recheck);
+ }
+
+ pgstat_progress_incr_param(PROGRESS_CLUSTER_HEAP_TUPLES_UPDATED, 1);
+}
+
+static void
+apply_concurrent_delete(Relation rel, HeapTuple tup_target,
+ ConcurrentChange *change)
+{
+ simple_heap_delete(rel, &tup_target->t_self);
+
+ pgstat_progress_incr_param(PROGRESS_CLUSTER_HEAP_TUPLES_DELETED, 1);
+}
+
+/*
+ * Find the tuple to be updated or deleted.
+ *
+ * 'key' is a pre-initialized scan key, into which the function will put the
+ * key values.
+ *
+ * 'tup_key' is a tuple containing the key values for the scan.
+ *
+ * On exit,'*scan_p' contains the scan descriptor used. The caller must close
+ * it when he no longer needs the tuple returned.
+ */
+static HeapTuple
+find_target_tuple(Relation rel, ScanKey key, int nkeys, HeapTuple tup_key,
+ IndexInsertState *iistate,
+ TupleTableSlot *ident_slot, IndexScanDesc *scan_p)
+{
+ IndexScanDesc scan;
+ Form_pg_index ident_form;
+ int2vector *ident_indkey;
+ HeapTuple result = NULL;
+
+ scan = index_beginscan(rel, iistate->ident_index, GetActiveSnapshot(),
+ nkeys, 0);
+ *scan_p = scan;
+ index_rescan(scan, key, nkeys, NULL, 0);
+
+ /* Info needed to retrieve key values from heap tuple. */
+ ident_form = iistate->ident_index->rd_index;
+ ident_indkey = &ident_form->indkey;
+
+ /* Use the incoming tuple to finalize the scan key. */
+ for (int i = 0; i < scan->numberOfKeys; i++)
+ {
+ ScanKey entry;
+ bool isnull;
+ int16 attno_heap;
+
+ entry = &scan->keyData[i];
+ attno_heap = ident_indkey->values[i];
+ entry->sk_argument = heap_getattr(tup_key,
+ attno_heap,
+ rel->rd_att,
+ &isnull);
+ Assert(!isnull);
+ }
+ if (index_getnext_slot(scan, ForwardScanDirection, ident_slot))
+ {
+ bool shouldFree;
+
+ result = ExecFetchSlotHeapTuple(ident_slot, false, &shouldFree);
+ /* TTSOpsBufferHeapTuple has .get_heap_tuple != NULL. */
+ Assert(!shouldFree);
+ }
+
+ return result;
+}
+
+/*
+ * Decode and apply concurrent changes.
+ *
+ * Pass rel_src iff its reltoastrelid is needed.
+ */
+static void
+process_concurrent_changes(LogicalDecodingContext *ctx, XLogRecPtr end_of_wal,
+ Relation rel_dst, Relation rel_src, ScanKey ident_key,
+ int ident_key_nentries, IndexInsertState *iistate)
+{
+ ClusterDecodingState *dstate;
+
+ pgstat_progress_update_param(PROGRESS_CLUSTER_PHASE,
+ PROGRESS_CLUSTER_PHASE_CATCH_UP);
+
+ dstate = (ClusterDecodingState *) ctx->output_writer_private;
+
+ cluster_decode_concurrent_changes(ctx, end_of_wal);
+
+ if (dstate->nchanges == 0)
+ return;
+
+ PG_TRY();
+ {
+ /*
+ * Make sure that TOAST values can eventually be accessed via the old
+ * relation - see comment in copy_table_data().
+ */
+ if (rel_src)
+ rel_dst->rd_toastoid = rel_src->rd_rel->reltoastrelid;
+
+ apply_concurrent_changes(dstate, rel_dst, ident_key,
+ ident_key_nentries, iistate);
+ }
+ PG_FINALLY();
+ {
+ if (rel_src)
+ rel_dst->rd_toastoid = InvalidOid;
+ }
+ PG_END_TRY();
+}
+
+static IndexInsertState *
+get_index_insert_state(Relation relation, Oid ident_index_id)
+{
+ EState *estate;
+ int i;
+ IndexInsertState *result;
+
+ result = (IndexInsertState *) palloc0(sizeof(IndexInsertState));
+ estate = CreateExecutorState();
+ result->econtext = GetPerTupleExprContext(estate);
+
+ result->rri = (ResultRelInfo *) palloc(sizeof(ResultRelInfo));
+ InitResultRelInfo(result->rri, relation, 0, 0, 0);
+ ExecOpenIndices(result->rri, false);
+
+ /*
+ * Find the relcache entry of the identity index so that we spend no extra
+ * effort to open / close it.
+ */
+ for (i = 0; i < result->rri->ri_NumIndices; i++)
+ {
+ Relation ind_rel;
+
+ ind_rel = result->rri->ri_IndexRelationDescs[i];
+ if (ind_rel->rd_id == ident_index_id)
+ result->ident_index = ind_rel;
+ }
+ if (result->ident_index == NULL)
+ elog(ERROR, "Failed to open identity index");
+
+ /* Only initialize fields needed by ExecInsertIndexTuples(). */
+ result->estate = estate;
+
+ return result;
+}
+
+/*
+ * Build scan key to process logical changes.
+ */
+static ScanKey
+build_identity_key(Oid ident_idx_oid, Relation rel_src, int *nentries)
+{
+ Relation ident_idx_rel;
+ Form_pg_index ident_idx;
+ int n,
+ i;
+ ScanKey result;
+
+ Assert(OidIsValid(ident_idx_oid));
+ ident_idx_rel = index_open(ident_idx_oid, AccessShareLock);
+ ident_idx = ident_idx_rel->rd_index;
+ n = ident_idx->indnatts;
+ result = (ScanKey) palloc(sizeof(ScanKeyData) * n);
+ for (i = 0; i < n; i++)
+ {
+ ScanKey entry;
+ int16 relattno;
+ Form_pg_attribute att;
+ Oid opfamily,
+ opcintype,
+ opno,
+ opcode;
+
+ entry = &result[i];
+ relattno = ident_idx->indkey.values[i];
+ if (relattno >= 1)
+ {
+ TupleDesc desc;
+
+ desc = rel_src->rd_att;
+ att = TupleDescAttr(desc, relattno - 1);
+ }
+ else
+ elog(ERROR, "Unexpected attribute number %d in index", relattno);
+
+ opfamily = ident_idx_rel->rd_opfamily[i];
+ opcintype = ident_idx_rel->rd_opcintype[i];
+ opno = get_opfamily_member(opfamily, opcintype, opcintype,
+ BTEqualStrategyNumber);
+
+ if (!OidIsValid(opno))
+ elog(ERROR, "Failed to find = operator for type %u", opcintype);
+
+ opcode = get_opcode(opno);
+ if (!OidIsValid(opcode))
+ elog(ERROR, "Failed to find = operator for operator %u", opno);
+
+ /* Initialize everything but argument. */
+ ScanKeyInit(entry,
+ i + 1,
+ BTEqualStrategyNumber, opcode,
+ (Datum) NULL);
+ entry->sk_collation = att->attcollation;
+ }
+ index_close(ident_idx_rel, AccessShareLock);
+
+ *nentries = n;
+ return result;
+}
+
+static void
+free_index_insert_state(IndexInsertState *iistate)
+{
+ ExecCloseIndices(iistate->rri);
+ FreeExecutorState(iistate->estate);
+ pfree(iistate->rri);
+ pfree(iistate);
+}
+
+static void
+cleanup_logical_decoding(LogicalDecodingContext *ctx)
+{
+ ClusterDecodingState *dstate;
+
+ dstate = (ClusterDecodingState *) ctx->output_writer_private;
+
+ ExecDropSingleTupleTableSlot(dstate->tsslot);
+ FreeTupleDesc(dstate->tupdesc_change);
+ FreeTupleDesc(dstate->tupdesc);
+ tuplestore_end(dstate->tstore);
+
+ FreeDecodingContext(ctx);
+}
+
+/*
+ * The final steps of rebuild_relation() for concurrent processing.
+ *
+ * On entry, NewHeap is locked in AccessExclusiveLock mode. OldHeap and its
+ * clustering index (if one is passed) are still locked in a mode that allows
+ * concurrent data changes. On exit, both tables and their indexes are closed,
+ * but locked in AccessExclusiveLock mode.
+ */
+static void
+rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
+ Relation cl_index,
+ CatalogState *cat_state,
+ LogicalDecodingContext *ctx,
+ bool swap_toast_by_content,
+ TransactionId frozenXid,
+ MultiXactId cutoffMulti)
+{
+ LOCKMODE lockmode_old PG_USED_FOR_ASSERTS_ONLY;
+ List *ind_oids_new;
+ Oid old_table_oid = RelationGetRelid(OldHeap);
+ Oid new_table_oid = RelationGetRelid(NewHeap);
+ List *ind_oids_old = RelationGetIndexList(OldHeap);
+ ListCell *lc, *lc2;
+ char relpersistence;
+ bool is_system_catalog;
+ Oid ident_idx_old, ident_idx_new;
+ IndexInsertState *iistate;
+ ScanKey ident_key;
+ int ident_key_nentries;
+ XLogRecPtr wal_insert_ptr, end_of_wal;
+ char dummy_rec_data = '\0';
+ RelReopenInfo *rri = NULL;
+ int nrel;
+ Relation *ind_refs_all, *ind_refs_p;
+
+ /* Like in cluster_rel(). */
+ lockmode_old = ShareUpdateExclusiveLock;
+ Assert(CheckRelationLockedByMe(OldHeap, lockmode_old, false));
+ Assert(cl_index == NULL ||
+ CheckRelationLockedByMe(cl_index, lockmode_old, false));
+ /* This is expected from the caller. */
+ Assert(CheckRelationLockedByMe(NewHeap, AccessExclusiveLock, false));
+
+ ident_idx_old = RelationGetReplicaIndex(OldHeap);
+
+ /*
+ * Unlike the exclusive case, we build new indexes for the new relation
+ * rather than swapping the storage and reindexing the old relation. The
+ * point is that the index build can take some time, so we do it before we
+ * get AccessExclusiveLock on the old heap and therefore we cannot swap
+ * the heap storage yet.
+ *
+ * index_create() will lock the new indexes using AccessExclusiveLock
+ * creation - no need to change that.
+ */
+ ind_oids_new = build_new_indexes(NewHeap, OldHeap, ind_oids_old);
+
+ /*
+ * Processing shouldn't start w/o valid identity index.
+ */
+ Assert(OidIsValid(ident_idx_old));
+
+ /* Find "identity index" on the new relation. */
+ ident_idx_new = InvalidOid;
+ forboth(lc, ind_oids_old, lc2, ind_oids_new)
+ {
+ Oid ind_old = lfirst_oid(lc);
+ Oid ind_new = lfirst_oid(lc2);
+
+ if (ident_idx_old == ind_old)
+ {
+ ident_idx_new = ind_new;
+ break;
+ }
+ }
+ if (!OidIsValid(ident_idx_new))
+ /*
+ * Should not happen, given our lock on the old relation.
+ */
+ ereport(ERROR,
+ (errmsg("Identity index missing on the new relation")));
+
+ /* Executor state to update indexes. */
+ iistate = get_index_insert_state(NewHeap, ident_idx_new);
+
+ /*
+ * Build scan key that we'll use to look for rows to be updated / deleted
+ * during logical decoding.
+ */
+ ident_key = build_identity_key(ident_idx_new, OldHeap, &ident_key_nentries);
+
+ /*
+ * Flush all WAL records inserted so far (possibly except for the last
+ * incomplete page, see GetInsertRecPtr), to minimize the amount of data
+ * we need to flush while holding exclusive lock on the source table.
+ */
+ wal_insert_ptr = GetInsertRecPtr();
+ XLogFlush(wal_insert_ptr);
+ end_of_wal = GetFlushRecPtr(NULL);
+
+ /*
+ * Apply concurrent changes first time, to minimize the time we need to
+ * hold AccessExclusiveLock. (Quite some amount of WAL could have been
+ * written during the data copying and index creation.)
+ */
+ process_concurrent_changes(ctx, end_of_wal, NewHeap,
+ swap_toast_by_content ? OldHeap : NULL,
+ ident_key, ident_key_nentries, iistate);
+
+ /*
+ * Release the locks that allowed concurrent data changes, in order to
+ * acquire the AccessExclusiveLock.
+ */
+ nrel = 0;
+ /*
+ * We unlock the old relation (and its clustering index), but then we will
+ * lock the relation and *all* its indexes because we want to swap their
+ * storage.
+ *
+ * (NewHeap is already locked, as well as its indexes.)
+ */
+ rri = palloc_array(RelReopenInfo, 1 + list_length(ind_oids_old));
+ init_rel_reopen_info(&rri[nrel++], &OldHeap, InvalidOid,
+ ShareUpdateExclusiveLock, AccessExclusiveLock);
+ /* References to the re-opened indexes will be stored in this array. */
+ ind_refs_all = palloc_array(Relation, list_length(ind_oids_old));
+ ind_refs_p = ind_refs_all;
+ /* The clustering index is a special case. */
+ if (cl_index)
+ {
+ *ind_refs_p = cl_index;
+ init_rel_reopen_info(&rri[nrel], ind_refs_p, InvalidOid,
+ ShareUpdateExclusiveLock, AccessExclusiveLock);
+ nrel++;
+ ind_refs_p++;
+ }
+ /*
+ * Initialize also the entries for the other indexes (currently unlocked)
+ * because we will have to lock them.
+ */
+ foreach(lc, ind_oids_old)
+ {
+ Oid ind_oid;
+
+ ind_oid = lfirst_oid(lc);
+ /* Clustering index is already in the array, or there is none. */
+ if (cl_index && RelationGetRelid(cl_index) == ind_oid)
+ continue;
+
+ Assert(nrel < (1 + list_length(ind_oids_old)));
+
+ *ind_refs_p = NULL;
+ init_rel_reopen_info(&rri[nrel],
+ /*
+ * In this special case we do not have the
+ * relcache reference, use OID instead.
+ */
+ ind_refs_p,
+ ind_oid,
+ NoLock, /* Nothing to unlock. */
+ AccessExclusiveLock);
+
+ nrel++;
+ ind_refs_p++;
+ }
+ /* Perform the actual unlocking and re-locking. */
+ unlock_and_close_relations(rri, nrel);
+ reopen_relations(rri, nrel);
+
+ /*
+ * In addition, lock the OldHeap's TOAST relation that we skipped for the
+ * CONCURRENTLY option in copy_table_data(). This lock will be needed to
+ * swap the relation files.
+ */
+ if (OidIsValid(OldHeap->rd_rel->reltoastrelid))
+ LockRelationOid(OldHeap->rd_rel->reltoastrelid, AccessExclusiveLock);
+
+ /*
+ * Check if the new indexes match the old ones, i.e. no changes occurred
+ * while OldHeap was unlocked.
+ *
+ * XXX It's probably not necessary to check the relation tuple descriptor
+ * here because the logical decoding was already active when we released
+ * the lock, and thus the corresponding data changes won't be lost.
+ * However processing of those changes might take a lot of time.
+ */
+ check_catalog_changes(OldHeap, cat_state);
+
+ /*
+ * Tuples and pages of the old heap will be gone, but the heap will stay.
+ */
+ TransferPredicateLocksToHeapRelation(OldHeap);
+ /* The same for indexes. */
+ for (int i = 0; i < (nrel - 1); i++)
+ {
+ Relation index = ind_refs_all[i];
+
+ TransferPredicateLocksToHeapRelation(index);
+
+ /*
+ * References to indexes on the old relation are not needed anymore,
+ * however locks stay till the end of the transaction.
+ */
+ index_close(index, NoLock);
+ }
+ pfree(ind_refs_all);
+
+ /*
+ * Flush anything we see in WAL, to make sure that all changes committed
+ * while we were waiting for the exclusive lock are available for
+ * decoding. This should not be necessary if all backends had
+ * synchronous_commit set, but we can't rely on this setting.
+ *
+ * Unfortunately, GetInsertRecPtr() may lag behind the actual insert
+ * position, and GetLastImportantRecPtr() points at the start of the last
+ * record rather than at the end. Thus the simplest way to determine the
+ * insert position is to insert a dummy record and use its LSN.
+ *
+ * XXX Consider using GetLastImportantRecPtr() and adding the size of the
+ * last record (plus the total size of all the page headers the record
+ * spans)?
+ */
+ XLogBeginInsert();
+ XLogRegisterData(&dummy_rec_data, 1);
+ wal_insert_ptr = XLogInsert(RM_XLOG_ID, XLOG_NOOP);
+ XLogFlush(wal_insert_ptr);
+ end_of_wal = GetFlushRecPtr(NULL);
+
+ /* Apply the concurrent changes again. */
+ process_concurrent_changes(ctx, end_of_wal, NewHeap,
+ swap_toast_by_content ? OldHeap : NULL,
+ ident_key, ident_key_nentries, iistate);
+
+ /* Remember info about rel before closing OldHeap */
+ relpersistence = OldHeap->rd_rel->relpersistence;
+ is_system_catalog = IsSystemRelation(OldHeap);
+
+ pgstat_progress_update_param(PROGRESS_CLUSTER_PHASE,
+ PROGRESS_CLUSTER_PHASE_SWAP_REL_FILES);
+
+ forboth(lc, ind_oids_old, lc2, ind_oids_new)
+ {
+ Oid ind_old = lfirst_oid(lc);
+ Oid ind_new = lfirst_oid(lc2);
+ Oid mapped_tables[4];
+
+ /* Zero out possible results from swapped_relation_files */
+ memset(mapped_tables, 0, sizeof(mapped_tables));
+
+ swap_relation_files(ind_old, ind_new,
+ (old_table_oid == RelationRelationId),
+ swap_toast_by_content,
+ true,
+ InvalidTransactionId,
+ InvalidMultiXactId,
+ mapped_tables);
+
+#ifdef USE_ASSERT_CHECKING
+ /*
+ * Concurrent processing is not supported for system relations, so
+ * there should be no mapped tables.
+ */
+ for (int i = 0; i < 4; i++)
+ Assert(mapped_tables[i] == 0);
+#endif
+ }
+
+ /* The new indexes must be visible for deletion. */
+ CommandCounterIncrement();
+
+ /* Close the old heap but keep lock until transaction commit. */
+ table_close(OldHeap, NoLock);
+ /* Close the new heap. (We didn't have to open its indexes). */
+ table_close(NewHeap, NoLock);
+
+ /* Cleanup what we don't need anymore. (And close the identity index.) */
+ pfree(ident_key);
+ free_index_insert_state(iistate);
+
+ /*
+ * Swap the relations and their TOAST relations and TOAST indexes. This
+ * also drops the new relation and its indexes.
+ *
+ * (System catalogs are currently not supported.)
+ */
+ Assert(!is_system_catalog);
+ finish_heap_swap(old_table_oid, new_table_oid,
+ is_system_catalog,
+ swap_toast_by_content,
+ false, true, false,
+ frozenXid, cutoffMulti,
+ relpersistence);
+
+ pfree(rri);
+}
+
+/*
+ * Build indexes on NewHeap according to those on OldHeap.
+ *
+ * OldIndexes is the list of index OIDs on OldHeap.
+ *
+ * A list of OIDs of the corresponding indexes created on NewHeap is
+ * returned. The order of items does match, so we can use these arrays to swap
+ * index storage.
+ */
+static List *
+build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes)
+{
+ StringInfo ind_name;
+ ListCell *lc;
+ List *result = NIL;
+
+ pgstat_progress_update_param(PROGRESS_CLUSTER_PHASE,
+ PROGRESS_CLUSTER_PHASE_REBUILD_INDEX);
+
+ ind_name = makeStringInfo();
+
+ foreach(lc, OldIndexes)
+ {
+ Oid ind_oid,
+ ind_oid_new,
+ tbsp_oid;
+ Relation ind;
+ IndexInfo *ind_info;
+ int i,
+ heap_col_id;
+ List *colnames;
+ int16 indnatts;
+ Oid *collations,
+ *opclasses;
+ HeapTuple tup;
+ bool isnull;
+ Datum d;
+ oidvector *oidvec;
+ int2vector *int2vec;
+ size_t oid_arr_size;
+ size_t int2_arr_size;
+ int16 *indoptions;
+ text *reloptions = NULL;
+ bits16 flags;
+ Datum *opclassOptions;
+ NullableDatum *stattargets;
+
+ ind_oid = lfirst_oid(lc);
+ ind = index_open(ind_oid, AccessShareLock);
+ ind_info = BuildIndexInfo(ind);
+
+ tbsp_oid = ind->rd_rel->reltablespace;
+ /*
+ * Index name really doesn't matter, we'll eventually use only their
+ * storage. Just make them unique within the table.
+ */
+ resetStringInfo(ind_name);
+ appendStringInfo(ind_name, "ind_%d",
+ list_cell_number(OldIndexes, lc));
+
+ flags = 0;
+ if (ind->rd_index->indisprimary)
+ flags |= INDEX_CREATE_IS_PRIMARY;
+
+ colnames = NIL;
+ indnatts = ind->rd_index->indnatts;
+ oid_arr_size = sizeof(Oid) * indnatts;
+ int2_arr_size = sizeof(int16) * indnatts;
+
+ collations = (Oid *) palloc(oid_arr_size);
+ for (i = 0; i < indnatts; i++)
+ {
+ char *colname;
+
+ heap_col_id = ind->rd_index->indkey.values[i];
+ if (heap_col_id > 0)
+ {
+ Form_pg_attribute att;
+
+ /* Normal attribute. */
+ att = TupleDescAttr(OldHeap->rd_att, heap_col_id - 1);
+ colname = pstrdup(NameStr(att->attname));
+ collations[i] = att->attcollation;
+ }
+ else if (heap_col_id == 0)
+ {
+ HeapTuple tuple;
+ Form_pg_attribute att;
+
+ /*
+ * Expression column is not present in relcache. What we need
+ * here is an attribute of the *index* relation.
+ */
+ tuple = SearchSysCache2(ATTNUM,
+ ObjectIdGetDatum(ind_oid),
+ Int16GetDatum(i + 1));
+ if (!HeapTupleIsValid(tuple))
+ elog(ERROR,
+ "cache lookup failed for attribute %d of relation %u",
+ i + 1, ind_oid);
+ att = (Form_pg_attribute) GETSTRUCT(tuple);
+ colname = pstrdup(NameStr(att->attname));
+ collations[i] = att->attcollation;
+ ReleaseSysCache(tuple);
+ }
+ else
+ elog(ERROR, "Unexpected column number: %d",
+ heap_col_id);
+
+ colnames = lappend(colnames, colname);
+ }
+
+ /*
+ * Special effort needed for variable length attributes of
+ * Form_pg_index.
+ */
+ tup = SearchSysCache1(INDEXRELID, ObjectIdGetDatum(ind_oid));
+ if (!HeapTupleIsValid(tup))
+ elog(ERROR, "cache lookup failed for index %u", ind_oid);
+ d = SysCacheGetAttr(INDEXRELID, tup, Anum_pg_index_indclass, &isnull);
+ Assert(!isnull);
+ oidvec = (oidvector *) DatumGetPointer(d);
+ opclasses = (Oid *) palloc(oid_arr_size);
+ memcpy(opclasses, oidvec->values, oid_arr_size);
+
+ d = SysCacheGetAttr(INDEXRELID, tup, Anum_pg_index_indoption,
+ &isnull);
+ Assert(!isnull);
+ int2vec = (int2vector *) DatumGetPointer(d);
+ indoptions = (int16 *) palloc(int2_arr_size);
+ memcpy(indoptions, int2vec->values, int2_arr_size);
+ ReleaseSysCache(tup);
+
+ tup = SearchSysCache1(RELOID, ObjectIdGetDatum(ind_oid));
+ if (!HeapTupleIsValid(tup))
+ elog(ERROR, "cache lookup failed for index relation %u", ind_oid);
+ d = SysCacheGetAttr(RELOID, tup, Anum_pg_class_reloptions, &isnull);
+ reloptions = !isnull ? DatumGetTextPCopy(d) : NULL;
+ ReleaseSysCache(tup);
+
+ opclassOptions = palloc0(sizeof(Datum) * ind_info->ii_NumIndexAttrs);
+ for (i = 0; i < ind_info->ii_NumIndexAttrs; i++)
+ opclassOptions[i] = get_attoptions(ind_oid, i + 1);
+
+ stattargets = get_index_stattargets(ind_oid, ind_info);
+
+ /*
+ * Neither parentIndexRelid nor parentConstraintId needs to be passed
+ * since the new catalog entries (pg_constraint, pg_inherits) would
+ * eventually be dropped. Therefore there's no need to record valid
+ * dependency on parents.
+ */
+ ind_oid_new = index_create(NewHeap,
+ ind_name->data,
+ InvalidOid,
+ InvalidOid, /* parentIndexRelid */
+ InvalidOid, /* parentConstraintId */
+ InvalidOid,
+ ind_info,
+ colnames,
+ ind->rd_rel->relam,
+ tbsp_oid,
+ collations,
+ opclasses,
+ opclassOptions,
+ indoptions,
+ stattargets,
+ PointerGetDatum(reloptions),
+ flags, /* flags */
+ 0, /* constr_flags */
+ false, /* allow_system_table_mods */
+ false, /* is_internal */
+ NULL /* constraintId */
+ );
+ result = lappend_oid(result, ind_oid_new);
+
+ index_close(ind, AccessShareLock);
+ list_free_deep(colnames);
+ pfree(collations);
+ pfree(opclasses);
+ pfree(indoptions);
+ if (reloptions)
+ pfree(reloptions);
+ }
+
+ return result;
+}
+
+static void
+init_rel_reopen_info(RelReopenInfo *rri, Relation *rel_p, Oid relid,
+ LOCKMODE lockmode_orig, LOCKMODE lockmode_new)
+{
+ rri->rel_p = rel_p;
+ rri->relid = relid;
+ rri->lockmode_orig = lockmode_orig;
+ rri->lockmode_new = lockmode_new;
+}
+
+/*
+ * Unlock and close relations specified by items of the 'rels' array. 'nrels'
+ * is the number of items.
+ *
+ * Information needed to (re)open the relations (or to issue meaningful ERROR)
+ * is added to the array items.
+ */
+static void
+unlock_and_close_relations(RelReopenInfo *rels, int nrel)
+{
+ int i;
+ RelReopenInfo *rri;
+
+ /*
+ * First, retrieve the information that we will need for re-opening.
+ *
+ * We could close (and unlock) each relation as soon as we have gathered
+ * the related information, but then we would have to be careful not to
+ * unlock the table until we have the info on all its indexes. (Once we
+ * unlock the table, any index can be dropped, and thus we can fail to get
+ * the name we want to report if re-opening fails.) It seem simpler to
+ * separate the work into two iterations.
+ */
+ for (i = 0; i < nrel; i++)
+ {
+ Relation rel;
+
+ rri = &rels[i];
+ rel = *rri->rel_p;
+
+ if (rel)
+ {
+ Assert(CheckRelationLockedByMe(rel, rri->lockmode_orig, false));
+ Assert(!OidIsValid(rri->relid));
+
+ rri->relid = RelationGetRelid(rel);
+ rri->relkind = rel->rd_rel->relkind;
+ rri->relname = pstrdup(RelationGetRelationName(rel));
+ }
+ else
+ {
+ Assert(OidIsValid(rri->relid));
+
+ rri->relname = get_rel_name(rri->relid);
+ rri->relkind = get_rel_relkind(rri->relid);
+ }
+ }
+
+ /* Second, close the relations. */
+ for (i = 0; i < nrel; i++)
+ {
+ Relation rel;
+
+ rri = &rels[i];
+ rel = *rri->rel_p;
+
+ /* Close the relation if the caller passed one. */
+ if (rel)
+ {
+ if (rri->relkind == RELKIND_RELATION)
+ table_close(rel, rri->lockmode_orig);
+ else
+ {
+ Assert(rri->relkind == RELKIND_INDEX);
+
+ index_close(rel, rri->lockmode_orig);
+ }
+ }
+ }
+}
+
+/*
+ * Re-open the relations closed previously by unlock_and_close_relations().
+ */
+static void
+reopen_relations(RelReopenInfo *rels, int nrel)
+{
+ for (int i = 0; i < nrel; i++)
+ {
+ RelReopenInfo *rri = &rels[i];
+ Relation rel;
+
+ if (rri->relkind == RELKIND_RELATION)
+ {
+ rel = try_table_open(rri->relid, rri->lockmode_new);
+ }
+ else
+ {
+ Assert(rri->relkind == RELKIND_INDEX);
+
+ rel = try_index_open(rri->relid, rri->lockmode_new);
+ }
+
+ if (rel == NULL)
+ {
+ const char *kind_str;
+
+ kind_str = (rri->relkind == RELKIND_RELATION) ? "table" : "index";
+ ereport(ERROR,
+ (errmsg("could not open \%s \"%s\"", kind_str,
+ rri->relname),
+ errhint("The %s could have been dropped by another transaction.",
+ kind_str)));
+ }
+ *rri->rel_p = rel;
+
+ pfree(rri->relname);
+ }
+}
diff --git a/src/backend/commands/matview.c b/src/backend/commands/matview.c
index c12817091e..8f0197378d 100644
--- a/src/backend/commands/matview.c
+++ b/src/backend/commands/matview.c
@@ -905,7 +905,7 @@ refresh_by_match_merge(Oid matviewOid, Oid tempOid, Oid relowner,
static void
refresh_by_heap_swap(Oid matviewOid, Oid OIDNewHeap, char relpersistence)
{
- finish_heap_swap(matviewOid, OIDNewHeap, false, false, true, true,
+ finish_heap_swap(matviewOid, OIDNewHeap, false, false, true, true, true,
RecentXmin, ReadNextMultiXactId(), relpersistence);
}
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 4fc54bd6eb..f070ffc960 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -4500,6 +4500,16 @@ AlterTableInternal(Oid relid, List *cmds, bool recurse)
rel = relation_open(relid, lockmode);
+ /*
+ * If lockmode allows, check if VACUUM FULL / CLUSTER CONCURRENTLY is in
+ * progress. If lockmode is too weak, cluster_rel() should detect
+ * incompatible DDLs executed by us.
+ *
+ * XXX We might skip the changes for DDLs which do not change the tuple
+ * descriptor.
+ */
+ check_for_concurrent_cluster(relid, lockmode);
+
EventTriggerAlterTableRelid(relid);
ATController(NULL, rel, cmds, recurse, lockmode, NULL);
@@ -5929,6 +5939,7 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode,
finish_heap_swap(tab->relid, OIDNewHeap,
false, false, true,
!OidIsValid(tab->newTableSpace),
+ true,
RecentXmin,
ReadNextMultiXactId(),
persistence);
diff --git a/src/backend/commands/vacuum.c b/src/backend/commands/vacuum.c
index e6745e6145..516a064fed 100644
--- a/src/backend/commands/vacuum.c
+++ b/src/backend/commands/vacuum.c
@@ -111,7 +111,7 @@ static void vac_truncate_clog(TransactionId frozenXID,
TransactionId lastSaneFrozenXid,
MultiXactId lastSaneMinMulti);
static bool vacuum_rel(Oid relid, RangeVar *relation, VacuumParams *params,
- BufferAccessStrategy bstrategy);
+ BufferAccessStrategy bstrategy, bool isTopLevel);
static double compute_parallel_delay(void);
static VacOptValue get_vacoptval_from_boolean(DefElem *def);
static bool vac_tid_reaped(ItemPointer itemptr, void *state);
@@ -153,6 +153,7 @@ ExecVacuum(ParseState *pstate, VacuumStmt *vacstmt, bool isTopLevel)
bool analyze = false;
bool freeze = false;
bool full = false;
+ bool concurrent = false;
bool disable_page_skipping = false;
bool process_main = true;
bool process_toast = true;
@@ -226,6 +227,8 @@ ExecVacuum(ParseState *pstate, VacuumStmt *vacstmt, bool isTopLevel)
freeze = defGetBoolean(opt);
else if (strcmp(opt->defname, "full") == 0)
full = defGetBoolean(opt);
+ else if (strcmp(opt->defname, "concurrently") == 0)
+ concurrent = defGetBoolean(opt);
else if (strcmp(opt->defname, "disable_page_skipping") == 0)
disable_page_skipping = defGetBoolean(opt);
else if (strcmp(opt->defname, "index_cleanup") == 0)
@@ -300,7 +303,7 @@ ExecVacuum(ParseState *pstate, VacuumStmt *vacstmt, bool isTopLevel)
(skip_locked ? VACOPT_SKIP_LOCKED : 0) |
(analyze ? VACOPT_ANALYZE : 0) |
(freeze ? VACOPT_FREEZE : 0) |
- (full ? VACOPT_FULL : 0) |
+ (full ? (concurrent ? VACOPT_FULL_CONCURRENT : VACOPT_FULL_EXCLUSIVE) : 0) |
(disable_page_skipping ? VACOPT_DISABLE_PAGE_SKIPPING : 0) |
(process_main ? VACOPT_PROCESS_MAIN : 0) |
(process_toast ? VACOPT_PROCESS_TOAST : 0) |
@@ -380,6 +383,12 @@ ExecVacuum(ParseState *pstate, VacuumStmt *vacstmt, bool isTopLevel)
errmsg("ONLY_DATABASE_STATS cannot be specified with other VACUUM options")));
}
+ /* This problem cannot be identified from the options. */
+ if (concurrent && !full)
+ ereport(ERROR,
+ (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
+ errmsg("CONCURRENTLY can only be specified with VACUUM FULL")));
+
/*
* All freeze ages are zero if the FREEZE option is given; otherwise pass
* them as -1 which means to use the default values.
@@ -543,7 +552,17 @@ vacuum(List *relations, VacuumParams *params, BufferAccessStrategy bstrategy,
relations = newrels;
}
else
+ {
+ /*
+ * Concurrent processing is currently considered rather special so it
+ * is not performed in bulk.
+ */
+ if (params->options & VACOPT_FULL_CONCURRENT)
+ ereport(ERROR,
+ (errmsg("VACUUM (CONCURRENTLY) requires explicit list of tables")));
+
relations = get_all_vacuum_rels(vac_context, params->options);
+ }
/*
* Decide whether we need to start/commit our own transactions.
@@ -616,7 +635,8 @@ vacuum(List *relations, VacuumParams *params, BufferAccessStrategy bstrategy,
if (params->options & VACOPT_VACUUM)
{
- if (!vacuum_rel(vrel->oid, vrel->relation, params, bstrategy))
+ if (!vacuum_rel(vrel->oid, vrel->relation, params, bstrategy,
+ isTopLevel))
continue;
}
@@ -960,6 +980,17 @@ expand_vacuum_rel(VacuumRelation *vrel, MemoryContext vac_context,
(errmsg("VACUUM ONLY of partitioned table \"%s\" has no effect",
vrel->relation->relname)));
+ /*
+ * Concurrent processing is currently considered rather special
+ * (e.g. in terms of resources consumed) so it is not performed in
+ * bulk.
+ */
+ if (is_partitioned_table && (options & VACOPT_FULL_CONCURRENT))
+ ereport(ERROR,
+ (errmsg("VACUUM (CONCURRENTLY) not supported for partitioned tables"),
+ errhint("Consider running the command for individual partitions.")));
+
+
ReleaseSysCache(tuple);
/*
@@ -1954,10 +1985,10 @@ vac_truncate_clog(TransactionId frozenXID,
/*
* vacuum_rel() -- vacuum one heap relation
*
- * relid identifies the relation to vacuum. If relation is supplied,
- * use the name therein for reporting any failure to open/lock the rel;
- * do not use it once we've successfully opened the rel, since it might
- * be stale.
+ * relid identifies the relation to vacuum. If relation is supplied, use
+ * the name therein for reporting any failure to open/lock the rel; do
+ * not use it once we've successfully opened the rel, since it might be
+ * stale.
*
* Returns true if it's okay to proceed with a requested ANALYZE
* operation on this table.
@@ -1972,7 +2003,7 @@ vac_truncate_clog(TransactionId frozenXID,
*/
static bool
vacuum_rel(Oid relid, RangeVar *relation, VacuumParams *params,
- BufferAccessStrategy bstrategy)
+ BufferAccessStrategy bstrategy, bool isTopLevel)
{
LOCKMODE lmode;
Relation rel;
@@ -2035,10 +2066,11 @@ vacuum_rel(Oid relid, RangeVar *relation, VacuumParams *params,
/*
* Determine the type of lock we want --- hard exclusive lock for a FULL
- * vacuum, but just ShareUpdateExclusiveLock for concurrent vacuum. Either
- * way, we can be sure that no other backend is vacuuming the same table.
+ * exclusive vacuum, but a weaker lock (ShareUpdateExclusiveLock) for
+ * concurrent vacuum. Either way, we can be sure that no other backend is
+ * vacuuming the same table.
*/
- lmode = (params->options & VACOPT_FULL) ?
+ lmode = (params->options & VACOPT_FULL_EXCLUSIVE) ?
AccessExclusiveLock : ShareUpdateExclusiveLock;
/* open the relation and get the appropriate lock on it */
@@ -2053,6 +2085,22 @@ vacuum_rel(Oid relid, RangeVar *relation, VacuumParams *params,
return false;
}
+ /*
+ * Skip the relation if VACUUM FULL / CLUSTER CONCURRENTLY is in progress
+ * as it will drop the current storage of the relation.
+ *
+ * This check should not take place until we have a lock that prevents
+ * another backend from starting VACUUM FULL / CLUSTER CONCURRENTLY later.
+ */
+ Assert(lmode >= ShareUpdateExclusiveLock);
+ if (is_concurrent_cluster_in_progress(relid))
+ {
+ relation_close(rel, lmode);
+ PopActiveSnapshot();
+ CommitTransactionCommand();
+ return false;
+ }
+
/*
* When recursing to a TOAST table, check privileges on the parent. NB:
* This is only safe to do because we hold a session lock on the main
@@ -2126,19 +2174,6 @@ vacuum_rel(Oid relid, RangeVar *relation, VacuumParams *params,
return true;
}
- /*
- * Get a session-level lock too. This will protect our access to the
- * relation across multiple transactions, so that we can vacuum the
- * relation's TOAST table (if any) secure in the knowledge that no one is
- * deleting the parent relation.
- *
- * NOTE: this cannot block, even if someone else is waiting for access,
- * because the lock manager knows that both lock requests are from the
- * same process.
- */
- lockrelid = rel->rd_lockInfo.lockRelId;
- LockRelationIdForSession(&lockrelid, lmode);
-
/*
* Set index_cleanup option based on index_cleanup reloption if it wasn't
* specified in VACUUM command, or when running in an autovacuum worker
@@ -2191,6 +2226,30 @@ vacuum_rel(Oid relid, RangeVar *relation, VacuumParams *params,
else
toast_relid = InvalidOid;
+ /*
+ * Get a session-level lock too. This will protect our access to the
+ * relation across multiple transactions, so that we can vacuum the
+ * relation's TOAST table (if any) secure in the knowledge that no one is
+ * deleting the parent relation.
+ *
+ * NOTE: this cannot block, even if someone else is waiting for access,
+ * because the lock manager knows that both lock requests are from the
+ * same process.
+ */
+ if (OidIsValid(toast_relid))
+ {
+ /*
+ * You might worry that, in the VACUUM (FULL, CONCURRENTLY) case,
+ * cluster_rel() needs to release all the locks on the relation at
+ * some point, but this session lock makes it impossible. In fact,
+ * cluster_rel() will will eventually be called for the TOAST relation
+ * and raise ERROR because, in the concurrent mode, it cannot process
+ * TOAST relation alone anyway.
+ */
+ lockrelid = rel->rd_lockInfo.lockRelId;
+ LockRelationIdForSession(&lockrelid, lmode);
+ }
+
/*
* Switch to the table owner's userid, so that any index functions are run
* as that user. Also lock down security-restricted operations and
@@ -2218,11 +2277,22 @@ vacuum_rel(Oid relid, RangeVar *relation, VacuumParams *params,
{
ClusterParams cluster_params = {0};
+ /*
+ * Invalid toast_relid means that there is no session lock on the
+ * relation. Such a lock would be a problem because it would
+ * prevent cluster_rel() from releasing all locks when it tries to
+ * get AccessExclusiveLock.
+ */
+ Assert(!OidIsValid(toast_relid));
+
if ((params->options & VACOPT_VERBOSE) != 0)
cluster_params.options |= CLUOPT_VERBOSE;
+ if ((params->options & VACOPT_FULL_CONCURRENT) != 0)
+ cluster_params.options |= CLUOPT_CONCURRENT;
+
/* VACUUM FULL is now a variant of CLUSTER; see cluster.c */
- cluster_rel(rel, InvalidOid, &cluster_params);
+ cluster_rel(rel, InvalidOid, &cluster_params, isTopLevel, true);
/* cluster_rel closes the relation, but keeps lock */
rel = NULL;
@@ -2268,13 +2338,15 @@ vacuum_rel(Oid relid, RangeVar *relation, VacuumParams *params,
toast_vacuum_params.options |= VACOPT_PROCESS_MAIN;
toast_vacuum_params.toast_parent = relid;
- vacuum_rel(toast_relid, NULL, &toast_vacuum_params, bstrategy);
+ vacuum_rel(toast_relid, NULL, &toast_vacuum_params, bstrategy,
+ isTopLevel);
}
/*
* Now release the session-level lock on the main table.
*/
- UnlockRelationIdForSession(&lockrelid, lmode);
+ if (OidIsValid(toast_relid))
+ UnlockRelationIdForSession(&lockrelid, lmode);
/* Report that we really did it. */
return true;
diff --git a/src/backend/meson.build b/src/backend/meson.build
index 2b0db21480..4b00a9b8c6 100644
--- a/src/backend/meson.build
+++ b/src/backend/meson.build
@@ -194,5 +194,6 @@ pg_test_mod_args = pg_mod_args + {
subdir('jit/llvm')
subdir('replication/libpqwalreceiver')
subdir('replication/pgoutput')
+subdir('replication/pgoutput_cluster')
subdir('snowball')
subdir('utils/mb/conversion_procs')
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index 0bff0f1065..8f45a7a168 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -33,6 +33,7 @@
#include "access/xlogreader.h"
#include "access/xlogrecord.h"
#include "catalog/pg_control.h"
+#include "commands/cluster.h"
#include "replication/decode.h"
#include "replication/logical.h"
#include "replication/message.h"
@@ -467,6 +468,29 @@ heap_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
TransactionId xid = XLogRecGetXid(buf->record);
SnapBuild *builder = ctx->snapshot_builder;
+ /*
+ * Check if CLUSTER CONCURRENTLY is being performed by this backend. If
+ * so, only decode data changes of the table that it is processing, and
+ * the changes of its TOAST relation.
+ *
+ * (TOAST locator should not be set unless the main is.)
+ */
+ Assert(!OidIsValid(clustered_rel_toast_locator.relNumber) ||
+ OidIsValid(clustered_rel_locator.relNumber));
+
+ if (OidIsValid(clustered_rel_locator.relNumber))
+ {
+ XLogReaderState *r = buf->record;
+ RelFileLocator locator;
+
+ /* Not all records contain the block. */
+ if (XLogRecGetBlockTagExtended(r, 0, &locator, NULL, NULL, NULL) &&
+ !RelFileLocatorEquals(locator, clustered_rel_locator) &&
+ (!OidIsValid(clustered_rel_toast_locator.relNumber) ||
+ !RelFileLocatorEquals(locator, clustered_rel_toast_locator)))
+ return;
+ }
+
ReorderBufferProcessXid(ctx->reorder, xid, buf->origptr);
/*
diff --git a/src/backend/replication/logical/snapbuild.c b/src/backend/replication/logical/snapbuild.c
index 84bf0503a5..86a9b0335a 100644
--- a/src/backend/replication/logical/snapbuild.c
+++ b/src/backend/replication/logical/snapbuild.c
@@ -486,6 +486,26 @@ SnapBuildInitialSnapshot(SnapBuild *builder)
return SnapBuildMVCCFromHistoric(snap, true);
}
+/*
+ * Build an MVCC snapshot for the initial data load performed by CLUSTER
+ * CONCURRENTLY command.
+ *
+ * The snapshot will only be used to scan one particular relation, which is
+ * treated like a catalog (therefore ->building_full_snapshot is not
+ * important), and the caller should already have a replication slot setup (so
+ * we do not set MyProc->xmin). XXX Do we yet need to add some restrictions?
+ */
+Snapshot
+SnapBuildInitialSnapshotForCluster(SnapBuild *builder)
+{
+ Snapshot snap;
+
+ Assert(builder->state == SNAPBUILD_CONSISTENT);
+
+ snap = SnapBuildBuildSnapshot(builder);
+ return SnapBuildMVCCFromHistoric(snap, false);
+}
+
/*
* Turn a historic MVCC snapshot into an ordinary MVCC snapshot.
*
diff --git a/src/backend/replication/pgoutput_cluster/Makefile b/src/backend/replication/pgoutput_cluster/Makefile
new file mode 100644
index 0000000000..31471bb546
--- /dev/null
+++ b/src/backend/replication/pgoutput_cluster/Makefile
@@ -0,0 +1,32 @@
+#-------------------------------------------------------------------------
+#
+# Makefile--
+# Makefile for src/backend/replication/pgoutput_cluster
+#
+# IDENTIFICATION
+# src/backend/replication/pgoutput_cluster
+#
+#-------------------------------------------------------------------------
+
+subdir = src/backend/replication/pgoutput_cluster
+top_builddir = ../../../..
+include $(top_builddir)/src/Makefile.global
+
+OBJS = \
+ $(WIN32RES) \
+ pgoutput_cluster.o
+PGFILEDESC = "pgoutput_cluster - logical replication output plugin for CLUSTER command"
+NAME = pgoutput_cluster
+
+all: all-shared-lib
+
+include $(top_srcdir)/src/Makefile.shlib
+
+install: all installdirs install-lib
+
+installdirs: installdirs-lib
+
+uninstall: uninstall-lib
+
+clean distclean: clean-lib
+ rm -f $(OBJS)
diff --git a/src/backend/replication/pgoutput_cluster/meson.build b/src/backend/replication/pgoutput_cluster/meson.build
new file mode 100644
index 0000000000..0f033064f2
--- /dev/null
+++ b/src/backend/replication/pgoutput_cluster/meson.build
@@ -0,0 +1,18 @@
+# Copyright (c) 2022-2024, PostgreSQL Global Development Group
+
+pgoutput_cluster_sources = files(
+ 'pgoutput_cluster.c',
+)
+
+if host_system == 'windows'
+ pgoutput_cluster_sources += rc_lib_gen.process(win32ver_rc, extra_args: [
+ '--NAME', 'pgoutput_cluster',
+ '--FILEDESC', 'pgoutput_cluster - logical replication output plugin for CLUSTER command',])
+endif
+
+pgoutput_cluster = shared_module('pgoutput_cluster',
+ pgoutput_cluster_sources,
+ kwargs: pg_mod_args,
+)
+
+backend_targets += pgoutput_cluster
diff --git a/src/backend/replication/pgoutput_cluster/pgoutput_cluster.c b/src/backend/replication/pgoutput_cluster/pgoutput_cluster.c
new file mode 100644
index 0000000000..43f7b34297
--- /dev/null
+++ b/src/backend/replication/pgoutput_cluster/pgoutput_cluster.c
@@ -0,0 +1,288 @@
+/* TODO Move into src/backend/cluster/ (and rename?) */
+/*-------------------------------------------------------------------------
+ *
+ * pgoutput_cluster.c
+ * Logical Replication output plugin for CLUSTER command
+ *
+ * Copyright (c) 2012-2024, PostgreSQL Global Development Group
+ *
+ * IDENTIFICATION
+ * src/backend/replication/pgoutput_cluster/pgoutput_cluster.c
+ *
+ *-------------------------------------------------------------------------
+ */
+#include "postgres.h"
+
+#include "access/heaptoast.h"
+#include "commands/cluster.h"
+#include "replication/snapbuild.h"
+
+PG_MODULE_MAGIC;
+
+static void plugin_startup(LogicalDecodingContext *ctx,
+ OutputPluginOptions *opt, bool is_init);
+static void plugin_shutdown(LogicalDecodingContext *ctx);
+static void plugin_begin_txn(LogicalDecodingContext *ctx,
+ ReorderBufferTXN *txn);
+static void plugin_commit_txn(LogicalDecodingContext *ctx,
+ ReorderBufferTXN *txn, XLogRecPtr commit_lsn);
+static void plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
+ Relation rel, ReorderBufferChange *change);
+static void plugin_truncate(struct LogicalDecodingContext *ctx,
+ ReorderBufferTXN *txn, int nrelations,
+ Relation relations[],
+ ReorderBufferChange *change);
+static void store_change(LogicalDecodingContext *ctx,
+ ConcurrentChangeKind kind, HeapTuple tuple);
+
+void
+_PG_output_plugin_init(OutputPluginCallbacks *cb)
+{
+ AssertVariableIsOfType(&_PG_output_plugin_init, LogicalOutputPluginInit);
+
+ cb->startup_cb = plugin_startup;
+ cb->begin_cb = plugin_begin_txn;
+ cb->change_cb = plugin_change;
+ cb->truncate_cb = plugin_truncate;
+ cb->commit_cb = plugin_commit_txn;
+ cb->shutdown_cb = plugin_shutdown;
+}
+
+
+/* initialize this plugin */
+static void
+plugin_startup(LogicalDecodingContext *ctx, OutputPluginOptions *opt,
+ bool is_init)
+{
+ ctx->output_plugin_private = NULL;
+
+ /* Probably unnecessary, as we don't use the SQL interface ... */
+ opt->output_type = OUTPUT_PLUGIN_BINARY_OUTPUT;
+
+ if (ctx->output_plugin_options != NIL)
+ {
+ ereport(ERROR,
+ (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
+ errmsg("This plugin does not expect any options")));
+ }
+}
+
+static void
+plugin_shutdown(LogicalDecodingContext *ctx)
+{
+}
+
+/*
+ * As we don't release the slot during processing of particular table, there's
+ * no room for SQL interface, even for debugging purposes. Therefore we need
+ * neither OutputPluginPrepareWrite() nor OutputPluginWrite() in the plugin
+ * callbacks. (Although we might want to write custom callbacks, this API
+ * seems to be unnecessarily generic for our purposes.)
+ */
+
+/* BEGIN callback */
+static void
+plugin_begin_txn(LogicalDecodingContext *ctx, ReorderBufferTXN *txn)
+{
+}
+
+/* COMMIT callback */
+static void
+plugin_commit_txn(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
+ XLogRecPtr commit_lsn)
+{
+}
+
+/*
+ * Callback for individual changed tuples
+ */
+static void
+plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
+ Relation relation, ReorderBufferChange *change)
+{
+ ClusterDecodingState *dstate;
+
+ dstate = (ClusterDecodingState *) ctx->output_writer_private;
+
+ /* Only interested in one particular relation. */
+ if (relation->rd_id != dstate->relid)
+ return;
+
+ /* Decode entry depending on its type */
+ switch (change->action)
+ {
+ case REORDER_BUFFER_CHANGE_INSERT:
+ {
+ HeapTuple newtuple;
+
+ newtuple = change->data.tp.newtuple != NULL ?
+ change->data.tp.newtuple : NULL;
+
+ /*
+ * Identity checks in the main function should have made this
+ * impossible.
+ */
+ if (newtuple == NULL)
+ elog(ERROR, "Incomplete insert info.");
+
+ store_change(ctx, CHANGE_INSERT, newtuple);
+ }
+ break;
+ case REORDER_BUFFER_CHANGE_UPDATE:
+ {
+ HeapTuple oldtuple,
+ newtuple;
+
+ oldtuple = change->data.tp.oldtuple != NULL ?
+ change->data.tp.oldtuple : NULL;
+ newtuple = change->data.tp.newtuple != NULL ?
+ change->data.tp.newtuple : NULL;
+
+ if (newtuple == NULL)
+ elog(ERROR, "Incomplete update info.");
+
+ if (oldtuple != NULL)
+ store_change(ctx, CHANGE_UPDATE_OLD, oldtuple);
+
+ store_change(ctx, CHANGE_UPDATE_NEW, newtuple);
+ }
+ break;
+ case REORDER_BUFFER_CHANGE_DELETE:
+ {
+ HeapTuple oldtuple;
+
+ oldtuple = change->data.tp.oldtuple ?
+ change->data.tp.oldtuple : NULL;
+
+ if (oldtuple == NULL)
+ elog(ERROR, "Incomplete delete info.");
+
+ store_change(ctx, CHANGE_DELETE, oldtuple);
+ }
+ break;
+ default:
+ /* Should not come here */
+ Assert(false);
+ break;
+ }
+}
+
+static void
+plugin_truncate(struct LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
+ int nrelations, Relation relations[],
+ ReorderBufferChange *change)
+{
+ ClusterDecodingState *dstate;
+ int i;
+ Relation relation = NULL;
+
+ dstate = (ClusterDecodingState *) ctx->output_writer_private;
+
+ /* Find the relation we are processing. */
+ for (i = 0; i < nrelations; i++)
+ {
+ relation = relations[i];
+
+ if (RelationGetRelid(relation) == dstate->relid)
+ break;
+ }
+
+ /* Is this truncation of another relation? */
+ if (i == nrelations)
+ return;
+
+ store_change(ctx, CHANGE_TRUNCATE, NULL);
+}
+
+/* Store concurrent data change. */
+static void
+store_change(LogicalDecodingContext *ctx, ConcurrentChangeKind kind,
+ HeapTuple tuple)
+{
+ ClusterDecodingState *dstate;
+ char *change_raw;
+ ConcurrentChange change;
+ bool flattened = false;
+ Size size;
+ Datum values[1];
+ bool isnull[1];
+ char *dst, *dst_start;
+
+ dstate = (ClusterDecodingState *) ctx->output_writer_private;
+
+ size = MAXALIGN(VARHDRSZ) + SizeOfConcurrentChange;
+
+ if (tuple)
+ {
+ /*
+ * ReorderBufferCommit() stores the TOAST chunks in its private memory
+ * context and frees them after having called
+ * apply_change(). Therefore we need flat copy (including TOAST) that
+ * we eventually copy into the memory context which is available to
+ * decode_concurrent_changes().
+ */
+ if (HeapTupleHasExternal(tuple))
+ {
+ /*
+ * toast_flatten_tuple_to_datum() might be more convenient but we
+ * don't want the decompression it does.
+ */
+ tuple = toast_flatten_tuple(tuple, dstate->tupdesc);
+ flattened = true;
+ }
+
+ size += tuple->t_len;
+ }
+
+ /* XXX Isn't there any function / macro to do this? */
+ if (size >= 0x3FFFFFFF)
+ elog(ERROR, "Change is too big.");
+
+ /* Construct the change. */
+ change_raw = (char *) palloc0(size);
+ SET_VARSIZE(change_raw, size);
+ /*
+ * Since the varlena alignment might not be sufficient for the structure,
+ * set the fields in a local instance and remember where it should
+ * eventually be copied.
+ */
+ change.kind = kind;
+ dst_start = (char *) VARDATA(change_raw);
+
+ /* No other information is needed for TRUNCATE. */
+ if (change.kind == CHANGE_TRUNCATE)
+ {
+ memcpy(dst_start, &change, SizeOfConcurrentChange);
+ goto store;
+ }
+
+ /*
+ * Copy the tuple.
+ *
+ * CAUTION: change->tup_data.t_data must be fixed on retrieval!
+ */
+ memcpy(&change.tup_data, tuple, sizeof(HeapTupleData));
+ dst = dst_start + SizeOfConcurrentChange;
+ memcpy(dst, tuple->t_data, tuple->t_len);
+
+ /* The data has been copied. */
+ if (flattened)
+ pfree(tuple);
+
+store:
+ /* Copy the structure so it can be stored. */
+ memcpy(dst_start, &change, SizeOfConcurrentChange);
+
+ /* Store as tuple of 1 bytea column. */
+ values[0] = PointerGetDatum(change_raw);
+ isnull[0] = false;
+ tuplestore_putvalues(dstate->tstore, dstate->tupdesc_change,
+ values, isnull);
+
+ /* Accounting. */
+ dstate->nchanges++;
+
+ /* Cleanup. */
+ pfree(change_raw);
+}
+
diff --git a/src/backend/storage/ipc/ipci.c b/src/backend/storage/ipc/ipci.c
index 174eed7036..56df243a8a 100644
--- a/src/backend/storage/ipc/ipci.c
+++ b/src/backend/storage/ipc/ipci.c
@@ -25,6 +25,7 @@
#include "access/xlogprefetcher.h"
#include "access/xlogrecovery.h"
#include "commands/async.h"
+#include "commands/cluster.h"
#include "miscadmin.h"
#include "pgstat.h"
#include "postmaster/autovacuum.h"
@@ -148,6 +149,7 @@ CalculateShmemSize(int *num_semaphores)
size = add_size(size, WaitEventCustomShmemSize());
size = add_size(size, InjectionPointShmemSize());
size = add_size(size, SlotSyncShmemSize());
+ size = add_size(size, ClusterShmemSize());
/* include additional requested shmem from preload libraries */
size = add_size(size, total_addin_request);
@@ -340,6 +342,7 @@ CreateOrAttachShmemStructs(void)
StatsShmemInit();
WaitEventCustomShmemInit();
InjectionPointShmemInit();
+ ClusterShmemInit();
}
/*
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 25fe3d5801..7e5484dae4 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1302,6 +1302,17 @@ ProcessUtilitySlow(ParseState *pstate,
lockmode = AlterTableGetLockLevel(atstmt->cmds);
relid = AlterTableLookupRelation(atstmt, lockmode);
+ /*
+ * If lockmode allows, check if VACUUM FULL / CLUSTER
+ * CONCURRENT is in progress. If lockmode is too weak,
+ * cluster_rel() should detect incompatible DDLs executed
+ * by us.
+ *
+ * XXX We might skip the changes for DDLs which do not
+ * change the tuple descriptor.
+ */
+ check_for_concurrent_cluster(relid, lockmode);
+
if (OidIsValid(relid))
{
AlterTableUtilityContext atcontext;
diff --git a/src/backend/utils/activity/backend_progress.c b/src/backend/utils/activity/backend_progress.c
index eebc968193..e2c84baba9 100644
--- a/src/backend/utils/activity/backend_progress.c
+++ b/src/backend/utils/activity/backend_progress.c
@@ -162,3 +162,19 @@ pgstat_progress_end_command(void)
beentry->st_progress.command_target = InvalidOid;
PGSTAT_END_WRITE_ACTIVITY(beentry);
}
+
+void
+pgstat_progress_restore_state(PgBackendProgress *backup)
+{
+ volatile PgBackendStatus *beentry = MyBEEntry;
+
+ if (!beentry || !pgstat_track_activities)
+ return;
+
+ PGSTAT_BEGIN_WRITE_ACTIVITY(beentry);
+ beentry->st_progress.command = backup->command;
+ beentry->st_progress.command_target = backup->command_target;
+ memcpy(MyBEEntry->st_progress.param, backup->param,
+ sizeof(beentry->st_progress.param));
+ PGSTAT_END_WRITE_ACTIVITY(beentry);
+}
diff --git a/src/backend/utils/activity/wait_event_names.txt b/src/backend/utils/activity/wait_event_names.txt
index 0b53cba807..a28fba18a7 100644
--- a/src/backend/utils/activity/wait_event_names.txt
+++ b/src/backend/utils/activity/wait_event_names.txt
@@ -345,6 +345,7 @@ WALSummarizer "Waiting to read or update WAL summarization state."
DSMRegistry "Waiting to read or update the dynamic shared memory registry."
InjectionPoint "Waiting to read or update information related to injection points."
SerialControl "Waiting to read or update shared <filename>pg_serial</filename> state."
+ClusteredRels "Waiting to read or update information on tables being clustered concurrently."
#
# END OF PREDEFINED LWLOCKS (DO NOT CHANGE THIS LINE)
diff --git a/src/backend/utils/cache/inval.c b/src/backend/utils/cache/inval.c
index 254947e943..e2d030e6a1 100644
--- a/src/backend/utils/cache/inval.c
+++ b/src/backend/utils/cache/inval.c
@@ -1565,6 +1565,28 @@ CacheInvalidateRelcache(Relation relation)
databaseId, relationId);
}
+/*
+ * CacheInvalidateRelcacheImmediate
+ * Send invalidation message for the specified relation's relcache entry.
+ *
+ * Currently this is used in VACUUM FULL/CLUSTER CONCURRENTLY, to make sure
+ * that other backends are aware that the command is being executed for the
+ * relation.
+ */
+void
+CacheInvalidateRelcacheImmediate(Relation relation)
+{
+ SharedInvalidationMessage msg;
+
+ msg.rc.id = SHAREDINVALRELCACHE_ID;
+ msg.rc.dbId = MyDatabaseId;
+ msg.rc.relId = RelationGetRelid(relation);
+ /* check AddCatcacheInvalidationMessage() for an explanation */
+ VALGRIND_MAKE_MEM_DEFINED(&msg, sizeof(msg));
+
+ SendSharedInvalidMessages(&msg, 1);
+}
+
/*
* CacheInvalidateRelcacheAll
* Register invalidation of the whole relcache at the end of command.
diff --git a/src/backend/utils/cache/relcache.c b/src/backend/utils/cache/relcache.c
index 43219a9629..55557c7589 100644
--- a/src/backend/utils/cache/relcache.c
+++ b/src/backend/utils/cache/relcache.c
@@ -64,6 +64,7 @@
#include "catalog/pg_type.h"
#include "catalog/schemapg.h"
#include "catalog/storage.h"
+#include "commands/cluster.h"
#include "commands/policy.h"
#include "commands/publicationcmds.h"
#include "commands/trigger.h"
@@ -1246,6 +1247,10 @@ retry:
/* make sure relation is marked as having no open file yet */
relation->rd_smgr = NULL;
+ /* Is CLUSTER CONCURRENTLY in progress? */
+ relation->rd_cluster_concurrent =
+ is_concurrent_cluster_in_progress(targetRelId);
+
/*
* now we can free the memory allocated for pg_class_tuple
*/
diff --git a/src/backend/utils/time/snapmgr.c b/src/backend/utils/time/snapmgr.c
index 42bded373b..103d1249bb 100644
--- a/src/backend/utils/time/snapmgr.c
+++ b/src/backend/utils/time/snapmgr.c
@@ -154,7 +154,6 @@ static List *exportedSnapshots = NIL;
/* Prototypes for local functions */
static void UnregisterSnapshotNoOwner(Snapshot snapshot);
-static void FreeSnapshot(Snapshot snapshot);
static void SnapshotResetXmin(void);
/* ResourceOwner callbacks to track snapshot references */
@@ -587,7 +586,7 @@ CopySnapshot(Snapshot snapshot)
* FreeSnapshot
* Free the memory associated with a snapshot.
*/
-static void
+void
FreeSnapshot(Snapshot snapshot)
{
Assert(snapshot->regd_count == 0);
diff --git a/src/bin/psql/tab-complete.in.c b/src/bin/psql/tab-complete.in.c
index 81cbf10aa2..14c2058997 100644
--- a/src/bin/psql/tab-complete.in.c
+++ b/src/bin/psql/tab-complete.in.c
@@ -3119,7 +3119,7 @@ match_previous_words(int pattern_id,
* one word, so the above test is correct.
*/
if (ends_with(prev_wd, '(') || ends_with(prev_wd, ','))
- COMPLETE_WITH("VERBOSE");
+ COMPLETE_WITH("VERBOSE", "CONCURRENTLY");
}
/* COMMENT */
@@ -5133,7 +5133,8 @@ match_previous_words(int pattern_id,
"DISABLE_PAGE_SKIPPING", "SKIP_LOCKED",
"INDEX_CLEANUP", "PROCESS_MAIN", "PROCESS_TOAST",
"TRUNCATE", "PARALLEL", "SKIP_DATABASE_STATS",
- "ONLY_DATABASE_STATS", "BUFFER_USAGE_LIMIT");
+ "ONLY_DATABASE_STATS", "BUFFER_USAGE_LIMIT",
+ "CONCURRENTLY");
else if (TailMatches("FULL|FREEZE|ANALYZE|VERBOSE|DISABLE_PAGE_SKIPPING|SKIP_LOCKED|PROCESS_MAIN|PROCESS_TOAST|TRUNCATE|SKIP_DATABASE_STATS|ONLY_DATABASE_STATS"))
COMPLETE_WITH("ON", "OFF");
else if (TailMatches("INDEX_CLEANUP"))
diff --git a/src/include/access/heapam.h b/src/include/access/heapam.h
index 7d06dad83f..9b1fb15d8c 100644
--- a/src/include/access/heapam.h
+++ b/src/include/access/heapam.h
@@ -413,6 +413,10 @@ extern HTSV_Result HeapTupleSatisfiesVacuumHorizon(HeapTuple htup, Buffer buffer
TransactionId *dead_after);
extern void HeapTupleSetHintBits(HeapTupleHeader tuple, Buffer buffer,
uint16 infomask, TransactionId xid);
+extern bool HeapTupleMVCCInserted(HeapTuple htup, Snapshot snapshot,
+ Buffer buffer);
+extern bool HeapTupleMVCCNotDeleted(HeapTuple htup, Snapshot snapshot,
+ Buffer buffer);
extern bool HeapTupleHeaderIsOnlyLocked(HeapTupleHeader tuple);
extern bool HeapTupleIsSurelyDead(HeapTuple htup,
struct GlobalVisState *vistest);
diff --git a/src/include/access/tableam.h b/src/include/access/tableam.h
index 09b9b394e0..8efe1c23de 100644
--- a/src/include/access/tableam.h
+++ b/src/include/access/tableam.h
@@ -21,6 +21,7 @@
#include "access/sdir.h"
#include "access/xact.h"
#include "executor/tuptable.h"
+#include "replication/logical.h"
#include "storage/read_stream.h"
#include "utils/rel.h"
#include "utils/snapshot.h"
@@ -629,6 +630,8 @@ typedef struct TableAmRoutine
Relation OldIndex,
bool use_sort,
TransactionId OldestXmin,
+ Snapshot snapshot,
+ LogicalDecodingContext *decoding_ctx,
TransactionId *xid_cutoff,
MultiXactId *multi_cutoff,
double *num_tuples,
@@ -1672,6 +1675,10 @@ table_relation_copy_data(Relation rel, const RelFileLocator *newrlocator)
* not needed for the relation's AM
* - *xid_cutoff - ditto
* - *multi_cutoff - ditto
+ * - snapshot - if != NULL, ignore data changes done by transactions that this
+ * (MVCC) snapshot considers still in-progress or in the future.
+ * - decoding_ctx - logical decoding context, to capture concurrent data
+ * changes.
*
* Output parameters:
* - *xid_cutoff - rel's new relfrozenxid value, may be invalid
@@ -1684,6 +1691,8 @@ table_relation_copy_for_cluster(Relation OldTable, Relation NewTable,
Relation OldIndex,
bool use_sort,
TransactionId OldestXmin,
+ Snapshot snapshot,
+ LogicalDecodingContext *decoding_ctx,
TransactionId *xid_cutoff,
MultiXactId *multi_cutoff,
double *num_tuples,
@@ -1692,6 +1701,7 @@ table_relation_copy_for_cluster(Relation OldTable, Relation NewTable,
{
OldTable->rd_tableam->relation_copy_for_cluster(OldTable, NewTable, OldIndex,
use_sort, OldestXmin,
+ snapshot, decoding_ctx,
xid_cutoff, multi_cutoff,
num_tuples, tups_vacuumed,
tups_recently_dead);
diff --git a/src/include/catalog/index.h b/src/include/catalog/index.h
index 4daa8bef5e..66431cc19e 100644
--- a/src/include/catalog/index.h
+++ b/src/include/catalog/index.h
@@ -100,6 +100,9 @@ extern Oid index_concurrently_create_copy(Relation heapRelation,
Oid tablespaceOid,
const char *newName);
+extern NullableDatum *get_index_stattargets(Oid indexid,
+ IndexInfo *indInfo);
+
extern void index_concurrently_build(Oid heapRelationId,
Oid indexRelationId);
diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h
index 60088a64cb..d420930d6b 100644
--- a/src/include/commands/cluster.h
+++ b/src/include/commands/cluster.h
@@ -13,10 +13,15 @@
#ifndef CLUSTER_H
#define CLUSTER_H
+#include "nodes/execnodes.h"
#include "nodes/parsenodes.h"
#include "parser/parse_node.h"
+#include "replication/logical.h"
#include "storage/lock.h"
+#include "storage/relfilelocator.h"
#include "utils/relcache.h"
+#include "utils/resowner.h"
+#include "utils/tuplestore.h"
/* flag bits for ClusterParams->options */
@@ -24,6 +29,7 @@
#define CLUOPT_RECHECK 0x02 /* recheck relation state */
#define CLUOPT_RECHECK_ISCLUSTERED 0x04 /* recheck relation state for
* indisclustered */
+#define CLUOPT_CONCURRENT 0x08 /* allow concurrent data changes */
/* options for CLUSTER */
typedef struct ClusterParams
@@ -31,12 +37,91 @@ typedef struct ClusterParams
bits32 options; /* bitmask of CLUOPT_* */
} ClusterParams;
+/*
+ * The following definitions are used for concurrent processing.
+ */
+
+extern RelFileLocator clustered_rel_locator;
+extern RelFileLocator clustered_rel_toast_locator;
+
+typedef enum
+{
+ CHANGE_INSERT,
+ CHANGE_UPDATE_OLD,
+ CHANGE_UPDATE_NEW,
+ CHANGE_DELETE,
+ CHANGE_TRUNCATE
+} ConcurrentChangeKind;
+
+typedef struct ConcurrentChange
+{
+ /* See the enum above. */
+ ConcurrentChangeKind kind;
+
+ /*
+ * The actual tuple.
+ *
+ * The tuple data follows the ConcurrentChange structure. Before use make
+ * sure the tuple is correctly aligned (ConcurrentChange can be stored as
+ * bytea) and that tuple->t_data is fixed.
+ */
+ HeapTupleData tup_data;
+} ConcurrentChange;
+
+#define SizeOfConcurrentChange (offsetof(ConcurrentChange, tup_data) + \
+ sizeof(HeapTupleData))
+
+/*
+ * Logical decoding state.
+ *
+ * Here we store the data changes that we decode from WAL while the table
+ * contents is being copied to a new storage. Also the necessary metadata
+ * needed to apply these changes to the table is stored here.
+ */
+typedef struct ClusterDecodingState
+{
+ /* The relation whose changes we're decoding. */
+ Oid relid;
+
+ /*
+ * Decoded changes are stored here. Although we try to avoid excessive
+ * batches, it can happen that the changes need to be stored to disk. The
+ * tuplestore does this transparently.
+ */
+ Tuplestorestate *tstore;
+
+ /* The current number of changes in tstore. */
+ double nchanges;
+
+ /*
+ * Descriptor to store the ConcurrentChange structure serialized (bytea).
+ * We can't store the tuple directly because tuplestore only supports
+ * minimum tuple and we may need to transfer OID system column from the
+ * output plugin. Also we need to transfer the change kind, so it's better
+ * to put everything in the structure than to use 2 tuplestores "in
+ * parallel".
+ */
+ TupleDesc tupdesc_change;
+
+ /* Tuple descriptor needed to update indexes. */
+ TupleDesc tupdesc;
+
+ /* Slot to retrieve data from tstore. */
+ TupleTableSlot *tsslot;
+
+ ResourceOwner resowner;
+} ClusterDecodingState;
+
extern void cluster(ParseState *pstate, ClusterStmt *stmt, bool isTopLevel);
-extern void cluster_rel(Relation OldHeap, Oid indexOid, ClusterParams *params);
+extern void cluster_rel(Relation OldHeap, Oid indexOid, ClusterParams *params,
+ bool isTopLevel, bool isVacuum);
extern void check_index_is_clusterable(Relation OldHeap, Oid indexOid,
LOCKMODE lockmode);
extern void mark_index_clustered(Relation rel, Oid indexOid, bool is_internal);
-
+extern void check_relation_is_clusterable_concurrently(Relation rel,
+ bool is_vacuum);
+extern void cluster_decode_concurrent_changes(LogicalDecodingContext *ctx,
+ XLogRecPtr end_of_wal);
extern Oid make_new_heap(Oid OIDOldHeap, Oid NewTableSpace, Oid NewAccessMethod,
char relpersistence, LOCKMODE lockmode);
extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap,
@@ -44,8 +129,13 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap,
bool swap_toast_by_content,
bool check_constraints,
bool is_internal,
+ bool reindex,
TransactionId frozenXid,
MultiXactId cutoffMulti,
char newrelpersistence);
+extern Size ClusterShmemSize(void);
+extern void ClusterShmemInit(void);
+extern bool is_concurrent_cluster_in_progress(Oid relid);
+extern void check_for_concurrent_cluster(Oid relid, LOCKMODE lockmode);
#endif /* CLUSTER_H */
diff --git a/src/include/commands/progress.h b/src/include/commands/progress.h
index 18e3179ef6..e1d53f2a0a 100644
--- a/src/include/commands/progress.h
+++ b/src/include/commands/progress.h
@@ -59,19 +59,22 @@
#define PROGRESS_CLUSTER_PHASE 1
#define PROGRESS_CLUSTER_INDEX_RELID 2
#define PROGRESS_CLUSTER_HEAP_TUPLES_SCANNED 3
-#define PROGRESS_CLUSTER_HEAP_TUPLES_WRITTEN 4
-#define PROGRESS_CLUSTER_TOTAL_HEAP_BLKS 5
-#define PROGRESS_CLUSTER_HEAP_BLKS_SCANNED 6
-#define PROGRESS_CLUSTER_INDEX_REBUILD_COUNT 7
+#define PROGRESS_CLUSTER_HEAP_TUPLES_INSERTED 4
+#define PROGRESS_CLUSTER_HEAP_TUPLES_UPDATED 5
+#define PROGRESS_CLUSTER_HEAP_TUPLES_DELETED 6
+#define PROGRESS_CLUSTER_TOTAL_HEAP_BLKS 7
+#define PROGRESS_CLUSTER_HEAP_BLKS_SCANNED 8
+#define PROGRESS_CLUSTER_INDEX_REBUILD_COUNT 9
/* Phases of cluster (as advertised via PROGRESS_CLUSTER_PHASE) */
#define PROGRESS_CLUSTER_PHASE_SEQ_SCAN_HEAP 1
#define PROGRESS_CLUSTER_PHASE_INDEX_SCAN_HEAP 2
#define PROGRESS_CLUSTER_PHASE_SORT_TUPLES 3
#define PROGRESS_CLUSTER_PHASE_WRITE_NEW_HEAP 4
-#define PROGRESS_CLUSTER_PHASE_SWAP_REL_FILES 5
-#define PROGRESS_CLUSTER_PHASE_REBUILD_INDEX 6
-#define PROGRESS_CLUSTER_PHASE_FINAL_CLEANUP 7
+#define PROGRESS_CLUSTER_PHASE_CATCH_UP 5
+#define PROGRESS_CLUSTER_PHASE_SWAP_REL_FILES 6
+#define PROGRESS_CLUSTER_PHASE_REBUILD_INDEX 7
+#define PROGRESS_CLUSTER_PHASE_FINAL_CLEANUP 8
/* Commands of PROGRESS_CLUSTER */
#define PROGRESS_CLUSTER_COMMAND_CLUSTER 1
diff --git a/src/include/commands/vacuum.h b/src/include/commands/vacuum.h
index 12d0b61950..05c1fce969 100644
--- a/src/include/commands/vacuum.h
+++ b/src/include/commands/vacuum.h
@@ -181,13 +181,16 @@ typedef struct VacAttrStats
#define VACOPT_ANALYZE 0x02 /* do ANALYZE */
#define VACOPT_VERBOSE 0x04 /* output INFO instrumentation messages */
#define VACOPT_FREEZE 0x08 /* FREEZE option */
-#define VACOPT_FULL 0x10 /* FULL (non-concurrent) vacuum */
-#define VACOPT_SKIP_LOCKED 0x20 /* skip if cannot get lock */
-#define VACOPT_PROCESS_MAIN 0x40 /* process main relation */
-#define VACOPT_PROCESS_TOAST 0x80 /* process the TOAST table, if any */
-#define VACOPT_DISABLE_PAGE_SKIPPING 0x100 /* don't skip any pages */
-#define VACOPT_SKIP_DATABASE_STATS 0x200 /* skip vac_update_datfrozenxid() */
-#define VACOPT_ONLY_DATABASE_STATS 0x400 /* only vac_update_datfrozenxid() */
+#define VACOPT_FULL_EXCLUSIVE 0x10 /* FULL (non-concurrent) vacuum */
+#define VACOPT_FULL_CONCURRENT 0x20 /* FULL (concurrent) vacuum */
+#define VACOPT_SKIP_LOCKED 0x40 /* skip if cannot get lock */
+#define VACOPT_PROCESS_MAIN 0x80 /* process main relation */
+#define VACOPT_PROCESS_TOAST 0x100 /* process the TOAST table, if any */
+#define VACOPT_DISABLE_PAGE_SKIPPING 0x200 /* don't skip any pages */
+#define VACOPT_SKIP_DATABASE_STATS 0x400 /* skip vac_update_datfrozenxid() */
+#define VACOPT_ONLY_DATABASE_STATS 0x800 /* only vac_update_datfrozenxid() */
+
+#define VACOPT_FULL (VACOPT_FULL_EXCLUSIVE | VACOPT_FULL_CONCURRENT)
/*
* Values used by index_cleanup and truncate params.
diff --git a/src/include/replication/snapbuild.h b/src/include/replication/snapbuild.h
index 6d4d2d1814..7c2e1f2ceb 100644
--- a/src/include/replication/snapbuild.h
+++ b/src/include/replication/snapbuild.h
@@ -73,6 +73,7 @@ extern void FreeSnapshotBuilder(SnapBuild *builder);
extern void SnapBuildSnapDecRefcount(Snapshot snap);
extern Snapshot SnapBuildInitialSnapshot(SnapBuild *builder);
+extern Snapshot SnapBuildInitialSnapshotForCluster(SnapBuild *builder);
extern Snapshot SnapBuildMVCCFromHistoric(Snapshot snapshot, bool in_place);
extern const char *SnapBuildExportSnapshot(SnapBuild *builder);
extern void SnapBuildClearExportedSnapshot(void);
diff --git a/src/include/storage/lockdefs.h b/src/include/storage/lockdefs.h
index 7f3ba0352f..b7b94dacf8 100644
--- a/src/include/storage/lockdefs.h
+++ b/src/include/storage/lockdefs.h
@@ -36,8 +36,9 @@ typedef int LOCKMODE;
#define AccessShareLock 1 /* SELECT */
#define RowShareLock 2 /* SELECT FOR UPDATE/FOR SHARE */
#define RowExclusiveLock 3 /* INSERT, UPDATE, DELETE */
-#define ShareUpdateExclusiveLock 4 /* VACUUM (non-FULL), ANALYZE, CREATE
- * INDEX CONCURRENTLY */
+#define ShareUpdateExclusiveLock 4 /* VACUUM (non-exclusive), ANALYZE, CREATE
+ * INDEX CONCURRENTLY, CLUSTER
+ * CONCURRENTLY */
#define ShareLock 5 /* CREATE INDEX (WITHOUT CONCURRENTLY) */
#define ShareRowExclusiveLock 6 /* like EXCLUSIVE MODE, but allows ROW
* SHARE */
diff --git a/src/include/storage/lwlocklist.h b/src/include/storage/lwlocklist.h
index cf56545238..9211124d10 100644
--- a/src/include/storage/lwlocklist.h
+++ b/src/include/storage/lwlocklist.h
@@ -83,3 +83,4 @@ PG_LWLOCK(49, WALSummarizer)
PG_LWLOCK(50, DSMRegistry)
PG_LWLOCK(51, InjectionPoint)
PG_LWLOCK(52, SerialControl)
+PG_LWLOCK(54, ClusteredRels)
diff --git a/src/include/utils/backend_progress.h b/src/include/utils/backend_progress.h
index 739629cb21..5b9903de08 100644
--- a/src/include/utils/backend_progress.h
+++ b/src/include/utils/backend_progress.h
@@ -35,7 +35,7 @@ typedef enum ProgressCommandType
/*
* Any command which wishes can advertise that it is running by setting
- * command, command_target, and param[]. command_target should be the OID of
+ * ommand, command_target, and param[]. command_target should be the OID of
* the relation which the command targets (we assume there's just one, as this
* is meant for utility commands), but the meaning of each element in the
* param array is command-specific.
@@ -55,6 +55,7 @@ extern void pgstat_progress_parallel_incr_param(int index, int64 incr);
extern void pgstat_progress_update_multi_param(int nparam, const int *index,
const int64 *val);
extern void pgstat_progress_end_command(void);
+extern void pgstat_progress_restore_state(PgBackendProgress *backup);
#endif /* BACKEND_PROGRESS_H */
diff --git a/src/include/utils/inval.h b/src/include/utils/inval.h
index 40658ba2ff..6b2faed672 100644
--- a/src/include/utils/inval.h
+++ b/src/include/utils/inval.h
@@ -49,6 +49,8 @@ extern void CacheInvalidateCatalog(Oid catalogId);
extern void CacheInvalidateRelcache(Relation relation);
+extern void CacheInvalidateRelcacheImmediate(Relation relation);
+
extern void CacheInvalidateRelcacheAll(void);
extern void CacheInvalidateRelcacheByTuple(HeapTuple classTuple);
diff --git a/src/include/utils/rel.h b/src/include/utils/rel.h
index 33d1e4a4e2..243b80d7fa 100644
--- a/src/include/utils/rel.h
+++ b/src/include/utils/rel.h
@@ -253,6 +253,9 @@ typedef struct RelationData
bool pgstat_enabled; /* should relation stats be counted */
/* use "struct" here to avoid needing to include pgstat.h: */
struct PgStat_TableStatus *pgstat_info; /* statistics collection area */
+
+ /* Is CLUSTER CONCURRENTLY being performed on this relation? */
+ bool rd_cluster_concurrent;
} RelationData;
@@ -684,7 +687,9 @@ RelationCloseSmgr(Relation relation)
#define RelationIsAccessibleInLogicalDecoding(relation) \
(XLogLogicalInfoActive() && \
RelationNeedsWAL(relation) && \
- (IsCatalogRelation(relation) || RelationIsUsedAsCatalogTable(relation)))
+ (IsCatalogRelation(relation) || \
+ RelationIsUsedAsCatalogTable(relation) || \
+ (relation)->rd_cluster_concurrent))
/*
* RelationIsLogicallyLogged
diff --git a/src/include/utils/snapmgr.h b/src/include/utils/snapmgr.h
index 147b190210..5eeabdc6c4 100644
--- a/src/include/utils/snapmgr.h
+++ b/src/include/utils/snapmgr.h
@@ -61,6 +61,8 @@ extern Snapshot GetLatestSnapshot(void);
extern void SnapshotSetCommandId(CommandId curcid);
extern Snapshot CopySnapshot(Snapshot snapshot);
+extern void FreeSnapshot(Snapshot snapshot);
+
extern Snapshot GetCatalogSnapshot(Oid relid);
extern Snapshot GetNonHistoricCatalogSnapshot(Oid relid);
extern void InvalidateCatalogSnapshot(void);
diff --git a/src/test/regress/expected/rules.out b/src/test/regress/expected/rules.out
index 3014d047fe..81300642a5 100644
--- a/src/test/regress/expected/rules.out
+++ b/src/test/regress/expected/rules.out
@@ -1962,17 +1962,20 @@ pg_stat_progress_cluster| SELECT s.pid,
WHEN 2 THEN 'index scanning heap'::text
WHEN 3 THEN 'sorting tuples'::text
WHEN 4 THEN 'writing new heap'::text
- WHEN 5 THEN 'swapping relation files'::text
- WHEN 6 THEN 'rebuilding index'::text
- WHEN 7 THEN 'performing final cleanup'::text
+ WHEN 5 THEN 'catch-up'::text
+ WHEN 6 THEN 'swapping relation files'::text
+ WHEN 7 THEN 'rebuilding index'::text
+ WHEN 8 THEN 'performing final cleanup'::text
ELSE NULL::text
END AS phase,
(s.param3)::oid AS cluster_index_relid,
s.param4 AS heap_tuples_scanned,
- s.param5 AS heap_tuples_written,
- s.param6 AS heap_blks_total,
- s.param7 AS heap_blks_scanned,
- s.param8 AS index_rebuild_count
+ s.param5 AS heap_tuples_inserted,
+ s.param6 AS heap_tuples_updated,
+ s.param7 AS heap_tuples_deleted,
+ s.param8 AS heap_blks_total,
+ s.param9 AS heap_blks_scanned,
+ s.param10 AS index_rebuild_count
FROM (pg_stat_get_progress_info('CLUSTER'::text) s(pid, datid, relid, param1, param2, param3, param4, param5, param6, param7, param8, param9, param10, param11, param12, param13, param14, param15, param16, param17, param18, param19, param20)
LEFT JOIN pg_database d ON ((s.datid = d.oid)));
pg_stat_progress_copy| SELECT s.pid,
--
2.45.2
Attachments:
[text/x-diff] v07-0002-Move-progress-related-fields-from-PgBackendStatus-to.patch (6.6K, ../6250.1736776111@antos/2-v07-0002-Move-progress-related-fields-from-PgBackendStatus-to.patch)
download | inline diff:
From 8fc954ac4f917fa696391a16d75bafeb4f8b2ee0 Mon Sep 17 00:00:00 2001
From: Antonin Houska <[email protected]>
Date: Mon, 13 Jan 2025 14:29:54 +0100
Subject: [PATCH 2/8] Move progress related fields from PgBackendStatus to
PgBackendProgress.
VACUUM FULL / CLUSTER CONCURRENTLY will need to save and restore these fields
at some point. This is because plan_cluster_use_sort() has to be called in a
subtransaction (so that it does not leave any additional locks on the table)
and rollback of that subtransaction clears the progress information.
---
src/backend/utils/activity/backend_progress.c | 18 +++++++++---------
src/backend/utils/activity/backend_status.c | 4 ++--
src/backend/utils/adt/pgstatfuncs.c | 6 +++---
src/include/utils/backend_progress.h | 14 ++++++++++++++
src/include/utils/backend_status.h | 14 ++------------
5 files changed, 30 insertions(+), 26 deletions(-)
diff --git a/src/backend/utils/activity/backend_progress.c b/src/backend/utils/activity/backend_progress.c
index 99a8c73bf0..eebc968193 100644
--- a/src/backend/utils/activity/backend_progress.c
+++ b/src/backend/utils/activity/backend_progress.c
@@ -32,9 +32,9 @@ pgstat_progress_start_command(ProgressCommandType cmdtype, Oid relid)
return;
PGSTAT_BEGIN_WRITE_ACTIVITY(beentry);
- beentry->st_progress_command = cmdtype;
- beentry->st_progress_command_target = relid;
- MemSet(&beentry->st_progress_param, 0, sizeof(beentry->st_progress_param));
+ beentry->st_progress.command = cmdtype;
+ beentry->st_progress.command_target = relid;
+ MemSet(&beentry->st_progress.param, 0, sizeof(beentry->st_progress.param));
PGSTAT_END_WRITE_ACTIVITY(beentry);
}
@@ -55,7 +55,7 @@ pgstat_progress_update_param(int index, int64 val)
return;
PGSTAT_BEGIN_WRITE_ACTIVITY(beentry);
- beentry->st_progress_param[index] = val;
+ beentry->st_progress.param[index] = val;
PGSTAT_END_WRITE_ACTIVITY(beentry);
}
@@ -76,7 +76,7 @@ pgstat_progress_incr_param(int index, int64 incr)
return;
PGSTAT_BEGIN_WRITE_ACTIVITY(beentry);
- beentry->st_progress_param[index] += incr;
+ beentry->st_progress.param[index] += incr;
PGSTAT_END_WRITE_ACTIVITY(beentry);
}
@@ -133,7 +133,7 @@ pgstat_progress_update_multi_param(int nparam, const int *index,
{
Assert(index[i] >= 0 && index[i] < PGSTAT_NUM_PROGRESS_PARAM);
- beentry->st_progress_param[index[i]] = val[i];
+ beentry->st_progress.param[index[i]] = val[i];
}
PGSTAT_END_WRITE_ACTIVITY(beentry);
@@ -154,11 +154,11 @@ pgstat_progress_end_command(void)
if (!beentry || !pgstat_track_activities)
return;
- if (beentry->st_progress_command == PROGRESS_COMMAND_INVALID)
+ if (beentry->st_progress.command == PROGRESS_COMMAND_INVALID)
return;
PGSTAT_BEGIN_WRITE_ACTIVITY(beentry);
- beentry->st_progress_command = PROGRESS_COMMAND_INVALID;
- beentry->st_progress_command_target = InvalidOid;
+ beentry->st_progress.command = PROGRESS_COMMAND_INVALID;
+ beentry->st_progress.command_target = InvalidOid;
PGSTAT_END_WRITE_ACTIVITY(beentry);
}
diff --git a/src/backend/utils/activity/backend_status.c b/src/backend/utils/activity/backend_status.c
index 731342799a..b49cbf147d 100644
--- a/src/backend/utils/activity/backend_status.c
+++ b/src/backend/utils/activity/backend_status.c
@@ -376,8 +376,8 @@ pgstat_bestart(void)
#endif
lbeentry.st_state = STATE_UNDEFINED;
- lbeentry.st_progress_command = PROGRESS_COMMAND_INVALID;
- lbeentry.st_progress_command_target = InvalidOid;
+ lbeentry.st_progress.command = PROGRESS_COMMAND_INVALID;
+ lbeentry.st_progress.command_target = InvalidOid;
lbeentry.st_query_id = UINT64CONST(0);
/*
diff --git a/src/backend/utils/adt/pgstatfuncs.c b/src/backend/utils/adt/pgstatfuncs.c
index 5f8d20a406..25887f06e7 100644
--- a/src/backend/utils/adt/pgstatfuncs.c
+++ b/src/backend/utils/adt/pgstatfuncs.c
@@ -269,7 +269,7 @@ pg_stat_get_progress_info(PG_FUNCTION_ARGS)
* Report values for only those backends which are running the given
* command.
*/
- if (beentry->st_progress_command != cmdtype)
+ if (beentry->st_progress.command != cmdtype)
continue;
/* Value available to all callers */
@@ -279,9 +279,9 @@ pg_stat_get_progress_info(PG_FUNCTION_ARGS)
/* show rest of the values including relid only to role members */
if (HAS_PGSTAT_PERMISSIONS(beentry->st_userid))
{
- values[2] = ObjectIdGetDatum(beentry->st_progress_command_target);
+ values[2] = ObjectIdGetDatum(beentry->st_progress.command_target);
for (i = 0; i < PGSTAT_NUM_PROGRESS_PARAM; i++)
- values[i + 3] = Int64GetDatum(beentry->st_progress_param[i]);
+ values[i + 3] = Int64GetDatum(beentry->st_progress.param[i]);
}
else
{
diff --git a/src/include/utils/backend_progress.h b/src/include/utils/backend_progress.h
index dda813ab40..739629cb21 100644
--- a/src/include/utils/backend_progress.h
+++ b/src/include/utils/backend_progress.h
@@ -30,8 +30,22 @@ typedef enum ProgressCommandType
PROGRESS_COMMAND_COPY,
} ProgressCommandType;
+
#define PGSTAT_NUM_PROGRESS_PARAM 20
+/*
+ * Any command which wishes can advertise that it is running by setting
+ * command, command_target, and param[]. command_target should be the OID of
+ * the relation which the command targets (we assume there's just one, as this
+ * is meant for utility commands), but the meaning of each element in the
+ * param array is command-specific.
+ */
+typedef struct PgBackendProgress
+{
+ ProgressCommandType command;
+ Oid command_target;
+ int64 param[PGSTAT_NUM_PROGRESS_PARAM];
+} PgBackendProgress;
extern void pgstat_progress_start_command(ProgressCommandType cmdtype,
Oid relid);
diff --git a/src/include/utils/backend_status.h b/src/include/utils/backend_status.h
index d3d4ff6c5c..a73c76a442 100644
--- a/src/include/utils/backend_status.h
+++ b/src/include/utils/backend_status.h
@@ -155,18 +155,8 @@ typedef struct PgBackendStatus
*/
char *st_activity_raw;
- /*
- * Command progress reporting. Any command which wishes can advertise
- * that it is running by setting st_progress_command,
- * st_progress_command_target, and st_progress_param[].
- * st_progress_command_target should be the OID of the relation which the
- * command targets (we assume there's just one, as this is meant for
- * utility commands), but the meaning of each element in the
- * st_progress_param array is command-specific.
- */
- ProgressCommandType st_progress_command;
- Oid st_progress_command_target;
- int64 st_progress_param[PGSTAT_NUM_PROGRESS_PARAM];
+ /* Command progress reporting. */
+ PgBackendProgress st_progress;
/* query identifier, optionally computed using post_parse_analyze_hook */
uint64 st_query_id;
--
2.45.2
[text/x-diff] v07-0003-Move-conversion-of-a-historic-to-MVCC-snapshot-to-a-.patch (5.4K, ../6250.1736776111@antos/3-v07-0003-Move-conversion-of-a-historic-to-MVCC-snapshot-to-a-.patch)
download | inline diff:
From aa264f420d1d616f2e29aa0a966b77fa818b2ed1 Mon Sep 17 00:00:00 2001
From: Antonin Houska <[email protected]>
Date: Mon, 13 Jan 2025 14:29:54 +0100
Subject: [PATCH 3/8] Move conversion of a "historic" to MVCC snapshot to a
separate function.
The conversion is now handled by SnapBuildMVCCFromHistoric(). The VACUUM FULL
/ CLUSTER will also need it.
---
src/backend/replication/logical/snapbuild.c | 51 +++++++++++++++++----
src/backend/utils/time/snapmgr.c | 3 +-
src/include/replication/snapbuild.h | 1 +
src/include/utils/snapmgr.h | 1 +
4 files changed, 45 insertions(+), 11 deletions(-)
diff --git a/src/backend/replication/logical/snapbuild.c b/src/backend/replication/logical/snapbuild.c
index bbedd3de31..84bf0503a5 100644
--- a/src/backend/replication/logical/snapbuild.c
+++ b/src/backend/replication/logical/snapbuild.c
@@ -440,10 +440,7 @@ Snapshot
SnapBuildInitialSnapshot(SnapBuild *builder)
{
Snapshot snap;
- TransactionId xid;
TransactionId safeXid;
- TransactionId *newxip;
- int newxcnt = 0;
Assert(XactIsoLevel == XACT_REPEATABLE_READ);
Assert(builder->building_full_snapshot);
@@ -485,6 +482,31 @@ SnapBuildInitialSnapshot(SnapBuild *builder)
MyProc->xmin = snap->xmin;
+ /* Convert the historic snapshot to MVCC snapshot. */
+ return SnapBuildMVCCFromHistoric(snap, true);
+}
+
+/*
+ * Turn a historic MVCC snapshot into an ordinary MVCC snapshot.
+ *
+ * Unlike a regular (non-historic) MVCC snapshot, the xip array of this
+ * snapshot contains not only running main transactions, but also their
+ * subtransactions. This difference does has no impact on XidInMVCCSnapshot().
+ *
+ * Pass true for 'in_place' if you don't care about modifying the source
+ * snapshot. If you need a new instance, and one that was allocated as a
+ * single chunk of memory, pass false.
+ */
+Snapshot
+SnapBuildMVCCFromHistoric(Snapshot snapshot, bool in_place)
+{
+ TransactionId xid;
+ TransactionId *oldxip = snapshot->xip;
+ uint32 oldxcnt = snapshot->xcnt;
+ TransactionId *newxip;
+ int newxcnt = 0;
+ Snapshot result;
+
/* allocate in transaction context */
newxip = (TransactionId *)
palloc(sizeof(TransactionId) * GetMaxSnapshotXidCount());
@@ -495,7 +517,7 @@ SnapBuildInitialSnapshot(SnapBuild *builder)
* classical snapshot by marking all non-committed transactions as
* in-progress. This can be expensive.
*/
- for (xid = snap->xmin; NormalTransactionIdPrecedes(xid, snap->xmax);)
+ for (xid = snapshot->xmin; NormalTransactionIdPrecedes(xid, snapshot->xmax);)
{
void *test;
@@ -503,7 +525,7 @@ SnapBuildInitialSnapshot(SnapBuild *builder)
* Check whether transaction committed using the decoding snapshot
* meaning of ->xip.
*/
- test = bsearch(&xid, snap->xip, snap->xcnt,
+ test = bsearch(&xid, snapshot->xip, snapshot->xcnt,
sizeof(TransactionId), xidComparator);
if (test == NULL)
@@ -520,11 +542,22 @@ SnapBuildInitialSnapshot(SnapBuild *builder)
}
/* adjust remaining snapshot fields as needed */
- snap->snapshot_type = SNAPSHOT_MVCC;
- snap->xcnt = newxcnt;
- snap->xip = newxip;
+ snapshot->xcnt = newxcnt;
+ snapshot->xip = newxip;
+
+ if (in_place)
+ result = snapshot;
+ else
+ {
+ result = CopySnapshot(snapshot);
+
+ /* Restore the original values so the source is intact. */
+ snapshot->xip = oldxip;
+ snapshot->xcnt = oldxcnt;
+ }
+ result->snapshot_type = SNAPSHOT_MVCC;
- return snap;
+ return result;
}
/*
diff --git a/src/backend/utils/time/snapmgr.c b/src/backend/utils/time/snapmgr.c
index 8f1508b1ee..42bded373b 100644
--- a/src/backend/utils/time/snapmgr.c
+++ b/src/backend/utils/time/snapmgr.c
@@ -153,7 +153,6 @@ typedef struct ExportedSnapshot
static List *exportedSnapshots = NIL;
/* Prototypes for local functions */
-static Snapshot CopySnapshot(Snapshot snapshot);
static void UnregisterSnapshotNoOwner(Snapshot snapshot);
static void FreeSnapshot(Snapshot snapshot);
static void SnapshotResetXmin(void);
@@ -532,7 +531,7 @@ SetTransactionSnapshot(Snapshot sourcesnap, VirtualTransactionId *sourcevxid,
* The copy is palloc'd in TopTransactionContext and has initial refcounts set
* to 0. The returned snapshot has the copied flag set.
*/
-static Snapshot
+Snapshot
CopySnapshot(Snapshot snapshot)
{
Snapshot newsnap;
diff --git a/src/include/replication/snapbuild.h b/src/include/replication/snapbuild.h
index 44031dcf6e..6d4d2d1814 100644
--- a/src/include/replication/snapbuild.h
+++ b/src/include/replication/snapbuild.h
@@ -73,6 +73,7 @@ extern void FreeSnapshotBuilder(SnapBuild *builder);
extern void SnapBuildSnapDecRefcount(Snapshot snap);
extern Snapshot SnapBuildInitialSnapshot(SnapBuild *builder);
+extern Snapshot SnapBuildMVCCFromHistoric(Snapshot snapshot, bool in_place);
extern const char *SnapBuildExportSnapshot(SnapBuild *builder);
extern void SnapBuildClearExportedSnapshot(void);
extern void SnapBuildResetExportedSnapshotState(void);
diff --git a/src/include/utils/snapmgr.h b/src/include/utils/snapmgr.h
index d346be7164..147b190210 100644
--- a/src/include/utils/snapmgr.h
+++ b/src/include/utils/snapmgr.h
@@ -60,6 +60,7 @@ extern Snapshot GetTransactionSnapshot(void);
extern Snapshot GetLatestSnapshot(void);
extern void SnapshotSetCommandId(CommandId curcid);
+extern Snapshot CopySnapshot(Snapshot snapshot);
extern Snapshot GetCatalogSnapshot(Oid relid);
extern Snapshot GetNonHistoricCatalogSnapshot(Oid relid);
extern void InvalidateCatalogSnapshot(void);
--
2.45.2
[text/plain] v07-0004-Add-CONCURRENTLY-option-to-both-VACUUM-FULL-and-CLUS.patch (171.4K, ../6250.1736776111@antos/4-v07-0004-Add-CONCURRENTLY-option-to-both-VACUUM-FULL-and-CLUS.patch)
download | inline diff:
From bf2ec8c5d753de340140839f1b061044ec4c1149 Mon Sep 17 00:00:00 2001
From: Antonin Houska <[email protected]>
Date: Mon, 13 Jan 2025 14:29:54 +0100
Subject: [PATCH 4/8] Add CONCURRENTLY option to both VACUUM FULL and CLUSTER
commands.
Both VACUUM FULL and CLUSTER commands copy the relation data into a new file,
create new indexes and eventually swap the files. To make sure that the old
file does not change during the copying, the relation is locked in an
exclusive mode, which prevents applications from both reading and writing. (To
keep the data consistent, we'd only need to prevent the applications from
writing, but even reading needs to be blocked before we can swap the files -
otherwise some applications could continue using the old file. Since we cannot
get stronger lock without releasing the weaker one first, we acquire the
exclusive lock in the beginning and keep it till the end of the processing.)
This patch introduces an alternative workflow, which only requires the
exclusive lock when the relation (and index) files are being swapped.
(Supposedly, the swapping should be pretty fast.) On the other hand, when we
copy the data to the new file, we allow applications to read from the relation
and even write into it.
First, we scan the relation using a "historic snapshot", and insert all the
tuples satisfying this snapshot into the new file. Note that, before creating
that snapshot, we need to make sure that all the other backends treat the
relation as a system catalog: in particular, they must log information on new
command IDs (CIDs). We achieve that by adding the relation ID into a shared
hash table and waiting until all the transactions currently writing into the
table (i.e. transactions possibly not aware of the new entry) have finished.
Second, logical decoding is used to capture the data changes done by
applications during the copying (i.e. changes that do not satisfy the historic
snapshot mentioned above), and those are applied to the new file before we
acquire the exclusive lock we need to swap the files. (Of course, more data
changes can take place while we are waiting for the lock - these will be
applied to the new file after we have acquired the lock, before we swap the
files.)
While copying the data into the new file, we hold a lock that prevents
applications from changing the relation tuple descriptor (tuples inserted into
the old file must fit into the new file). However, as we have to release that
lock before getting the exclusive one, it's possible that someone adds or
drops a column, or changes the data type of an existing one. Therefore we have
to check the tuple descriptor before we swap the files. If we find out that
the tuple descriptor changed, ERROR is raised and all the changes are rolled
back. Since a lot of effort can be wasted in such a case, the ALTER TABLE
command also tries to check if VACUUM FULL / CLUSTER with the CONCURRENTLY
option is running on the same relation, and raises an ERROR if it is.
Like the existing implementation of both VACUUM FULL and CLUSTER commands, the
variant with the CONCURRENTLY option also requires an extra space for the new
relation and index files (which coexist with the old files for some time). In
addition, the CONCURRENTLY option might introduce a lag in releasing WAL
segments for archiving / recycling. This is due to the decoding of the data
changes done by application concurrently. However, this lag should not be more
than a single WAL segment.
---
doc/src/sgml/monitoring.sgml | 36 +-
doc/src/sgml/ref/cluster.sgml | 116 +-
doc/src/sgml/ref/vacuum.sgml | 22 +-
src/Makefile | 1 +
src/backend/access/heap/heapam.c | 8 +-
src/backend/access/heap/heapam_handler.c | 145 +-
src/backend/access/heap/heapam_visibility.c | 30 +-
src/backend/catalog/index.c | 43 +-
src/backend/catalog/system_views.sql | 17 +-
src/backend/commands/cluster.c | 2585 ++++++++++++++++-
src/backend/commands/matview.c | 2 +-
src/backend/commands/tablecmds.c | 11 +
src/backend/commands/vacuum.c | 126 +-
src/backend/meson.build | 1 +
src/backend/replication/logical/decode.c | 24 +
src/backend/replication/logical/snapbuild.c | 20 +
.../replication/pgoutput_cluster/Makefile | 32 +
.../replication/pgoutput_cluster/meson.build | 18 +
.../pgoutput_cluster/pgoutput_cluster.c | 288 ++
src/backend/storage/ipc/ipci.c | 3 +
src/backend/tcop/utility.c | 11 +
src/backend/utils/activity/backend_progress.c | 16 +
.../utils/activity/wait_event_names.txt | 1 +
src/backend/utils/cache/inval.c | 22 +
src/backend/utils/cache/relcache.c | 5 +
src/backend/utils/time/snapmgr.c | 3 +-
src/bin/psql/tab-complete.in.c | 5 +-
src/include/access/heapam.h | 4 +
src/include/access/tableam.h | 10 +
src/include/catalog/index.h | 3 +
src/include/commands/cluster.h | 94 +-
src/include/commands/progress.h | 17 +-
src/include/commands/vacuum.h | 17 +-
src/include/replication/snapbuild.h | 1 +
src/include/storage/lockdefs.h | 5 +-
src/include/storage/lwlocklist.h | 1 +
src/include/utils/backend_progress.h | 3 +-
src/include/utils/inval.h | 2 +
src/include/utils/rel.h | 7 +-
src/include/utils/snapmgr.h | 2 +
src/test/regress/expected/rules.out | 17 +-
41 files changed, 3572 insertions(+), 202 deletions(-)
create mode 100644 src/backend/replication/pgoutput_cluster/Makefile
create mode 100644 src/backend/replication/pgoutput_cluster/meson.build
create mode 100644 src/backend/replication/pgoutput_cluster/pgoutput_cluster.c
diff --git a/doc/src/sgml/monitoring.sgml b/doc/src/sgml/monitoring.sgml
index d0d176cc54..985b20a81e 100644
--- a/doc/src/sgml/monitoring.sgml
+++ b/doc/src/sgml/monitoring.sgml
@@ -5727,14 +5727,35 @@ FROM pg_stat_get_backend_idset() AS backendid;
<row>
<entry role="catalog_table_entry"><para role="column_definition">
- <structfield>heap_tuples_written</structfield> <type>bigint</type>
+ <structfield>heap_tuples_inserted</structfield> <type>bigint</type>
</para>
<para>
- Number of heap tuples written.
+ Number of heap tuples inserted.
This counter only advances when the phase is
<literal>seq scanning heap</literal>,
- <literal>index scanning heap</literal>
- or <literal>writing new heap</literal>.
+ <literal>index scanning heap</literal>,
+ <literal>writing new heap</literal>
+ or <literal>catch-up</literal>.
+ </para></entry>
+ </row>
+
+ <row>
+ <entry role="catalog_table_entry"><para role="column_definition">
+ <structfield>heap_tuples_updated</structfield> <type>bigint</type>
+ </para>
+ <para>
+ Number of heap tuples updated.
+ This counter only advances when the phase is <literal>catch-up</literal>.
+ </para></entry>
+ </row>
+
+ <row>
+ <entry role="catalog_table_entry"><para role="column_definition">
+ <structfield>heap_tuples_deleted</structfield> <type>bigint</type>
+ </para>
+ <para>
+ Number of heap tuples deleted.
+ This counter only advances when the phase is <literal>catch-up</literal>.
</para></entry>
</row>
@@ -5815,6 +5836,13 @@ FROM pg_stat_get_backend_idset() AS backendid;
<command>CLUSTER</command> is currently writing the new heap.
</entry>
</row>
+ <row>
+ <entry><literal>catch-up</literal></entry>
+ <entry>
+ <command>CLUSTER</command> is currently processing the DML commands
+ that other transactions executed during any of the preceding phase.
+ </entry>
+ </row>
<row>
<entry><literal>swapping relation files</literal></entry>
<entry>
diff --git a/doc/src/sgml/ref/cluster.sgml b/doc/src/sgml/ref/cluster.sgml
index 8811f169ea..356b40e3fe 100644
--- a/doc/src/sgml/ref/cluster.sgml
+++ b/doc/src/sgml/ref/cluster.sgml
@@ -26,6 +26,7 @@ CLUSTER [ ( <replaceable class="parameter">option</replaceable> [, ...] ) ] [ <r
<phrase>where <replaceable class="parameter">option</replaceable> can be one of:</phrase>
VERBOSE [ <replaceable class="parameter">boolean</replaceable> ]
+ CONCURRENTLY [ <replaceable class="parameter">boolean</replaceable> ]
</synopsis>
</refsynopsisdiv>
@@ -69,14 +70,17 @@ CLUSTER [ ( <replaceable class="parameter">option</replaceable> [, ...] ) ] [ <r
<replaceable class="parameter">table_name</replaceable> reclusters all the
previously-clustered tables in the current database that the calling user
has privileges for. This form of <command>CLUSTER</command> cannot be
- executed inside a transaction block.
+ executed inside a transaction block. Also, this form is not allowed if
+ the <literal>CONCURRENTLY</literal> option is used.
</para>
<para>
- When a table is being clustered, an <literal>ACCESS
- EXCLUSIVE</literal> lock is acquired on it. This prevents any other
- database operations (both reads and writes) from operating on the
- table until the <command>CLUSTER</command> is finished.
+ When a table is being clustered, an <literal>ACCESS EXCLUSIVE</literal>
+ lock is acquired on it. This prevents any other database operations (both
+ reads and writes) from operating on the table until
+ the <command>CLUSTER</command> is finished. If you want to keep the table
+ accessible during the clustering, consider using
+ the <literal>CONCURRENTLY</literal> option.
</para>
</refsect1>
@@ -112,6 +116,108 @@ CLUSTER [ ( <replaceable class="parameter">option</replaceable> [, ...] ) ] [ <r
</listitem>
</varlistentry>
+ <varlistentry>
+ <term><literal>CONCURRENTLY</literal></term>
+ <listitem>
+ <para>
+ Allow other transactions to use the table while it is being clustered.
+ </para>
+
+ <para>
+ Internally, <command>CLUSTER</command> copies the contents of the table
+ (ignoring dead tuples) into a new file, sorted by the specified index,
+ and also creates a new file for each index. Then it swaps the old and
+ new files for the table and all the indexes, and deletes the old
+ files. The <literal>ACCESS EXCLUSIVE</literal> lock is needed to make
+ sure that the old files do not change during the processing because the
+ changes would get lost due to the swap.
+ </para>
+
+ <para>
+ With the <literal>CONCURRENTLY</literal> option, the <literal>ACCESS
+ EXCLUSIVE</literal> lock is only acquired to swap the table and index
+ files. The data changes that took place during the creation of the new
+ table and index files are captured using logical decoding
+ (<xref linkend="logicaldecoding"/>) and applied before
+ the <literal>ACCESS EXCLUSIVE</literal> lock is requested. Thus the lock
+ is typically held only for the time needed to swap the files, which
+ should be pretty short.
+ </para>
+
+ <para>
+ Note that <command>CLUSTER</command> with the
+ the <literal>CONCURRENTLY</literal> option does not try to order the
+ rows inserted into the table after the clustering started. Also
+ note <command>CLUSTER</command> might fail to complete due to DDL
+ commands executed on the table by other transactions during the
+ clustering.
+ </para>
+
+ <note>
+ <para>
+ In addition to the temporary space requirements explained below,
+ the <literal>CONCURRENTLY</literal> option can add to the usage of
+ temporary space a bit more. The reason is that other transactions can
+ perform DML operations which cannot be applied to the new file until
+ <command>CLUSTER</command> has copied all the tuples from the old
+ file. Thus the tuples inserted into the old file during the copying are
+ also stored in separately in a temporary file, so they can eventually
+ be applied to the new file.
+ </para>
+
+ <para>
+ Furthermore, the data changes performed during the copying are
+ extracted from <link linkend="wal">write-ahead log</link> (WAL), and
+ this extraction (decoding) only takes place when certain amount of WAL
+ has been written. Therefore, WAL removal can be delayed by this
+ threshold. Currently the threshold is equal to the value of
+ the <link linkend="guc-wal-segment-size"><varname>wal_segment_size</varname></link>
+ configuration parameter.
+ </para>
+ </note>
+
+ <para>
+ The <literal>CONCURRENTLY</literal> option cannot be used in the
+ following cases:
+
+ <itemizedlist>
+ <listitem>
+ <para>
+ The table is partitioned.
+ </para>
+ </listitem>
+
+ <listitem>
+ <para>
+ The table is a system catalog or a <acronym>TOAST</acronym> table.
+ </para>
+ </listitem>
+
+ <listitem>
+ <para>
+ <command>CLUSTER</command> is executed inside a transaction block.
+ </para>
+ </listitem>
+
+ <listitem>
+ <para>
+ The <link linkend="guc-wal-level"><varname>wal_level</varname></link>
+ configuration parameter is less than <literal>logical</literal>.
+ </para>
+ </listitem>
+
+ <listitem>
+ <para>
+ The <link linkend="guc-max-replication-slots"><varname>max_replication_slots</varname></link>
+ configuration parameter does not allow for creation of an additional
+ replication slot.
+ </para>
+ </listitem>
+ </itemizedlist>
+ </para>
+ </listitem>
+ </varlistentry>
+
<varlistentry>
<term><replaceable class="parameter">boolean</replaceable></term>
<listitem>
diff --git a/doc/src/sgml/ref/vacuum.sgml b/doc/src/sgml/ref/vacuum.sgml
index 971b1237d4..ba8669026a 100644
--- a/doc/src/sgml/ref/vacuum.sgml
+++ b/doc/src/sgml/ref/vacuum.sgml
@@ -39,6 +39,7 @@ VACUUM [ ( <replaceable class="parameter">option</replaceable> [, ...] ) ] [ <re
SKIP_DATABASE_STATS [ <replaceable class="parameter">boolean</replaceable> ]
ONLY_DATABASE_STATS [ <replaceable class="parameter">boolean</replaceable> ]
BUFFER_USAGE_LIMIT <replaceable class="parameter">size</replaceable>
+ CONCURRENTLY [ <replaceable class="parameter">boolean</replaceable> ]
<phrase>and <replaceable class="parameter">table_and_columns</replaceable> is:</phrase>
@@ -62,7 +63,8 @@ VACUUM [ ( <replaceable class="parameter">option</replaceable> [, ...] ) ] [ <re
Without a <replaceable class="parameter">table_and_columns</replaceable>
list, <command>VACUUM</command> processes every table and materialized view
in the current database that the current user has permission to vacuum.
- With a list, <command>VACUUM</command> processes only those table(s).
+ With a list, <command>VACUUM</command> processes only those table(s). The
+ list is required if the <literal>CONCURRENTLY</literal> option is used.
</para>
<para>
@@ -361,6 +363,24 @@ VACUUM [ ( <replaceable class="parameter">option</replaceable> [, ...] ) ] [ <re
</listitem>
</varlistentry>
+ <varlistentry>
+ <term><literal>CONCURRENTLY</literal></term>
+ <listitem>
+ <para>
+ Allow other transactions to use the table while it is being vacuumed. If
+ this option is specified, <command>VACUUM</command> can only process
+ tables which have already been clustered. For more information, see the
+ description of the <literal>CONCURRENTLY</literal> option of the
+ <xref linkend="sql-cluster"/> command.
+ </para>
+
+ <para>
+ The <literal>CONCURRENTLY</literal> option can only be used
+ if <literal>FULL</literal> is used at the same time.
+ </para>
+ </listitem>
+ </varlistentry>
+
<varlistentry>
<term><replaceable class="parameter">boolean</replaceable></term>
<listitem>
diff --git a/src/Makefile b/src/Makefile
index 2f31a2f20a..8b9d30ff72 100644
--- a/src/Makefile
+++ b/src/Makefile
@@ -23,6 +23,7 @@ SUBDIRS = \
interfaces \
backend/replication/libpqwalreceiver \
backend/replication/pgoutput \
+ backend/replication/pgoutput_cluster \
fe_utils \
bin \
pl \
diff --git a/src/backend/access/heap/heapam.c b/src/backend/access/heap/heapam.c
index 485525f4d6..552993d4ef 100644
--- a/src/backend/access/heap/heapam.c
+++ b/src/backend/access/heap/heapam.c
@@ -2073,8 +2073,14 @@ heap_insert(Relation relation, HeapTuple tup, CommandId cid,
/*
* If this is a catalog, we need to transmit combo CIDs to properly
* decode, so log that as well.
+ *
+ * For the main heap (as opposed to TOAST), we only receive
+ * HEAP_INSERT_NO_LOGICAL when doing VACUUM FULL / CLUSTER, in which
+ * case the visibility information does not change. Therefore, there's
+ * no need to update the decoding snapshot.
*/
- if (RelationIsAccessibleInLogicalDecoding(relation))
+ if ((options & HEAP_INSERT_NO_LOGICAL) == 0 &&
+ RelationIsAccessibleInLogicalDecoding(relation))
log_heap_new_cid(relation, heaptup);
/*
diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c
index e817f8f8f8..c5ec21ca2f 100644
--- a/src/backend/access/heap/heapam_handler.c
+++ b/src/backend/access/heap/heapam_handler.c
@@ -33,6 +33,7 @@
#include "catalog/index.h"
#include "catalog/storage.h"
#include "catalog/storage_xlog.h"
+#include "commands/cluster.h"
#include "commands/progress.h"
#include "executor/executor.h"
#include "miscadmin.h"
@@ -53,6 +54,9 @@ static void reform_and_rewrite_tuple(HeapTuple tuple,
static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer,
HeapTuple tuple,
OffsetNumber tupoffset);
+static HeapTuple accept_tuple_for_concurrent_copy(HeapTuple tuple,
+ Snapshot snapshot,
+ Buffer buffer);
static BlockNumber heapam_scan_get_blocks_done(HeapScanDesc hscan);
@@ -681,6 +685,8 @@ static void
heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap,
Relation OldIndex, bool use_sort,
TransactionId OldestXmin,
+ Snapshot snapshot,
+ LogicalDecodingContext *decoding_ctx,
TransactionId *xid_cutoff,
MultiXactId *multi_cutoff,
double *num_tuples,
@@ -701,6 +707,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap,
bool *isnull;
BufferHeapTupleTableSlot *hslot;
BlockNumber prev_cblock = InvalidBlockNumber;
+ bool concurrent = snapshot != NULL;
+ XLogRecPtr end_of_wal_prev = GetFlushRecPtr(NULL);
/* Remember if it's a system catalog */
is_system_catalog = IsSystemRelation(OldHeap);
@@ -779,8 +787,10 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap,
for (;;)
{
HeapTuple tuple;
+ bool tuple_copied = false;
Buffer buf;
bool isdead;
+ HTSV_Result vis;
CHECK_FOR_INTERRUPTS();
@@ -835,7 +845,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap,
LockBuffer(buf, BUFFER_LOCK_SHARE);
- switch (HeapTupleSatisfiesVacuum(tuple, OldestXmin, buf))
+ switch ((vis = HeapTupleSatisfiesVacuum(tuple, OldestXmin, buf)))
{
case HEAPTUPLE_DEAD:
/* Definitely dead */
@@ -851,14 +861,15 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap,
case HEAPTUPLE_INSERT_IN_PROGRESS:
/*
- * Since we hold exclusive lock on the relation, normally the
- * only way to see this is if it was inserted earlier in our
- * own transaction. However, it can happen in system
+ * As long as we hold exclusive lock on the relation, normally
+ * the only way to see this is if it was inserted earlier in
+ * our own transaction. However, it can happen in system
* catalogs, since we tend to release write lock before commit
- * there. Give a warning if neither case applies; but in any
- * case we had better copy it.
+ * there. Also, there's no exclusive lock during concurrent
+ * processing. Give a warning if neither case applies; but in
+ * any case we had better copy it.
*/
- if (!is_system_catalog &&
+ if (!is_system_catalog && !concurrent &&
!TransactionIdIsCurrentTransactionId(HeapTupleHeaderGetXmin(tuple->t_data)))
elog(WARNING, "concurrent insert in progress within table \"%s\"",
RelationGetRelationName(OldHeap));
@@ -870,7 +881,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap,
/*
* Similar situation to INSERT_IN_PROGRESS case.
*/
- if (!is_system_catalog &&
+ if (!is_system_catalog && !concurrent &&
!TransactionIdIsCurrentTransactionId(HeapTupleHeaderGetUpdateXid(tuple->t_data)))
elog(WARNING, "concurrent delete in progress within table \"%s\"",
RelationGetRelationName(OldHeap));
@@ -884,8 +895,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap,
break;
}
- LockBuffer(buf, BUFFER_LOCK_UNLOCK);
-
if (isdead)
{
*tups_vacuumed += 1;
@@ -896,9 +905,47 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap,
*tups_vacuumed += 1;
*tups_recently_dead -= 1;
}
+
+ LockBuffer(buf, BUFFER_LOCK_UNLOCK);
continue;
}
+ if (concurrent)
+ {
+ /*
+ * Ignore concurrent changes now, they'll be processed later via
+ * logical decoding.
+ *
+ * INSERT_IN_PROGRESS is rejected right away because our snapshot
+ * should represent a point in time which should precede (or be
+ * equal to) the state of transactions as it was when the
+ * "SatisfiesVacuum" test was performed. Thus
+ * accept_tuple_for_concurrent_copy() should not consider the
+ * tuple inserted.
+ */
+ if (vis == HEAPTUPLE_INSERT_IN_PROGRESS)
+ tuple = NULL;
+ else
+ tuple = accept_tuple_for_concurrent_copy(tuple, snapshot,
+ buf);
+ /* Tuple not suitable for the new heap? */
+ if (tuple == NULL)
+ {
+ LockBuffer(buf, BUFFER_LOCK_UNLOCK);
+ continue;
+ }
+
+ /* Remember that we have to free the tuple eventually. */
+ tuple_copied = true;
+ }
+
+ /*
+ * In the concurrent case, we have a copy of the tuple, so we don't
+ * worry whether the source tuple will be deleted / updated after we
+ * release the lock.
+ */
+ LockBuffer(buf, BUFFER_LOCK_UNLOCK);
+
*num_tuples += 1;
if (tuplesort != NULL)
{
@@ -915,7 +962,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap,
{
const int ct_index[] = {
PROGRESS_CLUSTER_HEAP_TUPLES_SCANNED,
- PROGRESS_CLUSTER_HEAP_TUPLES_WRITTEN
+ PROGRESS_CLUSTER_HEAP_TUPLES_INSERTED
};
int64 ct_val[2];
@@ -930,6 +977,33 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap,
ct_val[1] = *num_tuples;
pgstat_progress_update_multi_param(2, ct_index, ct_val);
}
+ if (tuple_copied)
+ heap_freetuple(tuple);
+
+ /*
+ * Process the WAL produced by the load, as well as by other
+ * transactions, so that the replication slot can advance and WAL does
+ * not pile up. Use wal_segment_size as a threshold so that we do not
+ * introduce the decoding overhead too often.
+ *
+ * Of course, we must not apply the changes until the initial load has
+ * completed.
+ *
+ * Note that our insertions into the new table should not be decoded
+ * as we (intentionally) do not write the logical decoding specific
+ * information to WAL.
+ */
+ if (concurrent)
+ {
+ XLogRecPtr end_of_wal;
+
+ end_of_wal = GetFlushRecPtr(NULL);
+ if ((end_of_wal - end_of_wal_prev) > wal_segment_size)
+ {
+ cluster_decode_concurrent_changes(decoding_ctx, end_of_wal);
+ end_of_wal_prev = end_of_wal;
+ }
+ }
}
if (indexScan != NULL)
@@ -973,7 +1047,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap,
values, isnull,
rwstate);
/* Report n_tuples */
- pgstat_progress_update_param(PROGRESS_CLUSTER_HEAP_TUPLES_WRITTEN,
+ pgstat_progress_update_param(PROGRESS_CLUSTER_HEAP_TUPLES_INSERTED,
n_tuples);
}
@@ -2609,6 +2683,53 @@ SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer,
}
}
+/*
+ * Return copy of 'tuple' if it has been inserted according to 'snapshot', or
+ * NULL if the insertion took place in the future. If the tuple is already
+ * marked as deleted or updated by a transaction that 'snapshot' still
+ * considers running, clear the deletion / update XID in the header of the
+ * copied tuple. This way the returned tuple is suitable for insertion into
+ * the new heap.
+ */
+static HeapTuple
+accept_tuple_for_concurrent_copy(HeapTuple tuple, Snapshot snapshot,
+ Buffer buffer)
+{
+ HeapTuple result;
+
+ Assert(snapshot->snapshot_type == SNAPSHOT_MVCC);
+
+ /*
+ * First, check if the tuple insertion is visible by our snapshot.
+ */
+ if (!HeapTupleMVCCInserted(tuple, snapshot, buffer))
+ return NULL;
+
+ result = heap_copytuple(tuple);
+
+ /*
+ * If the tuple was deleted / updated but our snapshot still sees it, we
+ * need to keep it. In that case, clear the information that indicates the
+ * deletion / update. Otherwise the tuple chain would stay incomplete (as
+ * we will reject the new tuple above), and the delete / update would fail
+ * if executed later during logical decoding.
+ */
+ if (TransactionIdIsNormal(HeapTupleHeaderGetRawXmax(result->t_data)) &&
+ HeapTupleMVCCNotDeleted(result, snapshot, buffer))
+ {
+ /* TODO More work needed here?*/
+ result->t_data->t_infomask |= HEAP_XMAX_INVALID;
+ HeapTupleHeaderSetXmax(result->t_data, 0);
+ }
+
+ /*
+ * Accept the tuple even if our snapshot considers it deleted - older
+ * snapshots can still see the tuple, while the decoded transactions
+ * should not try to update / delete it again.
+ */
+ return result;
+}
+
/* ------------------------------------------------------------------------
* Definition of the heap table access method.
diff --git a/src/backend/access/heap/heapam_visibility.c b/src/backend/access/heap/heapam_visibility.c
index e146605bd5..d9be93aadc 100644
--- a/src/backend/access/heap/heapam_visibility.c
+++ b/src/backend/access/heap/heapam_visibility.c
@@ -955,16 +955,31 @@ HeapTupleSatisfiesDirty(HeapTuple htup, Snapshot snapshot,
* did TransactionIdIsInProgress in each call --- to no avail, as long as the
* inserting/deleting transaction was still running --- which was more cycles
* and more contention on ProcArrayLock.
+ *
+ * The checks are split into two functions, HeapTupleMVCCInserted() and
+ * HeapTupleMVCCNotDeleted(), because they are also useful separately.
*/
static bool
HeapTupleSatisfiesMVCC(HeapTuple htup, Snapshot snapshot,
Buffer buffer)
{
- HeapTupleHeader tuple = htup->t_data;
-
Assert(ItemPointerIsValid(&htup->t_self));
Assert(htup->t_tableOid != InvalidOid);
+ return HeapTupleMVCCInserted(htup, snapshot, buffer) &&
+ HeapTupleMVCCNotDeleted(htup, snapshot, buffer);
+}
+
+/*
+ * HeapTupleMVCCInserted
+ * True iff heap tuple was successfully inserted for the given MVCC
+ * snapshot.
+ */
+bool
+HeapTupleMVCCInserted(HeapTuple htup, Snapshot snapshot, Buffer buffer)
+{
+ HeapTupleHeader tuple = htup->t_data;
+
if (!HeapTupleHeaderXminCommitted(tuple))
{
if (HeapTupleHeaderXminInvalid(tuple))
@@ -1073,6 +1088,17 @@ HeapTupleSatisfiesMVCC(HeapTuple htup, Snapshot snapshot,
}
/* by here, the inserting transaction has committed */
+ return true;
+}
+
+/*
+ * HeapTupleMVCCNotDeleted
+ * True iff heap tuple was not deleted for the given MVCC snapshot.
+ */
+bool
+HeapTupleMVCCNotDeleted(HeapTuple htup, Snapshot snapshot, Buffer buffer)
+{
+ HeapTupleHeader tuple = htup->t_data;
if (tuple->t_infomask & HEAP_XMAX_INVALID) /* xid invalid or aborted */
return true;
diff --git a/src/backend/catalog/index.c b/src/backend/catalog/index.c
index 7377912b41..1f9aa5bf8f 100644
--- a/src/backend/catalog/index.c
+++ b/src/backend/catalog/index.c
@@ -1417,22 +1417,7 @@ index_concurrently_create_copy(Relation heapRelation, Oid oldIndexId,
for (int i = 0; i < newInfo->ii_NumIndexAttrs; i++)
opclassOptions[i] = get_attoptions(oldIndexId, i + 1);
- /* Extract statistic targets for each attribute */
- stattargets = palloc0_array(NullableDatum, newInfo->ii_NumIndexAttrs);
- for (int i = 0; i < newInfo->ii_NumIndexAttrs; i++)
- {
- HeapTuple tp;
- Datum dat;
-
- tp = SearchSysCache2(ATTNUM, ObjectIdGetDatum(oldIndexId), Int16GetDatum(i + 1));
- if (!HeapTupleIsValid(tp))
- elog(ERROR, "cache lookup failed for attribute %d of relation %u",
- i + 1, oldIndexId);
- dat = SysCacheGetAttr(ATTNUM, tp, Anum_pg_attribute_attstattarget, &isnull);
- ReleaseSysCache(tp);
- stattargets[i].value = dat;
- stattargets[i].isnull = isnull;
- }
+ stattargets = get_index_stattargets(oldIndexId, newInfo);
/*
* Now create the new index.
@@ -1471,6 +1456,32 @@ index_concurrently_create_copy(Relation heapRelation, Oid oldIndexId,
return newIndexId;
}
+NullableDatum *
+get_index_stattargets(Oid indexid, IndexInfo *indInfo)
+{
+ NullableDatum *stattargets;
+
+ /* Extract statistic targets for each attribute */
+ stattargets = palloc0_array(NullableDatum, indInfo->ii_NumIndexAttrs);
+ for (int i = 0; i < indInfo->ii_NumIndexAttrs; i++)
+ {
+ HeapTuple tp;
+ Datum dat;
+ bool isnull;
+
+ tp = SearchSysCache2(ATTNUM, ObjectIdGetDatum(indexid), Int16GetDatum(i + 1));
+ if (!HeapTupleIsValid(tp))
+ elog(ERROR, "cache lookup failed for attribute %d of relation %u",
+ i + 1, indexid);
+ dat = SysCacheGetAttr(ATTNUM, tp, Anum_pg_attribute_attstattarget, &isnull);
+ ReleaseSysCache(tp);
+ stattargets[i].value = dat;
+ stattargets[i].isnull = isnull;
+ }
+
+ return stattargets;
+}
+
/*
* index_concurrently_build
*
diff --git a/src/backend/catalog/system_views.sql b/src/backend/catalog/system_views.sql
index 7a595c84db..d1e75b5f44 100644
--- a/src/backend/catalog/system_views.sql
+++ b/src/backend/catalog/system_views.sql
@@ -1240,16 +1240,19 @@ CREATE VIEW pg_stat_progress_cluster AS
WHEN 2 THEN 'index scanning heap'
WHEN 3 THEN 'sorting tuples'
WHEN 4 THEN 'writing new heap'
- WHEN 5 THEN 'swapping relation files'
- WHEN 6 THEN 'rebuilding index'
- WHEN 7 THEN 'performing final cleanup'
+ WHEN 5 THEN 'catch-up'
+ WHEN 6 THEN 'swapping relation files'
+ WHEN 7 THEN 'rebuilding index'
+ WHEN 8 THEN 'performing final cleanup'
END AS phase,
CAST(S.param3 AS oid) AS cluster_index_relid,
S.param4 AS heap_tuples_scanned,
- S.param5 AS heap_tuples_written,
- S.param6 AS heap_blks_total,
- S.param7 AS heap_blks_scanned,
- S.param8 AS index_rebuild_count
+ S.param5 AS heap_tuples_inserted,
+ S.param6 AS heap_tuples_updated,
+ S.param7 AS heap_tuples_deleted,
+ S.param8 AS heap_blks_total,
+ S.param9 AS heap_blks_scanned,
+ S.param10 AS index_rebuild_count
FROM pg_stat_get_progress_info('CLUSTER') AS S
LEFT JOIN pg_database D ON S.datid = D.oid;
diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index 99193f5c88..c9cc061c45 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -25,6 +25,10 @@
#include "access/toast_internals.h"
#include "access/transam.h"
#include "access/xact.h"
+#include "access/xlog.h"
+#include "access/xlog_internal.h"
+#include "access/xloginsert.h"
+#include "access/xlogutils.h"
#include "catalog/catalog.h"
#include "catalog/dependency.h"
#include "catalog/heap.h"
@@ -32,6 +36,7 @@
#include "catalog/namespace.h"
#include "catalog/objectaccess.h"
#include "catalog/pg_am.h"
+#include "catalog/pg_control.h"
#include "catalog/pg_inherits.h"
#include "catalog/toasting.h"
#include "commands/cluster.h"
@@ -39,10 +44,15 @@
#include "commands/progress.h"
#include "commands/tablecmds.h"
#include "commands/vacuum.h"
+#include "executor/executor.h"
#include "miscadmin.h"
#include "optimizer/optimizer.h"
#include "pgstat.h"
+#include "replication/decode.h"
+#include "replication/logical.h"
+#include "replication/snapbuild.h"
#include "storage/bufmgr.h"
+#include "storage/ipc.h"
#include "storage/lmgr.h"
#include "storage/predicate.h"
#include "utils/acl.h"
@@ -56,6 +66,8 @@
#include "utils/snapmgr.h"
#include "utils/syscache.h"
+typedef struct RewriteStateData *RewriteState;
+
/*
* 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
@@ -67,17 +79,183 @@ typedef struct
Oid indexOid;
} RelToCluster;
+/*
+ * The following definitions are used for concurrent processing.
+ */
+
+/*
+ * OID of the table being processed by this backend.
+ */
+static Oid clustered_rel = InvalidOid;
+/* The same for its TOAST relation. */
+static Oid clustered_rel_toast = InvalidOid;
+
+/*
+ * The locators are used to avoid logical decoding of data that we do not need
+ * for our table.
+ */
+RelFileLocator clustered_rel_locator = {.relNumber = InvalidOid};
+RelFileLocator clustered_rel_toast_locator = {.relNumber = InvalidOid};
+
+/* XXX Do we also need to mention VACUUM FULL CONCURRENTLY? */
+#define CLUSTER_IN_PROGRESS_MESSAGE \
+ "relation \"%s\" is already being processed by CLUSTER CONCURRENTLY"
+
+/*
+ * Everything we need to call ExecInsertIndexTuples().
+ */
+typedef struct IndexInsertState
+{
+ ResultRelInfo *rri;
+ EState *estate;
+ ExprContext *econtext;
+
+ Relation ident_index;
+} IndexInsertState;
-static void cluster_multiple_rels(List *rtcs, ClusterParams *params);
-static void rebuild_relation(Relation OldHeap, Relation index, bool verbose);
+/*
+ * Catalog information to check if another backend changed the relation in
+ * such a way that makes CLUSTER CONCURRENTLY unable to continue. Such changes
+ * are possible because cluster_rel() has to release its lock on the relation
+ * in order to acquire AccessExclusiveLock that it needs to swap the relation
+ * files.
+ *
+ * The most obvious problem is that the tuple descriptor has changed, since
+ * then the tuples we try to insert into the new storage are not guaranteed to
+ * fit into the storage.
+ *
+ * Another problem is relfilenode changed by another backend. It's not
+ * necessarily a correctness issue (e.g. when the other backend ran
+ * cluster_rel()), but it's safer for us to terminate the table processing in
+ * such cases. However, this information is also needs to be checked during
+ * logical decoding, so we store it in global variables clustered_rel_locator
+ * and clustered_rel_toast_locator above.
+ *
+ * Where possible, commands which might change the relation in an incompatible
+ * way should check if CLUSTER CONCURRENTLY is running, before they start to
+ * do the actual changes (see is_concurrent_cluster_in_progress()). Anything
+ * else must be caught by check_catalog_changes(), which uses this structure.
+ */
+typedef struct CatalogState
+{
+ /* Tuple descriptor of the relation. */
+ TupleDesc tupdesc;
+
+ /* The number of indexes tracked. */
+ int ninds;
+ /* The index OIDs. */
+ Oid *ind_oids;
+ /* The index tuple descriptors. */
+ TupleDesc *ind_tupdescs;
+
+ /* The following are copies of the corresponding fields of pg_class. */
+ char relpersistence;
+ char replident;
+
+ /* rd_replidindex */
+ Oid replidindex;
+} CatalogState;
+
+/* The WAL segment being decoded. */
+static XLogSegNo cluster_current_segment = 0;
+
+static void cluster_multiple_rels(List *rtcs, ClusterParams *params,
+ LOCKMODE lockmode, bool isTopLevel);
+static void rebuild_relation(Relation OldHeap, Relation index, bool verbose,
+ bool concurrent, bool is_vacuum);
static void copy_table_data(Relation NewHeap, Relation OldHeap, Relation OldIndex,
+ Snapshot snapshot, LogicalDecodingContext *decoding_ctx,
bool verbose, bool *pSwapToastByContent,
TransactionId *pFreezeXid, MultiXactId *pCutoffMulti);
static List *get_tables_to_cluster(MemoryContext cluster_context);
static List *get_tables_to_cluster_partitioned(MemoryContext cluster_context,
Oid indexOid);
static bool cluster_is_permitted_for_relation(Oid relid, Oid userid);
+static void check_concurrent_cluster_requirements(Relation rel,
+ bool isTopLevel,
+ bool isCluster);
+static void begin_concurrent_cluster(Relation *rel_p, Relation *index_p,
+ bool *entered_p);
+static void end_concurrent_cluster(bool error);
+static void cluster_before_shmem_exit_callback(int code, Datum arg);
+static CatalogState *get_catalog_state(Relation rel, bool is_vacuum);
+static void free_catalog_state(CatalogState *state);
+static void check_catalog_changes(Relation rel, CatalogState *cat_state);
+static LogicalDecodingContext *setup_logical_decoding(Oid relid,
+ const char *slotname,
+ TupleDesc tupdesc);
+static HeapTuple get_changed_tuple(char *change);
+static void apply_concurrent_changes(ClusterDecodingState *dstate,
+ Relation rel, ScanKey key, int nkeys,
+ IndexInsertState *iistate);
+static void apply_concurrent_insert(Relation rel, ConcurrentChange *change,
+ HeapTuple tup, IndexInsertState *iistate,
+ TupleTableSlot *index_slot);
+static void apply_concurrent_update(Relation rel, HeapTuple tup,
+ HeapTuple tup_target,
+ ConcurrentChange *change,
+ IndexInsertState *iistate,
+ TupleTableSlot *index_slot);
+static void apply_concurrent_delete(Relation rel, HeapTuple tup_target,
+ ConcurrentChange *change);
+static HeapTuple find_target_tuple(Relation rel, ScanKey key, int nkeys,
+ HeapTuple tup_key,
+ IndexInsertState *iistate,
+ TupleTableSlot *ident_slot,
+ IndexScanDesc *scan_p);
+static void process_concurrent_changes(LogicalDecodingContext *ctx,
+ XLogRecPtr end_of_wal,
+ Relation rel_dst,
+ Relation rel_src,
+ ScanKey ident_key,
+ int ident_key_nentries,
+ IndexInsertState *iistate);
+static IndexInsertState *get_index_insert_state(Relation relation,
+ Oid ident_index_id);
+static ScanKey build_identity_key(Oid ident_idx_oid, Relation rel_src,
+ int *nentries);
+static void free_index_insert_state(IndexInsertState *iistate);
+static void cleanup_logical_decoding(LogicalDecodingContext *ctx);
+static void rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
+ Relation cl_index,
+ CatalogState *cat_state,
+ LogicalDecodingContext *ctx,
+ bool swap_toast_by_content,
+ TransactionId frozenXid,
+ MultiXactId cutoffMulti);
+static List *build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes);
+
+/*
+ * Use this API when relation needs to be unlocked, closed and re-opened. If
+ * the relation got dropped while being unlocked, raise ERROR that mentions
+ * the relation name rather than OID.
+ */
+typedef struct RelReopenInfo
+{
+ /*
+ * The relation to be closed. Pointer to the value is stored here so that
+ * the user gets his reference updated automatically on re-opening.
+ *
+ * When calling unlock_and_close_relations(), 'relid' can be passed
+ * instead of 'rel_p' when the caller only needs to gather information for
+ * subsequent opening.
+ */
+ Relation *rel_p;
+ Oid relid;
+ char relkind;
+ LOCKMODE lockmode_orig; /* The existing lock mode */
+ LOCKMODE lockmode_new; /* The lock mode after the relation is
+ * re-opened */
+
+ char *relname; /* Relation name, initialized automatically. */
+} RelReopenInfo;
+
+static void init_rel_reopen_info(RelReopenInfo *rri, Relation *rel_p,
+ Oid relid, LOCKMODE lockmode_orig,
+ LOCKMODE lockmode_new);
+static void unlock_and_close_relations(RelReopenInfo *rels, int nrel);
+static void reopen_relations(RelReopenInfo *rels, int nrel);
/*---------------------------------------------------------------------------
* This cluster code allows for clustering multiple tables at once. Because
@@ -109,10 +287,12 @@ cluster(ParseState *pstate, ClusterStmt *stmt, bool isTopLevel)
ListCell *lc;
ClusterParams params = {0};
bool verbose = false;
+ bool concurrent = false;
Relation rel = NULL;
Oid indexOid = InvalidOid;
MemoryContext cluster_context;
List *rtcs;
+ LOCKMODE lockmode;
/* Parse option list */
foreach(lc, stmt->params)
@@ -121,6 +301,8 @@ cluster(ParseState *pstate, ClusterStmt *stmt, bool isTopLevel)
if (strcmp(opt->defname, "verbose") == 0)
verbose = defGetBoolean(opt);
+ else if (strcmp(opt->defname, "concurrently") == 0)
+ concurrent = defGetBoolean(opt);
else
ereport(ERROR,
(errcode(ERRCODE_SYNTAX_ERROR),
@@ -129,20 +311,30 @@ cluster(ParseState *pstate, ClusterStmt *stmt, bool isTopLevel)
parser_errposition(pstate, opt->location)));
}
- params.options = (verbose ? CLUOPT_VERBOSE : 0);
+ params.options =
+ (verbose ? CLUOPT_VERBOSE : 0) |
+ (concurrent ? CLUOPT_CONCURRENT : 0);
+
+ /*
+ * Determine the lock mode expected by cluster_rel().
+ *
+ * In the exclusive case, we obtain AccessExclusiveLock right away to
+ * avoid lock-upgrade hazard in the single-transaction case. In the
+ * CONCURRENT case, the AccessExclusiveLock will only be used at the end
+ * of processing, supposedly for very short time. Until then, we'll have
+ * to unlock the relation temporarily, so there's no lock-upgrade hazard.
+ */
+ lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ?
+ AccessExclusiveLock : ShareUpdateExclusiveLock;
if (stmt->relation != NULL)
{
/* This is the single-relation case. */
Oid tableOid;
- /*
- * Find, lock, and check permissions on the table. We obtain
- * AccessExclusiveLock right away to avoid lock-upgrade hazard in the
- * single-transaction case.
- */
+ /* Find, lock, and check permissions on the table. */
tableOid = RangeVarGetRelidExtended(stmt->relation,
- AccessExclusiveLock,
+ lockmode,
0,
RangeVarCallbackMaintainsTable,
NULL);
@@ -194,7 +386,7 @@ cluster(ParseState *pstate, ClusterStmt *stmt, bool isTopLevel)
/* For non-partitioned tables, do what we came here to do. */
if (rel->rd_rel->relkind != RELKIND_PARTITIONED_TABLE)
{
- cluster_rel(rel, indexOid, ¶ms);
+ cluster_rel(rel, indexOid, ¶ms, isTopLevel, false);
/* cluster_rel closes the relation, but keeps lock */
return;
@@ -202,10 +394,29 @@ cluster(ParseState *pstate, ClusterStmt *stmt, bool isTopLevel)
}
/*
- * By here, we know we are in a multi-table situation. In order to avoid
- * holding locks for too long, we want to process each table in its own
- * transaction. This forces us to disallow running inside a user
- * transaction block.
+ * By here, we know we are in a multi-table situation.
+ *
+ * Concurrent processing is currently considered rather special (e.g. in
+ * terms of resources consumed) so it is not performed in bulk.
+ */
+ if (params.options & CLUOPT_CONCURRENT)
+ {
+ if (rel != NULL)
+ {
+ Assert(rel->rd_rel->relkind == RELKIND_PARTITIONED_TABLE);
+ ereport(ERROR,
+ (errmsg("CLUSTER (CONCURRENTLY) not supported for partitioned tables"),
+ errhint("Consider running the command for individual partitions.")));
+ }
+ else
+ ereport(ERROR,
+ (errmsg("CLUSTER (CONCURRENTLY) requires explicit table name")));
+ }
+
+ /*
+ * In order to avoid holding locks for too long, we want to process each
+ * table in its own transaction. This forces us to disallow running
+ * inside a user transaction block.
*/
PreventInTransactionBlock(isTopLevel, "CLUSTER");
@@ -230,11 +441,14 @@ cluster(ParseState *pstate, ClusterStmt *stmt, bool isTopLevel)
if (rel != NULL)
{
Assert(rel->rd_rel->relkind == RELKIND_PARTITIONED_TABLE);
+ /* See the ereport() above. */
+ Assert((params.options & CLUOPT_CONCURRENT) == 0);
+
check_index_is_clusterable(rel, indexOid, AccessShareLock);
rtcs = get_tables_to_cluster_partitioned(cluster_context, indexOid);
/* close relation, releasing lock on parent table */
- table_close(rel, AccessExclusiveLock);
+ table_close(rel, lockmode);
}
else
{
@@ -243,7 +457,7 @@ cluster(ParseState *pstate, ClusterStmt *stmt, bool isTopLevel)
}
/* Do the job. */
- cluster_multiple_rels(rtcs, ¶ms);
+ cluster_multiple_rels(rtcs, ¶ms, lockmode, isTopLevel);
/* Start a new transaction for the cleanup work. */
StartTransactionCommand();
@@ -260,7 +474,8 @@ cluster(ParseState *pstate, ClusterStmt *stmt, bool isTopLevel)
* return.
*/
static void
-cluster_multiple_rels(List *rtcs, ClusterParams *params)
+cluster_multiple_rels(List *rtcs, ClusterParams *params, LOCKMODE lockmode,
+ bool isTopLevel)
{
ListCell *lc;
@@ -280,10 +495,10 @@ cluster_multiple_rels(List *rtcs, ClusterParams *params)
/* functions in indexes may want a snapshot set */
PushActiveSnapshot(GetTransactionSnapshot());
- rel = table_open(rtc->tableOid, AccessExclusiveLock);
+ rel = table_open(rtc->tableOid, lockmode);
/* Process this table */
- cluster_rel(rel, rtc->indexOid, params);
+ cluster_rel(rel, rtc->indexOid, params, isTopLevel, false);
/* cluster_rel closes the relation, but keeps lock */
PopActiveSnapshot();
@@ -306,9 +521,16 @@ cluster_multiple_rels(List *rtcs, ClusterParams *params)
* If indexOid is InvalidOid, the table will be rewritten in physical order
* instead of index order. This is the new implementation of VACUUM FULL,
* and error messages should refer to the operation as VACUUM not CLUSTER.
+ *
+ * Note that, in the concurrent case, the function releases the lock at some
+ * point, in order to get AccessExclusiveLock for the final steps (i.e. to
+ * swap the relation files). To make things simpler, the caller should expect
+ * OldHeap to be closed on return, regardless CLUOPT_CONCURRENT. (The
+ * AccessExclusiveLock is kept till the end of the transaction.)
*/
void
-cluster_rel(Relation OldHeap, Oid indexOid, ClusterParams *params)
+cluster_rel(Relation OldHeap, Oid indexOid, ClusterParams *params,
+ bool isTopLevel, bool isVacuum)
{
Oid tableOid = RelationGetRelid(OldHeap);
Oid save_userid;
@@ -317,8 +539,46 @@ cluster_rel(Relation OldHeap, Oid indexOid, ClusterParams *params)
bool verbose = ((params->options & CLUOPT_VERBOSE) != 0);
bool recheck = ((params->options & CLUOPT_RECHECK) != 0);
Relation index;
+ bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0);
+ LOCKMODE lmode;
+ bool entered, success;
+
+ /*
+ * Check that the correct lock is held. The lock mode is
+ * AccessExclusiveLock for normal processing and ShareUpdateExclusiveLock
+ * for concurrent processing (so that SELECT, INSERT, UPDATE and DELETE
+ * commands work, but cluster_rel() cannot be called concurrently for the
+ * same relation).
+ */
+ lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock;
+
+ /*
+ * Skip the relation if it's being processed concurrently. In such a case,
+ * we cannot rely on a lock because the other backend needs to release it
+ * temporarily at some point.
+ *
+ * This check should not take place until we have a lock that prevents
+ * another backend from starting VACUUM FULL / CLUSTER CONCURRENTLY after
+ * our check.
+ */
+ Assert(CheckRelationLockedByMe(OldHeap, lmode, false));
+ if (is_concurrent_cluster_in_progress(tableOid))
+ {
+ ereport(NOTICE,
+ (errmsg(CLUSTER_IN_PROGRESS_MESSAGE,
+ RelationGetRelationName(OldHeap))));
+ table_close(OldHeap, lmode);
+ return;
+ }
+
+ /* There are specific requirements on concurrent processing. */
+ if (concurrent)
+ {
+ check_concurrent_cluster_requirements(OldHeap, isTopLevel,
+ OidIsValid(indexOid));
- Assert(CheckRelationLockedByMe(OldHeap, AccessExclusiveLock, false));
+ check_relation_is_clusterable_concurrently(OldHeap, isVacuum);
+ }
/* Check for user-requested abort. */
CHECK_FOR_INTERRUPTS();
@@ -355,7 +615,7 @@ cluster_rel(Relation OldHeap, Oid indexOid, ClusterParams *params)
/* Check that the user still has privileges for the relation */
if (!cluster_is_permitted_for_relation(tableOid, save_userid))
{
- relation_close(OldHeap, AccessExclusiveLock);
+ relation_close(OldHeap, lmode);
goto out;
}
@@ -370,7 +630,7 @@ cluster_rel(Relation OldHeap, Oid indexOid, ClusterParams *params)
*/
if (RELATION_IS_OTHER_TEMP(OldHeap))
{
- relation_close(OldHeap, AccessExclusiveLock);
+ relation_close(OldHeap, lmode);
goto out;
}
@@ -381,7 +641,7 @@ cluster_rel(Relation OldHeap, Oid indexOid, ClusterParams *params)
*/
if (!SearchSysCacheExists1(RELOID, ObjectIdGetDatum(indexOid)))
{
- relation_close(OldHeap, AccessExclusiveLock);
+ relation_close(OldHeap, lmode);
goto out;
}
@@ -392,7 +652,7 @@ cluster_rel(Relation OldHeap, Oid indexOid, ClusterParams *params)
if ((params->options & CLUOPT_RECHECK_ISCLUSTERED) != 0 &&
!get_index_isclustered(indexOid))
{
- relation_close(OldHeap, AccessExclusiveLock);
+ relation_close(OldHeap, lmode);
goto out;
}
}
@@ -408,6 +668,11 @@ cluster_rel(Relation OldHeap, Oid indexOid, ClusterParams *params)
ereport(ERROR,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("cannot cluster a shared catalog")));
+ /*
+ * The CONCURRENT case should have been rejected earlier because it does
+ * not support system catalogs.
+ */
+ Assert(!(OldHeap->rd_rel->relisshared && concurrent));
/*
* Don't process temp tables of other backends ... their local buffer
@@ -435,7 +700,7 @@ cluster_rel(Relation OldHeap, Oid indexOid, ClusterParams *params)
if (OidIsValid(indexOid))
{
/* verify the index is good and lock it */
- check_index_is_clusterable(OldHeap, indexOid, AccessExclusiveLock);
+ check_index_is_clusterable(OldHeap, indexOid, lmode);
/* also open it */
index = index_open(indexOid, NoLock);
}
@@ -452,7 +717,8 @@ cluster_rel(Relation OldHeap, Oid indexOid, ClusterParams *params)
if (OldHeap->rd_rel->relkind == RELKIND_MATVIEW &&
!RelationIsPopulated(OldHeap))
{
- relation_close(OldHeap, AccessExclusiveLock);
+ index_close(index, lmode);
+ relation_close(OldHeap, lmode);
goto out;
}
@@ -465,11 +731,42 @@ cluster_rel(Relation OldHeap, Oid indexOid, ClusterParams *params)
* invalid, because we move tuples around. Promote them to relation
* locks. Predicate locks on indexes will be promoted when they are
* reindexed.
+ *
+ * During concurrent processing, the heap as well as its indexes stay in
+ * operation, so we postpone this step until they are locked using
+ * AccessExclusiveLock near the end of the processing.
*/
- TransferPredicateLocksToHeapRelation(OldHeap);
+ if (!concurrent)
+ TransferPredicateLocksToHeapRelation(OldHeap);
/* rebuild_relation does all the dirty work */
- rebuild_relation(OldHeap, index, verbose);
+ entered = false;
+ success = false;
+ PG_TRY();
+ {
+ /*
+ * For concurrent processing, make sure other transactions treat this
+ * table as if it was a system / user catalog, and WAL the relevant
+ * additional information. ERROR is raised if another backend is
+ * processing the same table.
+ */
+ if (concurrent)
+ {
+ Relation *index_p = index ? &index : NULL;
+
+ begin_concurrent_cluster(&OldHeap, index_p, &entered);
+ }
+
+ rebuild_relation(OldHeap, index, verbose, concurrent, isVacuum);
+ success = true;
+ }
+ PG_FINALLY();
+ {
+ if (concurrent && entered)
+ end_concurrent_cluster(!success);
+ }
+ PG_END_TRY();
+
/* rebuild_relation closes OldHeap, and index if valid */
out:
@@ -615,18 +912,86 @@ mark_index_clustered(Relation rel, Oid indexOid, bool is_internal)
table_close(pg_index, RowExclusiveLock);
}
+/*
+ * Check if the CONCURRENTLY option is legal for the relation.
+ */
+void
+check_relation_is_clusterable_concurrently(Relation rel, bool is_vacuum)
+{
+ char relpersistence, replident;
+ Oid ident_idx;
+ const char *cmd = is_vacuum ? "VACUUM" : "CLUSTER";
+
+ /* Data changes in system relations are not logically decoded. */
+ if (IsCatalogRelation(rel))
+ ereport(ERROR,
+ (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
+ errmsg("cannot process relation \"%s\"",
+ RelationGetRelationName(rel)),
+ errhint("%s (CONCURRENTLY) is not supported for catalog relations.",
+ cmd)));
+
+ if (IsToastRelation(rel))
+ ereport(ERROR,
+ (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
+ errmsg("cannot process relation \"%s\"",
+ RelationGetRelationName(rel)),
+ errhint("%s (CONCURRENTLY) is not supported for TOAST relations, unless the main relation is processed too.",
+ cmd)));
+
+ relpersistence = rel->rd_rel->relpersistence;
+ if (relpersistence != RELPERSISTENCE_PERMANENT)
+ ereport(ERROR,
+ (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
+ errmsg("cannot process relation \"%s\"",
+ RelationGetRelationName(rel)),
+ errhint("%s (CONCURRENTLY) is only allowed for permanent relations.",
+ cmd)));
+
+ /* With NOTHING, WAL does not contain the old tuple. */
+ replident = rel->rd_rel->relreplident;
+ if (replident == REPLICA_IDENTITY_NOTHING)
+ ereport(ERROR,
+ (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
+ errmsg("cannot process relation \"%s\"",
+ RelationGetRelationName(rel)),
+ errhint("Relation \"%s\" has insufficient replication identity.",
+ RelationGetRelationName(rel))));
+
+ /*
+ * Identity index is not set if the replica identity is FULL, but PK might
+ * exist in such a case.
+ */
+ ident_idx = RelationGetReplicaIndex(rel);
+ if (!OidIsValid(ident_idx) && OidIsValid(rel->rd_pkindex))
+ ident_idx = rel->rd_pkindex;
+ if (!OidIsValid(ident_idx))
+ ereport(ERROR,
+ (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
+ errmsg("cannot process relation \"%s\"",
+ RelationGetRelationName(rel)),
+ (errhint("Relation \"%s\" has no identity index.",
+ RelationGetRelationName(rel)))));
+}
+
/*
* rebuild_relation: rebuild an existing relation in index or physical order
*
- * OldHeap: table to rebuild.
+ * OldHeap: table to rebuild. See cluster_rel() for comments on the required
+ * lock strength.
+ *
* index: index to cluster by, or NULL to rewrite in physical order.
*
- * On entry, heap and index (if one is given) must be open, and
- * AccessExclusiveLock held on them.
- * On exit, they are closed, but locks on them are not released.
+ * On entry, heap and index (if one is given) must be open, and the
+ * appropriate lock held on them (AccessExclusiveLock for exclusive processing
+ * and ShareUpdateExclusiveLock for concurrent processing)..
+ *
+ * On exit, they are closed, but still locked with AccessExclusiveLock (The
+ * function handles the lock upgrade if 'concurrent' is true.)
*/
static void
-rebuild_relation(Relation OldHeap, Relation index, bool verbose)
+rebuild_relation(Relation OldHeap, Relation index, bool verbose,
+ bool concurrent, bool is_vacuum)
{
Oid tableOid = RelationGetRelid(OldHeap);
Oid accessMethod = OldHeap->rd_rel->relam;
@@ -634,13 +999,81 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose)
Oid OIDNewHeap;
Relation NewHeap;
char relpersistence;
- bool is_system_catalog;
bool swap_toast_by_content;
TransactionId frozenXid;
MultiXactId cutoffMulti;
+ NameData slotname;
+ LogicalDecodingContext *ctx = NULL;
+ Snapshot snapshot = NULL;
+ CatalogState *cat_state = NULL;
+ LOCKMODE lmode;
+
+ lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock;
+
+ Assert(CheckRelationLockedByMe(OldHeap, lmode, false) &&
+ (index == NULL || CheckRelationLockedByMe(index, lmode, false)));
+
+ if (concurrent)
+ {
+ TupleDesc tupdesc;
+ RelReopenInfo rri[2];
+ int nrel;
+
+ /*
+ * CLUSTER CONCURRENTLY is not allowed in a transaction block, so this
+ * should never fire.
+ */
+ Assert(GetTopTransactionIdIfAny() == InvalidTransactionId);
+
+ /*
+ * A single backend should not execute multiple CLUSTER commands at a
+ * time, so use PID to make the slot unique.
+ */
+ snprintf(NameStr(slotname), NAMEDATALEN, "cluster_%d", MyProcPid);
+
+ /*
+ * Gather catalog information so that we can check later if the old
+ * relation has not changed while unlocked.
+ *
+ * Since this function also checks if the relation can be processed,
+ * it's important to call it before we spend notable amount of time to
+ * setup the logical decoding. Not sure though if it's necessary to do
+ * it even earlier.
+ */
+ cat_state = get_catalog_state(OldHeap, is_vacuum);
+
+ tupdesc = CreateTupleDescCopy(RelationGetDescr(OldHeap));
+
+ /*
+ * Unlock the relation (and possibly the clustering index) to avoid
+ * deadlock because setup_logical_decoding() will wait for all the
+ * running transactions (with XID assigned) to finish. Some of those
+ * transactions might be waiting for a lock on our relation.
+ */
+ nrel = 0;
+ init_rel_reopen_info(&rri[nrel++], &OldHeap, InvalidOid,
+ ShareUpdateExclusiveLock,
+ ShareUpdateExclusiveLock);
+ if (index)
+ init_rel_reopen_info(&rri[nrel++], &index, InvalidOid,
+ ShareUpdateExclusiveLock,
+ ShareUpdateExclusiveLock);
+ unlock_and_close_relations(rri, nrel);
+
+ /* Prepare to capture the concurrent data changes. */
+ ctx = setup_logical_decoding(tableOid, NameStr(slotname), tupdesc);
+
+ /* Lock the table (and index) again. */
+ reopen_relations(rri, nrel);
+
+ /*
+ * Check if a 'tupdesc' could have changed while the relation was
+ * unlocked.
+ */
+ check_catalog_changes(OldHeap, cat_state);
- Assert(CheckRelationLockedByMe(OldHeap, AccessExclusiveLock, false) &&
- (index == NULL || CheckRelationLockedByMe(index, AccessExclusiveLock, false)));
+ snapshot = SnapBuildInitialSnapshotForCluster(ctx->snapshot_builder);
+ }
if (index)
/* Mark the correct index as clustered */
@@ -648,7 +1081,6 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose)
/* Remember info about rel before closing OldHeap */
relpersistence = OldHeap->rd_rel->relpersistence;
- is_system_catalog = IsSystemRelation(OldHeap);
/*
* Create the transient table that will receive the re-ordered data.
@@ -664,30 +1096,51 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose)
NewHeap = table_open(OIDNewHeap, NoLock);
/* Copy the heap data into the new table in the desired order */
- copy_table_data(NewHeap, OldHeap, index, verbose,
+ copy_table_data(NewHeap, OldHeap, index, snapshot, ctx, verbose,
&swap_toast_by_content, &frozenXid, &cutoffMulti);
+ if (concurrent)
+ {
+ rebuild_relation_finish_concurrent(NewHeap, OldHeap, index,
+ cat_state, ctx,
+ swap_toast_by_content,
+ frozenXid, cutoffMulti);
+
+ pgstat_progress_update_param(PROGRESS_CLUSTER_PHASE,
+ PROGRESS_CLUSTER_PHASE_FINAL_CLEANUP);
+
+ /* Done with decoding. */
+ FreeSnapshot(snapshot);
+ free_catalog_state(cat_state);
+ cleanup_logical_decoding(ctx);
+ ReplicationSlotRelease();
+ ReplicationSlotDrop(NameStr(slotname), false);
+ }
+ else
+ {
+ bool is_system_catalog = IsSystemRelation(OldHeap);
- /* Close relcache entries, but keep lock until transaction commit */
- table_close(OldHeap, NoLock);
- if (index)
- index_close(index, NoLock);
+ /* Close relcache entries, but keep lock until transaction commit */
+ table_close(OldHeap, NoLock);
+ if (index)
+ index_close(index, NoLock);
- /*
- * Close the new relation so it can be dropped as soon as the storage is
- * swapped. The relation is not visible to others, so no need to unlock it
- * explicitly.
- */
- table_close(NewHeap, NoLock);
+ /*
+ * Close the new relation so it can be dropped as soon as the storage
+ * is swapped. The relation is not visible to others, so no need to
+ * unlock it explicitly.
+ */
+ table_close(NewHeap, NoLock);
- /*
- * Swap the physical files of the target and transient tables, then
- * rebuild the target's indexes and throw away the transient table.
- */
- finish_heap_swap(tableOid, OIDNewHeap, is_system_catalog,
- swap_toast_by_content, false, true,
- frozenXid, cutoffMulti,
- relpersistence);
+ /*
+ * Swap the physical files of the target and transient tables, then
+ * rebuild the target's indexes and throw away the transient table.
+ */
+ finish_heap_swap(tableOid, OIDNewHeap, is_system_catalog,
+ swap_toast_by_content, false, true, true,
+ frozenXid, cutoffMulti,
+ relpersistence);
+ }
}
@@ -822,15 +1275,19 @@ make_new_heap(Oid OIDOldHeap, Oid NewTableSpace, Oid NewAccessMethod,
/*
* Do the physical copying of table data.
*
+ * 'snapshot' and 'decoding_ctx': see table_relation_copy_for_cluster(). Pass
+ * iff concurrent processing is required.
+ *
* There are three output parameters:
* *pSwapToastByContent is set true if toast tables must be swapped by content.
* *pFreezeXid receives the TransactionId used as freeze cutoff point.
* *pCutoffMulti receives the MultiXactId used as a cutoff point.
*/
static void
-copy_table_data(Relation NewHeap, Relation OldHeap, Relation OldIndex, bool verbose,
- bool *pSwapToastByContent, TransactionId *pFreezeXid,
- MultiXactId *pCutoffMulti)
+copy_table_data(Relation NewHeap, Relation OldHeap, Relation OldIndex,
+ Snapshot snapshot, LogicalDecodingContext *decoding_ctx,
+ bool verbose, bool *pSwapToastByContent,
+ TransactionId *pFreezeXid, MultiXactId *pCutoffMulti)
{
Relation relRelation;
HeapTuple reltup;
@@ -847,6 +1304,7 @@ copy_table_data(Relation NewHeap, Relation OldHeap, Relation OldIndex, bool verb
int elevel = verbose ? INFO : DEBUG2;
PGRUsage ru0;
char *nspname;
+ bool concurrent = snapshot != NULL;
pg_rusage_init(&ru0);
@@ -873,8 +1331,12 @@ copy_table_data(Relation NewHeap, Relation OldHeap, Relation OldIndex, bool verb
*
* We don't need to open the toast relation here, just lock it. The lock
* will be held till end of transaction.
+ *
+ * In the CONCURRENT case, the lock does not help because we need to
+ * release it temporarily at some point. Instead, we expect VACUUM /
+ * CLUSTER to skip tables which are present in ClusteredRelsHash.
*/
- if (OldHeap->rd_rel->reltoastrelid)
+ if (OldHeap->rd_rel->reltoastrelid && !concurrent)
LockRelationOid(OldHeap->rd_rel->reltoastrelid, AccessExclusiveLock);
/*
@@ -950,8 +1412,46 @@ copy_table_data(Relation NewHeap, Relation OldHeap, Relation OldIndex, bool verb
* provided, else plain seqscan.
*/
if (OldIndex != NULL && OldIndex->rd_rel->relam == BTREE_AM_OID)
+ {
+ ResourceOwner oldowner = CurrentResourceOwner;
+
+ /*
+ * In the CONCURRENT case, do the planning in a subtransaction so that
+ * we don't leave any additional locks behind us that we cannot
+ * release easily.
+ */
+ if (concurrent)
+ {
+ Assert(CheckRelationLockedByMe(OldHeap, ShareUpdateExclusiveLock,
+ false));
+ Assert(CheckRelationLockedByMe(OldIndex, ShareUpdateExclusiveLock,
+ false));
+ BeginInternalSubTransaction("plan_cluster_use_sort");
+ }
+
use_sort = plan_cluster_use_sort(RelationGetRelid(OldHeap),
RelationGetRelid(OldIndex));
+
+ if (concurrent)
+ {
+ PgBackendProgress progress;
+
+ /*
+ * Command progress reporting gets terminated at subtransaction
+ * end. Save the status so it can be eventually restored.
+ */
+ memcpy(&progress, &MyBEEntry->st_progress,
+ sizeof(PgBackendProgress));
+
+ /* Release the locks by aborting the subtransaction. */
+ RollbackAndReleaseCurrentSubTransaction();
+
+ /* Restore the progress reporting status. */
+ pgstat_progress_restore_state(&progress);
+
+ CurrentResourceOwner = oldowner;
+ }
+ }
else
use_sort = false;
@@ -980,7 +1480,9 @@ copy_table_data(Relation NewHeap, Relation OldHeap, Relation OldIndex, bool verb
* values (e.g. because the AM doesn't use freezing).
*/
table_relation_copy_for_cluster(OldHeap, NewHeap, OldIndex, use_sort,
- cutoffs.OldestXmin, &cutoffs.FreezeLimit,
+ cutoffs.OldestXmin, snapshot,
+ decoding_ctx,
+ &cutoffs.FreezeLimit,
&cutoffs.MultiXactCutoff,
&num_tuples, &tups_vacuumed,
&tups_recently_dead);
@@ -989,7 +1491,11 @@ copy_table_data(Relation NewHeap, Relation OldHeap, Relation OldIndex, bool verb
*pFreezeXid = cutoffs.FreezeLimit;
*pCutoffMulti = cutoffs.MultiXactCutoff;
- /* Reset rd_toastoid just to be tidy --- it shouldn't be looked at again */
+ /*
+ * Reset rd_toastoid just to be tidy --- it shouldn't be looked at
+ * again. In the CONCURRENT case, we need to set it again before applying
+ * the concurrent changes.
+ */
NewHeap->rd_toastoid = InvalidOid;
num_pages = RelationGetNumberOfBlocks(NewHeap);
@@ -1442,14 +1948,13 @@ finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap,
bool swap_toast_by_content,
bool check_constraints,
bool is_internal,
+ bool reindex,
TransactionId frozenXid,
MultiXactId cutoffMulti,
char newrelpersistence)
{
ObjectAddress object;
Oid mapped_tables[4];
- int reindex_flags;
- ReindexParams reindex_params = {0};
int i;
/* Report that we are now swapping relation files */
@@ -1475,39 +1980,46 @@ finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap,
if (is_system_catalog)
CacheInvalidateCatalog(OIDOldHeap);
- /*
- * Rebuild each index on the relation (but not the toast table, which is
- * all-new at this point). It is important to do this before the DROP
- * step because if we are processing a system catalog that will be used
- * during DROP, we want to have its indexes available. There is no
- * advantage to the other order anyway because this is all transactional,
- * so no chance to reclaim disk space before commit. We do not need a
- * final CommandCounterIncrement() because reindex_relation does it.
- *
- * Note: because index_build is called via reindex_relation, it will never
- * set indcheckxmin true for the indexes. This is OK even though in some
- * sense we are building new indexes rather than rebuilding existing ones,
- * because the new heap won't contain any HOT chains at all, let alone
- * broken ones, so it can't be necessary to set indcheckxmin.
- */
- reindex_flags = REINDEX_REL_SUPPRESS_INDEX_USE;
- if (check_constraints)
- reindex_flags |= REINDEX_REL_CHECK_CONSTRAINTS;
+ if (reindex)
+ {
+ int reindex_flags;
+ ReindexParams reindex_params = {0};
- /*
- * Ensure that the indexes have the same persistence as the parent
- * relation.
- */
- if (newrelpersistence == RELPERSISTENCE_UNLOGGED)
- reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED;
- else if (newrelpersistence == RELPERSISTENCE_PERMANENT)
- reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT;
+ /*
+ * Rebuild each index on the relation (but not the toast table, which
+ * is all-new at this point). It is important to do this before the
+ * DROP step because if we are processing a system catalog that will
+ * be used during DROP, we want to have its indexes available. There
+ * is no advantage to the other order anyway because this is all
+ * transactional, so no chance to reclaim disk space before commit.
+ * We do not need a final CommandCounterIncrement() because
+ * reindex_relation does it.
+ *
+ * Note: because index_build is called via reindex_relation, it will never
+ * set indcheckxmin true for the indexes. This is OK even though in some
+ * sense we are building new indexes rather than rebuilding existing ones,
+ * because the new heap won't contain any HOT chains at all, let alone
+ * broken ones, so it can't be necessary to set indcheckxmin.
+ */
+ reindex_flags = REINDEX_REL_SUPPRESS_INDEX_USE;
+ if (check_constraints)
+ reindex_flags |= REINDEX_REL_CHECK_CONSTRAINTS;
- /* Report that we are now reindexing relations */
- pgstat_progress_update_param(PROGRESS_CLUSTER_PHASE,
- PROGRESS_CLUSTER_PHASE_REBUILD_INDEX);
+ /*
+ * Ensure that the indexes have the same persistence as the parent
+ * relation.
+ */
+ if (newrelpersistence == RELPERSISTENCE_UNLOGGED)
+ reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED;
+ else if (newrelpersistence == RELPERSISTENCE_PERMANENT)
+ reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT;
- reindex_relation(NULL, OIDOldHeap, reindex_flags, &reindex_params);
+ /* Report that we are now reindexing relations */
+ pgstat_progress_update_param(PROGRESS_CLUSTER_PHASE,
+ PROGRESS_CLUSTER_PHASE_REBUILD_INDEX);
+
+ reindex_relation(NULL, OIDOldHeap, reindex_flags, &reindex_params);
+ }
/* Report that we are now doing clean up */
pgstat_progress_update_param(PROGRESS_CLUSTER_PHASE,
@@ -1747,3 +2259,1886 @@ cluster_is_permitted_for_relation(Oid relid, Oid userid)
get_rel_name(relid))));
return false;
}
+
+#define REPL_PLUGIN_NAME "pgoutput_cluster"
+
+/*
+ * Each relation being processed by CLUSTER CONCURRENTLY must be in the
+ * clusteredRels hashtable.
+ */
+typedef struct ClusteredRel
+{
+ Oid relid;
+ Oid dbid;
+} ClusteredRel;
+
+static HTAB *ClusteredRelsHash = NULL;
+
+/* Maximum number of entries in the hashtable. */
+static int maxClusteredRels = 0;
+
+Size
+ClusterShmemSize(void)
+{
+ /*
+ * A replication slot is needed for the processing, so use this GUC to
+ * allocate memory for the hashtable. Reserve also space for TOAST
+ * relations.
+ */
+ maxClusteredRels = max_replication_slots * 2;
+
+ return hash_estimate_size(maxClusteredRels, sizeof(ClusteredRel));
+}
+
+void
+ClusterShmemInit(void)
+{
+ HASHCTL info;
+
+ info.keysize = sizeof(ClusteredRel);
+ info.entrysize = info.keysize;
+
+ ClusteredRelsHash = ShmemInitHash("Clustered Relations",
+ maxClusteredRels,
+ maxClusteredRels,
+ &info,
+ HASH_ELEM | HASH_BLOBS);
+}
+
+/*
+ * Perform a preliminary check whether CLUSTER / VACUUM FULL CONCURRENTLY is
+ * possible. Note that here we only check things that should not change if we
+ * release the relation lock temporarily. The information that can change due
+ * to unlocking is checked in get_catalog_state().
+ */
+static void
+check_concurrent_cluster_requirements(Relation rel, bool isTopLevel,
+ bool isCluster)
+{
+ const char *stmt;
+
+ if (isCluster)
+ stmt = "CLUSTER (CONCURRENTLY)";
+ else
+ stmt = "VACUUM (FULL, CONCURRENTLY)";
+
+ /*
+ * Make sure we have no XID assigned, otherwise call of
+ * setup_logical_decoding() can cause a deadlock.
+ */
+ PreventInTransactionBlock(isTopLevel, stmt);
+
+ CheckSlotPermissions();
+
+ /*
+ * Use an existing function to check if we can use logical
+ * decoding. However note that RecoveryInProgress() should already have
+ * caused error, as it does for the non-concurrent VACUUM FULL / CLUSTER.
+ */
+ CheckLogicalDecodingRequirements();
+
+ /* See ClusterShmemSize() */
+ if (max_replication_slots < 2)
+ ereport(ERROR,
+ errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
+ (errmsg("%s requires \"max_replication_slots\" to be at least 2",
+ stmt)));
+}
+
+/*
+ * Call this function before CLUSTER CONCURRENTLY starts to setup logical
+ * decoding. It makes sure that other users of the table put enough
+ * information into WAL.
+ *
+ * The point is that on various places we expect that the table we're
+ * processing is treated like a system catalog. For example, we need to be
+ * able to scan it using a "historic snapshot" anytime during the processing
+ * (as opposed to scanning only at the start point of the decoding, logical
+ * replication does during initial table synchronization), in order to apply
+ * concurrent UPDATE / DELETE commands.
+ *
+ * Since we need to close and reopen the relation here, the 'rel_p' and
+ * 'index_p' arguments are in/out.
+ *
+ * 'enter_p' receives a bool value telling whether relation OID was entered
+ * into the hashtable or not.
+ */
+static void
+begin_concurrent_cluster(Relation *rel_p, Relation *index_p,
+ bool *entered_p)
+{
+ Relation rel = *rel_p;
+ Oid relid, toastrelid;
+ ClusteredRel key, *entry;
+ bool found;
+ RelReopenInfo rri[2];
+ int nrel;
+ static bool before_shmem_exit_callback_setup = false;
+
+ relid = RelationGetRelid(rel);
+
+ /*
+ * Make sure that we do not leave an entry in ClusteredRelsHash if exiting
+ * due to FATAL.
+ */
+ if (!before_shmem_exit_callback_setup)
+ {
+ before_shmem_exit(cluster_before_shmem_exit_callback, 0);
+ before_shmem_exit_callback_setup = true;
+ }
+
+ memset(&key, 0, sizeof(key));
+ key.relid = relid;
+ key.dbid = MyDatabaseId;
+
+ *entered_p = false;
+ LWLockAcquire(ClusteredRelsLock, LW_EXCLUSIVE);
+ entry = (ClusteredRel *)
+ hash_search(ClusteredRelsHash, &key, HASH_ENTER_NULL, &found);
+ if (found)
+ {
+ /*
+ * Since CLUSTER CONCURRENTLY takes ShareRowExclusiveLock, a conflict
+ * should occur much earlier. However that lock may be released
+ * temporarily, see below. Anyway, we should complain whatever the
+ * reason of the conflict might be.
+ */
+ ereport(ERROR,
+ (errmsg(CLUSTER_IN_PROGRESS_MESSAGE,
+ RelationGetRelationName(rel))));
+ }
+ if (entry == NULL)
+ ereport(ERROR,
+ (errmsg("too many requests for CLUSTER CONCURRENTLY at a time")),
+ (errhint("Please consider increasing the \"max_replication_slots\" configuration parameter.")));
+
+ /*
+ * Even if the insertion of TOAST relid should fail below, the caller has
+ * to do cleanup.
+ */
+ *entered_p = true;
+
+ /*
+ * Enable the callback to remove the entry in case of exit. We should not
+ * do this earlier, otherwise an attempt to insert already existing entry
+ * could make us remove that entry (inserted by another backend) during
+ * ERROR handling.
+ */
+ Assert(!OidIsValid(clustered_rel));
+ clustered_rel = relid;
+
+ /*
+ * TOAST relation is not accessed using historic snapshot, but we enter it
+ * here to protect it from being VACUUMed by another backend. (Lock does
+ * not help in the CONCURRENT case because cannot hold it continuously
+ * till the end of the transaction.) See the comments on locking TOAST
+ * relation in copy_table_data().
+ */
+ toastrelid = rel->rd_rel->reltoastrelid;
+ if (OidIsValid(toastrelid))
+ {
+ key.relid = toastrelid;
+ entry = (ClusteredRel *)
+ hash_search(ClusteredRelsHash, &key, HASH_ENTER_NULL, &found);
+ if (found)
+ /*
+ * If we could enter the main fork the TOAST should succeed
+ * too. Nevertheless, check.
+ */
+ ereport(ERROR,
+ (errmsg("TOAST relation of \"%s\" is already being processed by CLUSTER CONCURRENTLY",
+ RelationGetRelationName(rel))));
+ if (entry == NULL)
+ ereport(ERROR,
+ (errmsg("too many requests for CLUSTER CONCURRENT at a time")),
+ (errhint("Please consider increasing the \"max_replication_slots\" configuration parameter.")));
+
+ Assert(!OidIsValid(clustered_rel_toast));
+ clustered_rel_toast = toastrelid;
+ }
+ LWLockRelease(ClusteredRelsLock);
+
+ /*
+ * Make sure that other backends are aware of the new hash entry.
+ *
+ * Besides sending the invalidation message, we need to force re-opening
+ * of the relation, which includes the actual invalidation (and thus
+ * checking of our hashtable on the next access).
+ */
+ CacheInvalidateRelcacheImmediate(rel);
+ /*
+ * Since the hashtable only needs to be checked by write transactions,
+ * lock the relation in a mode that conflicts with any DML command. (The
+ * reading transactions are supposed to close the relation before opening
+ * it with higher lock.) Once we have the relation (and its index) locked,
+ * we unlock it immediately and then re-lock using the original mode.
+ */
+ nrel = 0;
+ init_rel_reopen_info(&rri[nrel++], rel_p, InvalidOid,
+ ShareUpdateExclusiveLock, ShareLock);
+ if (index_p)
+ {
+ /*
+ * Another transaction might want to open both the relation and the
+ * index. If it already has the relation lock and is waiting for the
+ * index lock, we should release the index lock, otherwise our request
+ * for ShareLock on the relation can end up in a deadlock.
+ */
+ init_rel_reopen_info(&rri[nrel++], index_p, InvalidOid,
+ ShareUpdateExclusiveLock, ShareLock);
+ }
+ unlock_and_close_relations(rri, nrel);
+ /*
+ * XXX It's not strictly necessary to lock the index here, but it's
+ * probably not worth teaching the "reopen API" about this special case.
+ */
+ reopen_relations(rri, nrel);
+
+ /* Switch back to the original lock. */
+ nrel = 0;
+ init_rel_reopen_info(&rri[nrel++], rel_p, InvalidOid,
+ ShareLock, ShareUpdateExclusiveLock);
+ if (index_p)
+ init_rel_reopen_info(&rri[nrel++], index_p, InvalidOid,
+ ShareLock, ShareUpdateExclusiveLock);
+ unlock_and_close_relations(rri, nrel);
+ reopen_relations(rri, nrel);
+ /* Make sure the reopened relcache entry is used, not the old one. */
+ rel = *rel_p;
+
+ /* Avoid logical decoding of other relations by this backend. */
+ clustered_rel_locator = rel->rd_locator;
+ if (OidIsValid(toastrelid))
+ {
+ Relation toastrel;
+
+ /* Avoid logical decoding of other TOAST relations. */
+ toastrel = table_open(toastrelid, AccessShareLock);
+ clustered_rel_toast_locator = toastrel->rd_locator;
+ table_close(toastrel, AccessShareLock);
+ }
+}
+
+/*
+ * Call this when done with CLUSTER CONCURRENTLY.
+ *
+ * 'error' tells whether the function is being called in order to handle
+ * error.
+ */
+static void
+end_concurrent_cluster(bool error)
+{
+ ClusteredRel key;
+ ClusteredRel *entry = NULL, *entry_toast = NULL;
+ Oid relid = clustered_rel;
+ Oid toastrelid = clustered_rel_toast;
+
+ /* Remove the relation from the hash if we managed to insert one. */
+ if (OidIsValid(clustered_rel))
+ {
+ memset(&key, 0, sizeof(key));
+ key.relid = clustered_rel;
+ key.dbid = MyDatabaseId;
+ LWLockAcquire(ClusteredRelsLock, LW_EXCLUSIVE);
+ entry = hash_search(ClusteredRelsHash, &key, HASH_REMOVE, NULL);
+
+ /*
+ * By clearing this variable we also disable
+ * cluster_before_shmem_exit_callback().
+ */
+ clustered_rel = InvalidOid;
+ }
+
+ /* Remove the TOAST relation if there is one. */
+ if (OidIsValid(clustered_rel_toast))
+ {
+ key.relid = clustered_rel_toast;
+ entry_toast = hash_search(ClusteredRelsHash, &key, HASH_REMOVE,
+ NULL);
+
+ clustered_rel_toast = InvalidOid;
+ }
+ LWLockRelease(ClusteredRelsLock);
+
+ /* Restore normal function of logical decoding. */
+ clustered_rel_locator.relNumber = InvalidOid;
+ clustered_rel_toast_locator.relNumber = InvalidOid;
+
+ /*
+ * On normal completion (!error), we should not really fail to remove the
+ * entry. But if it wasn't there for any reason, raise ERROR to make sure
+ * the transaction is aborted: if other transactions, while changing the
+ * contents of the relation, didn't know that CLUSTER CONCURRENTLY was in
+ * progress, they could have missed to WAL enough information, and thus we
+ * could have produced an inconsistent table contents.
+ *
+ * On the other hand, if we are already handling an error, there's no
+ * reason to worry about inconsistent contents of the new storage because
+ * the transaction is going to be rolled back anyway. Furthermore, by
+ * raising ERROR here we'd shadow the original error.
+ */
+ if (!error)
+ {
+ char *relname;
+
+ if (OidIsValid(relid) && entry == NULL)
+ {
+ relname = get_rel_name(relid);
+ if (!relname)
+ ereport(ERROR,
+ (errmsg("cache lookup failed for relation %u",
+ relid)));
+
+ ereport(ERROR,
+ (errmsg("relation \"%s\" not found among clustered relations",
+ relname)));
+ }
+
+ /*
+ * Likewise, the TOAST relation should not have disappeared.
+ */
+ if (OidIsValid(toastrelid) && entry_toast == NULL)
+ {
+ relname = get_rel_name(key.relid);
+ if (!relname)
+ ereport(ERROR,
+ (errmsg("cache lookup failed for relation %u",
+ key.relid)));
+
+ ereport(ERROR,
+ (errmsg("relation \"%s\" not found among clustered relations",
+ relname)));
+ }
+ }
+
+ /*
+ * Note: unlike begin_concurrent_cluster(), here we do not lock/unlock the
+ * relation: 1) On normal completion, the caller is already holding
+ * AccessExclusiveLock (till the end of the transaction), 2) on ERROR /
+ * FATAL, we try to do the cleanup asap, but the worst case is that other
+ * backends will write unnecessary information to WAL until they close the
+ * relation.
+ */
+}
+
+/*
+ * A wrapper to call end_concurrent_cluster() as a before_shmem_exit callback.
+ */
+static void
+cluster_before_shmem_exit_callback(int code, Datum arg)
+{
+ if (OidIsValid(clustered_rel) || OidIsValid(clustered_rel_toast))
+ end_concurrent_cluster(true);
+}
+
+/*
+ * Check if relation is currently being processed by CLUSTER CONCURRENTLY.
+ */
+bool
+is_concurrent_cluster_in_progress(Oid relid)
+{
+ ClusteredRel key, *entry;
+
+ memset(&key, 0, sizeof(key));
+ key.relid = relid;
+ key.dbid = MyDatabaseId;
+
+ LWLockAcquire(ClusteredRelsLock, LW_SHARED);
+ entry = (ClusteredRel *)
+ hash_search(ClusteredRelsHash, &key, HASH_FIND, NULL);
+ LWLockRelease(ClusteredRelsLock);
+
+ return entry != NULL;
+}
+
+/*
+ * Check if VACUUM FULL / CLUSTER CONCURRENTLY is already running for given
+ * relation, and if so, raise ERROR. The problem is that cluster_rel() needs
+ * to release its lock on the relation temporarily at some point, so our lock
+ * alone does not help. Commands that might break what cluster_rel() is doing
+ * should call this function first.
+ *
+ * Return without checking if lockmode allows for race conditions which would
+ * make the result meaningless. In that case, cluster_rel() itself should
+ * throw ERROR if the relation was changed by us in an incompatible
+ * way. However, if it managed to do most of its work by then, a lot of CPU
+ * time might be wasted.
+ */
+void
+check_for_concurrent_cluster(Oid relid, LOCKMODE lockmode)
+{
+ /*
+ * If the caller does not have a lock that conflicts with
+ * ShareUpdateExclusiveLock, the check makes little sense because the
+ * VACUUM FULL / CLUSTER CONCURRENTLY can start anytime after the check.
+ */
+ if (lockmode < ShareUpdateExclusiveLock)
+ return;
+
+ if (is_concurrent_cluster_in_progress(relid))
+ ereport(ERROR,
+ (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
+ errmsg(CLUSTER_IN_PROGRESS_MESSAGE,
+ get_rel_name(relid))));
+
+}
+
+/*
+ * Check if relation is eligible for CLUSTER CONCURRENTLY and retrieve the
+ * catalog state to be passed later to check_catalog_changes.
+ *
+ * Caller is supposed to hold (at least) ShareUpdateExclusiveLock on the
+ * relation.
+ */
+static CatalogState *
+get_catalog_state(Relation rel, bool is_vacuum)
+{
+ CatalogState *result = palloc_object(CatalogState);
+ List *ind_oids;
+ ListCell *lc;
+ int ninds, i;
+ char relpersistence = rel->rd_rel->relpersistence;
+ char replident = rel->rd_rel->relreplident;
+ Oid ident_idx = RelationGetReplicaIndex(rel);
+ TupleDesc td_src = RelationGetDescr(rel);
+
+ /*
+ * While gathering the catalog information, check if there is a reason not
+ * to proceed.
+ *
+ * This function was already called, but the relation was unlocked since
+ * (see begin_concurrent_cluster()). check_catalog_changes() should catch
+ * any "disruptive" changes in the future.
+ */
+ check_relation_is_clusterable_concurrently(rel, is_vacuum);
+
+ /* No index should be dropped while we are checking it. */
+ Assert(CheckRelationLockedByMe(rel, ShareUpdateExclusiveLock, true));
+
+ ind_oids = RelationGetIndexList(rel);
+ result->ninds = ninds = list_length(ind_oids);
+ result->ind_oids = palloc_array(Oid, ninds);
+ result->ind_tupdescs = palloc_array(TupleDesc, ninds);
+ i = 0;
+ foreach(lc, ind_oids)
+ {
+ Oid ind_oid = lfirst_oid(lc);
+ Relation index;
+ TupleDesc td_ind_src, td_ind_dst;
+
+ /*
+ * Weaker lock should be o.k. for the index, but this one should not
+ * break anything either.
+ */
+ index = index_open(ind_oid, ShareUpdateExclusiveLock);
+
+ result->ind_oids[i] = RelationGetRelid(index);
+ td_ind_src = RelationGetDescr(index);
+ td_ind_dst = palloc(TupleDescSize(td_ind_src));
+ TupleDescCopy(td_ind_dst, td_ind_src);
+ result->ind_tupdescs[i] = td_ind_dst;
+ i++;
+
+ index_close(index, ShareUpdateExclusiveLock);
+ }
+
+ /* Fill-in the relation info. */
+ result->tupdesc = palloc(TupleDescSize(td_src));
+ TupleDescCopy(result->tupdesc, td_src);
+ result->relpersistence = relpersistence;
+ result->replident = replident;
+ result->replidindex = ident_idx;
+
+ return result;
+}
+
+static void
+free_catalog_state(CatalogState *state)
+{
+ /* We are only interested in indexes. */
+ if (state->ninds == 0)
+ return;
+
+ for (int i = 0; i < state->ninds; i++)
+ FreeTupleDesc(state->ind_tupdescs[i]);
+
+ FreeTupleDesc(state->tupdesc);
+ pfree(state->ind_oids);
+ pfree(state->ind_tupdescs);
+ pfree(state);
+}
+
+/*
+ * Raise ERROR if 'rel' changed in a way that does not allow further
+ * processing of CLUSTER CONCURRENTLY.
+ *
+ * Besides the relation's tuple descriptor, it's important to check indexes:
+ * concurrent change of index definition (can it happen in other way than
+ * dropping and re-creating the index, accidentally with the same OID?) can be
+ * a problem because we may already have the new index built. If an index was
+ * created or dropped concurrently, we'd fail to swap the index storage. In
+ * any case, we prefer to check the indexes early to get an explicit error
+ * message about the mismatch. Furthermore, the earlier we detect the change,
+ * the fewer CPU cycles we waste.
+ *
+ * Note that we do not check constraints because the transaction which changed
+ * them must have ensured that the existing tuples satisfy the new
+ * constraints. If any DML commands were necessary for that, we will simply
+ * decode them from WAL and apply them to the new storage.
+ *
+ * Caller is supposed to hold (at least) ShareUpdateExclusiveLock on the
+ * relation.
+ */
+static void
+check_catalog_changes(Relation rel, CatalogState *cat_state)
+{
+ Oid reltoastrelid = rel->rd_rel->reltoastrelid;
+ List *ind_oids;
+ ListCell *lc;
+ LOCKMODE lockmode;
+ Oid ident_idx;
+ TupleDesc td, td_cp;
+
+ /* First, check the relation info. */
+
+ /* TOAST is not easy to change, but check. */
+ if (reltoastrelid != clustered_rel_toast)
+ ereport(ERROR,
+ errmsg("TOAST relation of relation \"%s\" changed by another transaction",
+ RelationGetRelationName(rel)));
+
+ /*
+ * Likewise, check_for_concurrent_cluster() should prevent others from
+ * changing the relation file concurrently, but it's our responsibility to
+ * avoid data loss. (The original locators are stored outside cat_state,
+ * but the check belongs to this function.)
+ */
+ if (!RelFileLocatorEquals(rel->rd_locator, clustered_rel_locator))
+ ereport(ERROR,
+ (errmsg("file of relation \"%s\" changed by another transaction",
+ RelationGetRelationName(rel))));
+ if (OidIsValid(reltoastrelid))
+ {
+ Relation toastrel;
+
+ toastrel = table_open(reltoastrelid, AccessShareLock);
+ if (!RelFileLocatorEquals(toastrel->rd_locator,
+ clustered_rel_toast_locator))
+ ereport(ERROR,
+ (errmsg("file of relation \"%s\" changed by another transaction",
+ RelationGetRelationName(toastrel))));
+ table_close(toastrel, AccessShareLock);
+ }
+
+ if (rel->rd_rel->relpersistence != cat_state->relpersistence)
+ ereport(ERROR,
+ errmsg("persistence of relation \"%s\" changed by another transaction",
+ RelationGetRelationName(rel)));
+
+ if (cat_state->replident != rel->rd_rel->relreplident)
+ ereport(ERROR,
+ errmsg("replica identity of relation \"%s\" changed by another transaction",
+ RelationGetRelationName(rel)));
+
+ ident_idx = RelationGetReplicaIndex(rel);
+ if (ident_idx == InvalidOid && rel->rd_pkindex != InvalidOid)
+ ident_idx = rel->rd_pkindex;
+ if (cat_state->replidindex != ident_idx)
+ ereport(ERROR,
+ errmsg("identity index of relation \"%s\" changed by another transaction",
+ RelationGetRelationName(rel)));
+
+ /*
+ * As cat_state contains a copy (which has the constraint info cleared),
+ * create a temporary copy for the comparison.
+ */
+ td = RelationGetDescr(rel);
+ td_cp = palloc(TupleDescSize(td));
+ TupleDescCopy(td_cp, td);
+ if (!equalTupleDescs(cat_state->tupdesc, td_cp))
+ ereport(ERROR,
+ errmsg("definition of relation \"%s\" changed by another transaction",
+ RelationGetRelationName(rel)));
+ FreeTupleDesc(td_cp);
+
+ /* Now we are only interested in indexes. */
+ if (cat_state->ninds == 0)
+ return;
+
+ /* No index should be dropped while we are checking the relation. */
+ lockmode = ShareUpdateExclusiveLock;
+ Assert(CheckRelationLockedByMe(rel, lockmode, true));
+
+ ind_oids = RelationGetIndexList(rel);
+ if (list_length(ind_oids) != cat_state->ninds)
+ goto failed_index;
+
+ foreach(lc, ind_oids)
+ {
+ Oid ind_oid = lfirst_oid(lc);
+ int i;
+ TupleDesc tupdesc;
+ Relation index;
+
+ /* Find the index in cat_state. */
+ for (i = 0; i < cat_state->ninds; i++)
+ {
+ if (cat_state->ind_oids[i] == ind_oid)
+ break;
+ }
+ /*
+ * OID not found, i.e. the index was replaced by another one. XXX
+ * Should we yet try to find if an index having the desired tuple
+ * descriptor exists? Or should we always look for the tuple
+ * descriptor and not use OIDs at all?
+ */
+ if (i == cat_state->ninds)
+ goto failed_index;
+
+ /* Check the tuple descriptor. */
+ index = try_index_open(ind_oid, lockmode);
+ if (index == NULL)
+ goto failed_index;
+ tupdesc = RelationGetDescr(index);
+ if (!equalTupleDescs(cat_state->ind_tupdescs[i], tupdesc))
+ goto failed_index;
+ index_close(index, lockmode);
+ }
+
+ return;
+
+failed_index:
+ ereport(ERROR,
+ (errmsg("index(es) of relation \"%s\" changed by another transaction",
+ RelationGetRelationName(rel))));
+}
+
+/*
+ * This function is much like pg_create_logical_replication_slot() except that
+ * the new slot is neither released (if anyone else could read changes from
+ * our slot, we could miss changes other backends do while we copy the
+ * existing data into temporary table), nor persisted (it's easier to handle
+ * crash by restarting all the work from scratch).
+ *
+ * XXX Even though CreateInitDecodingContext() does not set state to
+ * RS_PERSISTENT, it does write the slot to disk. We rely on
+ * RestoreSlotFromDisk() to delete ephemeral slots during startup. (Both ERROR
+ * and FATAL should lead to cleanup even before the cluster goes down.)
+ */
+static LogicalDecodingContext *
+setup_logical_decoding(Oid relid, const char *slotname, TupleDesc tupdesc)
+{
+ LogicalDecodingContext *ctx;
+ ClusterDecodingState *dstate;
+
+ /* RS_TEMPORARY so that the slot gets cleaned up on ERROR. */
+ ReplicationSlotCreate(slotname, true, RS_TEMPORARY, false, false, false);
+
+ /*
+ * Neither prepare_write nor do_write callback nor update_progress is
+ * useful for us.
+ *
+ * Regarding the value of need_full_snapshot, we pass false because the
+ * table we are processing is present in ClusteredRelsHash and therefore,
+ * regarding logical decoding, treated like a catalog.
+ */
+ ctx = CreateInitDecodingContext(REPL_PLUGIN_NAME,
+ NIL,
+ false,
+ InvalidXLogRecPtr,
+ XL_ROUTINE(.page_read = read_local_xlog_page,
+ .segment_open = wal_segment_open,
+ .segment_close = wal_segment_close),
+ NULL, NULL, NULL);
+
+ /*
+ * We don't have control on setting fast_forward, so at least check it.
+ */
+ Assert(!ctx->fast_forward);
+
+ DecodingContextFindStartpoint(ctx);
+
+ /* Some WAL records should have been read. */
+ Assert(ctx->reader->EndRecPtr != InvalidXLogRecPtr);
+
+ XLByteToSeg(ctx->reader->EndRecPtr, cluster_current_segment,
+ wal_segment_size);
+
+ /*
+ * Setup structures to store decoded changes.
+ */
+ dstate = palloc0(sizeof(ClusterDecodingState));
+ dstate->relid = relid;
+ dstate->tstore = tuplestore_begin_heap(false, false,
+ maintenance_work_mem);
+ dstate->tupdesc = tupdesc;
+
+ /* Initialize the descriptor to store the changes ... */
+ dstate->tupdesc_change = CreateTemplateTupleDesc(1);
+
+ TupleDescInitEntry(dstate->tupdesc_change, 1, NULL, BYTEAOID, -1, 0);
+ /* ... as well as the corresponding slot. */
+ dstate->tsslot = MakeSingleTupleTableSlot(dstate->tupdesc_change,
+ &TTSOpsMinimalTuple);
+
+ dstate->resowner = ResourceOwnerCreate(CurrentResourceOwner,
+ "logical decoding");
+
+ ctx->output_writer_private = dstate;
+ return ctx;
+}
+
+/*
+ * Retrieve tuple from ConcurrentChange structure.
+ *
+ * The input data starts with the structure but it might not be appropriately
+ * aligned.
+ */
+static HeapTuple
+get_changed_tuple(char *change)
+{
+ HeapTupleData tup_data;
+ HeapTuple result;
+ char *src;
+
+ /*
+ * Ensure alignment before accessing the fields. (This is why we can't use
+ * heap_copytuple() instead of this function.)
+ */
+ src = change + offsetof(ConcurrentChange, tup_data);
+ memcpy(&tup_data, src, sizeof(HeapTupleData));
+
+ result = (HeapTuple) palloc(HEAPTUPLESIZE + tup_data.t_len);
+ memcpy(result, &tup_data, sizeof(HeapTupleData));
+ result->t_data = (HeapTupleHeader) ((char *) result + HEAPTUPLESIZE);
+ src = change + SizeOfConcurrentChange;
+ memcpy(result->t_data, src, result->t_len);
+
+ return result;
+}
+
+/*
+ * Decode logical changes from the WAL sequence up to end_of_wal.
+ */
+void
+cluster_decode_concurrent_changes(LogicalDecodingContext *ctx,
+ XLogRecPtr end_of_wal)
+{
+ ClusterDecodingState *dstate;
+ ResourceOwner resowner_old;
+ PgBackendProgress progress;
+
+ /*
+ * Invalidate the "present" cache before moving to "(recent) history".
+ */
+ InvalidateSystemCaches();
+
+ dstate = (ClusterDecodingState *) ctx->output_writer_private;
+ resowner_old = CurrentResourceOwner;
+ CurrentResourceOwner = dstate->resowner;
+
+ /*
+ * reorderbuffer.c uses internal subtransaction, whose abort ends the
+ * command progress reporting. Save the status here so we can restore when
+ * done with the decoding.
+ */
+ memcpy(&progress, &MyBEEntry->st_progress, sizeof(PgBackendProgress));
+
+ PG_TRY();
+ {
+ while (ctx->reader->EndRecPtr < end_of_wal)
+ {
+ XLogRecord *record;
+ XLogSegNo segno_new;
+ char *errm = NULL;
+ XLogRecPtr end_lsn;
+
+ record = XLogReadRecord(ctx->reader, &errm);
+ if (errm)
+ elog(ERROR, "%s", errm);
+
+ if (record != NULL)
+ LogicalDecodingProcessRecord(ctx, ctx->reader);
+
+ /*
+ * If WAL segment boundary has been crossed, inform the decoding
+ * system that the catalog_xmin can advance. (We can confirm more
+ * often, but a filling a single WAL segment should not take much
+ * time.)
+ */
+ end_lsn = ctx->reader->EndRecPtr;
+ XLByteToSeg(end_lsn, segno_new, wal_segment_size);
+ if (segno_new != cluster_current_segment)
+ {
+ LogicalConfirmReceivedLocation(end_lsn);
+ elog(DEBUG1, "cluster: confirmed receive location %X/%X",
+ (uint32) (end_lsn >> 32), (uint32) end_lsn);
+ cluster_current_segment = segno_new;
+ }
+
+ CHECK_FOR_INTERRUPTS();
+ }
+ InvalidateSystemCaches();
+ CurrentResourceOwner = resowner_old;
+ }
+ PG_CATCH();
+ {
+ InvalidateSystemCaches();
+ CurrentResourceOwner = resowner_old;
+ PG_RE_THROW();
+ }
+ PG_END_TRY();
+
+ /* Restore the progress reporting status. */
+ pgstat_progress_restore_state(&progress);
+}
+
+/*
+ * Apply changes that happened during the initial load.
+ *
+ * Scan key is passed by caller, so it does not have to be constructed
+ * multiple times. Key entries have all fields initialized, except for
+ * sk_argument.
+ */
+static void
+apply_concurrent_changes(ClusterDecodingState *dstate, Relation rel,
+ ScanKey key, int nkeys, IndexInsertState *iistate)
+{
+ TupleTableSlot *index_slot, *ident_slot;
+ HeapTuple tup_old = NULL;
+
+ if (dstate->nchanges == 0)
+ return;
+
+ /* TupleTableSlot is needed to pass the tuple to ExecInsertIndexTuples(). */
+ index_slot = MakeSingleTupleTableSlot(dstate->tupdesc, &TTSOpsHeapTuple);
+ iistate->econtext->ecxt_scantuple = index_slot;
+
+ /* A slot to fetch tuples from identity index. */
+ ident_slot = table_slot_create(rel, NULL);
+
+ while (tuplestore_gettupleslot(dstate->tstore, true, false,
+ dstate->tsslot))
+ {
+ bool shouldFree;
+ HeapTuple tup_change,
+ tup,
+ tup_exist;
+ char *change_raw, *src;
+ ConcurrentChange change;
+ bool isnull[1];
+ Datum values[1];
+
+ CHECK_FOR_INTERRUPTS();
+
+ /* Get the change from the single-column tuple. */
+ tup_change = ExecFetchSlotHeapTuple(dstate->tsslot, false, &shouldFree);
+ heap_deform_tuple(tup_change, dstate->tupdesc_change, values, isnull);
+ Assert(!isnull[0]);
+
+ /* Make sure we access aligned data. */
+ change_raw = (char *) DatumGetByteaP(values[0]);
+ src = (char *) VARDATA(change_raw);
+ memcpy(&change, src, SizeOfConcurrentChange);
+
+ /* TRUNCATE change contains no tuple, so process it separately. */
+ if (change.kind == CHANGE_TRUNCATE)
+ {
+ /*
+ * All the things that ExecuteTruncateGuts() does (such as firing
+ * triggers or handling the DROP_CASCADE behavior) should have
+ * taken place on the source relation. Thus we only do the actual
+ * truncation of the new relation (and its indexes).
+ */
+ heap_truncate_one_rel(rel);
+
+ pfree(tup_change);
+ continue;
+ }
+
+ /*
+ * Extract the tuple from the change. The tuple is copied here because
+ * it might be assigned to 'tup_old', in which case it needs to
+ * survive into the next iteration.
+ */
+ tup = get_changed_tuple(src);
+
+ if (change.kind == CHANGE_UPDATE_OLD)
+ {
+ Assert(tup_old == NULL);
+ tup_old = tup;
+ }
+ else if (change.kind == CHANGE_INSERT)
+ {
+ Assert(tup_old == NULL);
+
+ apply_concurrent_insert(rel, &change, tup, iistate, index_slot);
+
+ pfree(tup);
+ }
+ else if (change.kind == CHANGE_UPDATE_NEW ||
+ change.kind == CHANGE_DELETE)
+ {
+ IndexScanDesc ind_scan = NULL;
+ HeapTuple tup_key;
+
+ if (change.kind == CHANGE_UPDATE_NEW)
+ {
+ tup_key = tup_old != NULL ? tup_old : tup;
+ }
+ else
+ {
+ Assert(tup_old == NULL);
+ tup_key = tup;
+ }
+
+ /*
+ * Find the tuple to be updated or deleted.
+ */
+ tup_exist = find_target_tuple(rel, key, nkeys, tup_key,
+ iistate, ident_slot, &ind_scan);
+ if (tup_exist == NULL)
+ elog(ERROR, "Failed to find target tuple");
+
+ if (change.kind == CHANGE_UPDATE_NEW)
+ apply_concurrent_update(rel, tup, tup_exist, &change, iistate,
+ index_slot);
+ else
+ apply_concurrent_delete(rel, tup_exist, &change);
+
+ if (tup_old != NULL)
+ {
+ pfree(tup_old);
+ tup_old = NULL;
+ }
+
+ pfree(tup);
+ index_endscan(ind_scan);
+ }
+ else
+ elog(ERROR, "Unrecognized kind of change: %d", change.kind);
+
+ /* If there's any change, make it visible to the next iteration. */
+ if (change.kind != CHANGE_UPDATE_OLD)
+ {
+ CommandCounterIncrement();
+ UpdateActiveSnapshotCommandId();
+ }
+
+ /* TTSOpsMinimalTuple has .get_heap_tuple==NULL. */
+ Assert(shouldFree);
+ pfree(tup_change);
+ }
+
+ tuplestore_clear(dstate->tstore);
+ dstate->nchanges = 0;
+
+ /* Cleanup. */
+ ExecDropSingleTupleTableSlot(index_slot);
+ ExecDropSingleTupleTableSlot(ident_slot);
+}
+
+static void
+apply_concurrent_insert(Relation rel, ConcurrentChange *change, HeapTuple tup,
+ IndexInsertState *iistate, TupleTableSlot *index_slot)
+{
+ List *recheck;
+
+
+ heap_insert(rel, tup, GetCurrentCommandId(true), HEAP_INSERT_NO_LOGICAL, NULL);
+
+ /*
+ * Update indexes.
+ *
+ * In case functions in the index need the active snapshot and caller
+ * hasn't set one.
+ */
+ ExecStoreHeapTuple(tup, index_slot, false);
+ recheck = ExecInsertIndexTuples(iistate->rri,
+ index_slot,
+ iistate->estate,
+ false, /* update */
+ false, /* noDupErr */
+ NULL, /* specConflict */
+ NIL, /* arbiterIndexes */
+ false /* onlySummarizing */
+ );
+
+ /*
+ * If recheck is required, it must have been preformed on the source
+ * relation by now. (All the logical changes we process here are already
+ * committed.)
+ */
+ list_free(recheck);
+
+ pgstat_progress_incr_param(PROGRESS_CLUSTER_HEAP_TUPLES_INSERTED, 1);
+}
+
+static void
+apply_concurrent_update(Relation rel, HeapTuple tup, HeapTuple tup_target,
+ ConcurrentChange *change, IndexInsertState *iistate,
+ TupleTableSlot *index_slot)
+{
+ List *recheck;
+ TU_UpdateIndexes update_indexes;
+
+ /*
+ * Write the new tuple into the new heap. ('tup' gets the TID assigned
+ * here.)
+ */
+ simple_heap_update(rel, &tup_target->t_self, tup, &update_indexes);
+
+ ExecStoreHeapTuple(tup, index_slot, false);
+
+ if (update_indexes != TU_None)
+ {
+ recheck = ExecInsertIndexTuples(iistate->rri,
+ index_slot,
+ iistate->estate,
+ true, /* update */
+ false, /* noDupErr */
+ NULL, /* specConflict */
+ NIL, /* arbiterIndexes */
+ /* onlySummarizing */
+ update_indexes == TU_Summarizing);
+ list_free(recheck);
+ }
+
+ pgstat_progress_incr_param(PROGRESS_CLUSTER_HEAP_TUPLES_UPDATED, 1);
+}
+
+static void
+apply_concurrent_delete(Relation rel, HeapTuple tup_target,
+ ConcurrentChange *change)
+{
+ simple_heap_delete(rel, &tup_target->t_self);
+
+ pgstat_progress_incr_param(PROGRESS_CLUSTER_HEAP_TUPLES_DELETED, 1);
+}
+
+/*
+ * Find the tuple to be updated or deleted.
+ *
+ * 'key' is a pre-initialized scan key, into which the function will put the
+ * key values.
+ *
+ * 'tup_key' is a tuple containing the key values for the scan.
+ *
+ * On exit,'*scan_p' contains the scan descriptor used. The caller must close
+ * it when he no longer needs the tuple returned.
+ */
+static HeapTuple
+find_target_tuple(Relation rel, ScanKey key, int nkeys, HeapTuple tup_key,
+ IndexInsertState *iistate,
+ TupleTableSlot *ident_slot, IndexScanDesc *scan_p)
+{
+ IndexScanDesc scan;
+ Form_pg_index ident_form;
+ int2vector *ident_indkey;
+ HeapTuple result = NULL;
+
+ scan = index_beginscan(rel, iistate->ident_index, GetActiveSnapshot(),
+ nkeys, 0);
+ *scan_p = scan;
+ index_rescan(scan, key, nkeys, NULL, 0);
+
+ /* Info needed to retrieve key values from heap tuple. */
+ ident_form = iistate->ident_index->rd_index;
+ ident_indkey = &ident_form->indkey;
+
+ /* Use the incoming tuple to finalize the scan key. */
+ for (int i = 0; i < scan->numberOfKeys; i++)
+ {
+ ScanKey entry;
+ bool isnull;
+ int16 attno_heap;
+
+ entry = &scan->keyData[i];
+ attno_heap = ident_indkey->values[i];
+ entry->sk_argument = heap_getattr(tup_key,
+ attno_heap,
+ rel->rd_att,
+ &isnull);
+ Assert(!isnull);
+ }
+ if (index_getnext_slot(scan, ForwardScanDirection, ident_slot))
+ {
+ bool shouldFree;
+
+ result = ExecFetchSlotHeapTuple(ident_slot, false, &shouldFree);
+ /* TTSOpsBufferHeapTuple has .get_heap_tuple != NULL. */
+ Assert(!shouldFree);
+ }
+
+ return result;
+}
+
+/*
+ * Decode and apply concurrent changes.
+ *
+ * Pass rel_src iff its reltoastrelid is needed.
+ */
+static void
+process_concurrent_changes(LogicalDecodingContext *ctx, XLogRecPtr end_of_wal,
+ Relation rel_dst, Relation rel_src, ScanKey ident_key,
+ int ident_key_nentries, IndexInsertState *iistate)
+{
+ ClusterDecodingState *dstate;
+
+ pgstat_progress_update_param(PROGRESS_CLUSTER_PHASE,
+ PROGRESS_CLUSTER_PHASE_CATCH_UP);
+
+ dstate = (ClusterDecodingState *) ctx->output_writer_private;
+
+ cluster_decode_concurrent_changes(ctx, end_of_wal);
+
+ if (dstate->nchanges == 0)
+ return;
+
+ PG_TRY();
+ {
+ /*
+ * Make sure that TOAST values can eventually be accessed via the old
+ * relation - see comment in copy_table_data().
+ */
+ if (rel_src)
+ rel_dst->rd_toastoid = rel_src->rd_rel->reltoastrelid;
+
+ apply_concurrent_changes(dstate, rel_dst, ident_key,
+ ident_key_nentries, iistate);
+ }
+ PG_FINALLY();
+ {
+ if (rel_src)
+ rel_dst->rd_toastoid = InvalidOid;
+ }
+ PG_END_TRY();
+}
+
+static IndexInsertState *
+get_index_insert_state(Relation relation, Oid ident_index_id)
+{
+ EState *estate;
+ int i;
+ IndexInsertState *result;
+
+ result = (IndexInsertState *) palloc0(sizeof(IndexInsertState));
+ estate = CreateExecutorState();
+ result->econtext = GetPerTupleExprContext(estate);
+
+ result->rri = (ResultRelInfo *) palloc(sizeof(ResultRelInfo));
+ InitResultRelInfo(result->rri, relation, 0, 0, 0);
+ ExecOpenIndices(result->rri, false);
+
+ /*
+ * Find the relcache entry of the identity index so that we spend no extra
+ * effort to open / close it.
+ */
+ for (i = 0; i < result->rri->ri_NumIndices; i++)
+ {
+ Relation ind_rel;
+
+ ind_rel = result->rri->ri_IndexRelationDescs[i];
+ if (ind_rel->rd_id == ident_index_id)
+ result->ident_index = ind_rel;
+ }
+ if (result->ident_index == NULL)
+ elog(ERROR, "Failed to open identity index");
+
+ /* Only initialize fields needed by ExecInsertIndexTuples(). */
+ result->estate = estate;
+
+ return result;
+}
+
+/*
+ * Build scan key to process logical changes.
+ */
+static ScanKey
+build_identity_key(Oid ident_idx_oid, Relation rel_src, int *nentries)
+{
+ Relation ident_idx_rel;
+ Form_pg_index ident_idx;
+ int n,
+ i;
+ ScanKey result;
+
+ Assert(OidIsValid(ident_idx_oid));
+ ident_idx_rel = index_open(ident_idx_oid, AccessShareLock);
+ ident_idx = ident_idx_rel->rd_index;
+ n = ident_idx->indnatts;
+ result = (ScanKey) palloc(sizeof(ScanKeyData) * n);
+ for (i = 0; i < n; i++)
+ {
+ ScanKey entry;
+ int16 relattno;
+ Form_pg_attribute att;
+ Oid opfamily,
+ opcintype,
+ opno,
+ opcode;
+
+ entry = &result[i];
+ relattno = ident_idx->indkey.values[i];
+ if (relattno >= 1)
+ {
+ TupleDesc desc;
+
+ desc = rel_src->rd_att;
+ att = TupleDescAttr(desc, relattno - 1);
+ }
+ else
+ elog(ERROR, "Unexpected attribute number %d in index", relattno);
+
+ opfamily = ident_idx_rel->rd_opfamily[i];
+ opcintype = ident_idx_rel->rd_opcintype[i];
+ opno = get_opfamily_member(opfamily, opcintype, opcintype,
+ BTEqualStrategyNumber);
+
+ if (!OidIsValid(opno))
+ elog(ERROR, "Failed to find = operator for type %u", opcintype);
+
+ opcode = get_opcode(opno);
+ if (!OidIsValid(opcode))
+ elog(ERROR, "Failed to find = operator for operator %u", opno);
+
+ /* Initialize everything but argument. */
+ ScanKeyInit(entry,
+ i + 1,
+ BTEqualStrategyNumber, opcode,
+ (Datum) NULL);
+ entry->sk_collation = att->attcollation;
+ }
+ index_close(ident_idx_rel, AccessShareLock);
+
+ *nentries = n;
+ return result;
+}
+
+static void
+free_index_insert_state(IndexInsertState *iistate)
+{
+ ExecCloseIndices(iistate->rri);
+ FreeExecutorState(iistate->estate);
+ pfree(iistate->rri);
+ pfree(iistate);
+}
+
+static void
+cleanup_logical_decoding(LogicalDecodingContext *ctx)
+{
+ ClusterDecodingState *dstate;
+
+ dstate = (ClusterDecodingState *) ctx->output_writer_private;
+
+ ExecDropSingleTupleTableSlot(dstate->tsslot);
+ FreeTupleDesc(dstate->tupdesc_change);
+ FreeTupleDesc(dstate->tupdesc);
+ tuplestore_end(dstate->tstore);
+
+ FreeDecodingContext(ctx);
+}
+
+/*
+ * The final steps of rebuild_relation() for concurrent processing.
+ *
+ * On entry, NewHeap is locked in AccessExclusiveLock mode. OldHeap and its
+ * clustering index (if one is passed) are still locked in a mode that allows
+ * concurrent data changes. On exit, both tables and their indexes are closed,
+ * but locked in AccessExclusiveLock mode.
+ */
+static void
+rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
+ Relation cl_index,
+ CatalogState *cat_state,
+ LogicalDecodingContext *ctx,
+ bool swap_toast_by_content,
+ TransactionId frozenXid,
+ MultiXactId cutoffMulti)
+{
+ LOCKMODE lockmode_old PG_USED_FOR_ASSERTS_ONLY;
+ List *ind_oids_new;
+ Oid old_table_oid = RelationGetRelid(OldHeap);
+ Oid new_table_oid = RelationGetRelid(NewHeap);
+ List *ind_oids_old = RelationGetIndexList(OldHeap);
+ ListCell *lc, *lc2;
+ char relpersistence;
+ bool is_system_catalog;
+ Oid ident_idx_old, ident_idx_new;
+ IndexInsertState *iistate;
+ ScanKey ident_key;
+ int ident_key_nentries;
+ XLogRecPtr wal_insert_ptr, end_of_wal;
+ char dummy_rec_data = '\0';
+ RelReopenInfo *rri = NULL;
+ int nrel;
+ Relation *ind_refs_all, *ind_refs_p;
+
+ /* Like in cluster_rel(). */
+ lockmode_old = ShareUpdateExclusiveLock;
+ Assert(CheckRelationLockedByMe(OldHeap, lockmode_old, false));
+ Assert(cl_index == NULL ||
+ CheckRelationLockedByMe(cl_index, lockmode_old, false));
+ /* This is expected from the caller. */
+ Assert(CheckRelationLockedByMe(NewHeap, AccessExclusiveLock, false));
+
+ ident_idx_old = RelationGetReplicaIndex(OldHeap);
+
+ /*
+ * Unlike the exclusive case, we build new indexes for the new relation
+ * rather than swapping the storage and reindexing the old relation. The
+ * point is that the index build can take some time, so we do it before we
+ * get AccessExclusiveLock on the old heap and therefore we cannot swap
+ * the heap storage yet.
+ *
+ * index_create() will lock the new indexes using AccessExclusiveLock
+ * creation - no need to change that.
+ */
+ ind_oids_new = build_new_indexes(NewHeap, OldHeap, ind_oids_old);
+
+ /*
+ * Processing shouldn't start w/o valid identity index.
+ */
+ Assert(OidIsValid(ident_idx_old));
+
+ /* Find "identity index" on the new relation. */
+ ident_idx_new = InvalidOid;
+ forboth(lc, ind_oids_old, lc2, ind_oids_new)
+ {
+ Oid ind_old = lfirst_oid(lc);
+ Oid ind_new = lfirst_oid(lc2);
+
+ if (ident_idx_old == ind_old)
+ {
+ ident_idx_new = ind_new;
+ break;
+ }
+ }
+ if (!OidIsValid(ident_idx_new))
+ /*
+ * Should not happen, given our lock on the old relation.
+ */
+ ereport(ERROR,
+ (errmsg("Identity index missing on the new relation")));
+
+ /* Executor state to update indexes. */
+ iistate = get_index_insert_state(NewHeap, ident_idx_new);
+
+ /*
+ * Build scan key that we'll use to look for rows to be updated / deleted
+ * during logical decoding.
+ */
+ ident_key = build_identity_key(ident_idx_new, OldHeap, &ident_key_nentries);
+
+ /*
+ * Flush all WAL records inserted so far (possibly except for the last
+ * incomplete page, see GetInsertRecPtr), to minimize the amount of data
+ * we need to flush while holding exclusive lock on the source table.
+ */
+ wal_insert_ptr = GetInsertRecPtr();
+ XLogFlush(wal_insert_ptr);
+ end_of_wal = GetFlushRecPtr(NULL);
+
+ /*
+ * Apply concurrent changes first time, to minimize the time we need to
+ * hold AccessExclusiveLock. (Quite some amount of WAL could have been
+ * written during the data copying and index creation.)
+ */
+ process_concurrent_changes(ctx, end_of_wal, NewHeap,
+ swap_toast_by_content ? OldHeap : NULL,
+ ident_key, ident_key_nentries, iistate);
+
+ /*
+ * Release the locks that allowed concurrent data changes, in order to
+ * acquire the AccessExclusiveLock.
+ */
+ nrel = 0;
+ /*
+ * We unlock the old relation (and its clustering index), but then we will
+ * lock the relation and *all* its indexes because we want to swap their
+ * storage.
+ *
+ * (NewHeap is already locked, as well as its indexes.)
+ */
+ rri = palloc_array(RelReopenInfo, 1 + list_length(ind_oids_old));
+ init_rel_reopen_info(&rri[nrel++], &OldHeap, InvalidOid,
+ ShareUpdateExclusiveLock, AccessExclusiveLock);
+ /* References to the re-opened indexes will be stored in this array. */
+ ind_refs_all = palloc_array(Relation, list_length(ind_oids_old));
+ ind_refs_p = ind_refs_all;
+ /* The clustering index is a special case. */
+ if (cl_index)
+ {
+ *ind_refs_p = cl_index;
+ init_rel_reopen_info(&rri[nrel], ind_refs_p, InvalidOid,
+ ShareUpdateExclusiveLock, AccessExclusiveLock);
+ nrel++;
+ ind_refs_p++;
+ }
+ /*
+ * Initialize also the entries for the other indexes (currently unlocked)
+ * because we will have to lock them.
+ */
+ foreach(lc, ind_oids_old)
+ {
+ Oid ind_oid;
+
+ ind_oid = lfirst_oid(lc);
+ /* Clustering index is already in the array, or there is none. */
+ if (cl_index && RelationGetRelid(cl_index) == ind_oid)
+ continue;
+
+ Assert(nrel < (1 + list_length(ind_oids_old)));
+
+ *ind_refs_p = NULL;
+ init_rel_reopen_info(&rri[nrel],
+ /*
+ * In this special case we do not have the
+ * relcache reference, use OID instead.
+ */
+ ind_refs_p,
+ ind_oid,
+ NoLock, /* Nothing to unlock. */
+ AccessExclusiveLock);
+
+ nrel++;
+ ind_refs_p++;
+ }
+ /* Perform the actual unlocking and re-locking. */
+ unlock_and_close_relations(rri, nrel);
+ reopen_relations(rri, nrel);
+
+ /*
+ * In addition, lock the OldHeap's TOAST relation that we skipped for the
+ * CONCURRENTLY option in copy_table_data(). This lock will be needed to
+ * swap the relation files.
+ */
+ if (OidIsValid(OldHeap->rd_rel->reltoastrelid))
+ LockRelationOid(OldHeap->rd_rel->reltoastrelid, AccessExclusiveLock);
+
+ /*
+ * Check if the new indexes match the old ones, i.e. no changes occurred
+ * while OldHeap was unlocked.
+ *
+ * XXX It's probably not necessary to check the relation tuple descriptor
+ * here because the logical decoding was already active when we released
+ * the lock, and thus the corresponding data changes won't be lost.
+ * However processing of those changes might take a lot of time.
+ */
+ check_catalog_changes(OldHeap, cat_state);
+
+ /*
+ * Tuples and pages of the old heap will be gone, but the heap will stay.
+ */
+ TransferPredicateLocksToHeapRelation(OldHeap);
+ /* The same for indexes. */
+ for (int i = 0; i < (nrel - 1); i++)
+ {
+ Relation index = ind_refs_all[i];
+
+ TransferPredicateLocksToHeapRelation(index);
+
+ /*
+ * References to indexes on the old relation are not needed anymore,
+ * however locks stay till the end of the transaction.
+ */
+ index_close(index, NoLock);
+ }
+ pfree(ind_refs_all);
+
+ /*
+ * Flush anything we see in WAL, to make sure that all changes committed
+ * while we were waiting for the exclusive lock are available for
+ * decoding. This should not be necessary if all backends had
+ * synchronous_commit set, but we can't rely on this setting.
+ *
+ * Unfortunately, GetInsertRecPtr() may lag behind the actual insert
+ * position, and GetLastImportantRecPtr() points at the start of the last
+ * record rather than at the end. Thus the simplest way to determine the
+ * insert position is to insert a dummy record and use its LSN.
+ *
+ * XXX Consider using GetLastImportantRecPtr() and adding the size of the
+ * last record (plus the total size of all the page headers the record
+ * spans)?
+ */
+ XLogBeginInsert();
+ XLogRegisterData(&dummy_rec_data, 1);
+ wal_insert_ptr = XLogInsert(RM_XLOG_ID, XLOG_NOOP);
+ XLogFlush(wal_insert_ptr);
+ end_of_wal = GetFlushRecPtr(NULL);
+
+ /* Apply the concurrent changes again. */
+ process_concurrent_changes(ctx, end_of_wal, NewHeap,
+ swap_toast_by_content ? OldHeap : NULL,
+ ident_key, ident_key_nentries, iistate);
+
+ /* Remember info about rel before closing OldHeap */
+ relpersistence = OldHeap->rd_rel->relpersistence;
+ is_system_catalog = IsSystemRelation(OldHeap);
+
+ pgstat_progress_update_param(PROGRESS_CLUSTER_PHASE,
+ PROGRESS_CLUSTER_PHASE_SWAP_REL_FILES);
+
+ forboth(lc, ind_oids_old, lc2, ind_oids_new)
+ {
+ Oid ind_old = lfirst_oid(lc);
+ Oid ind_new = lfirst_oid(lc2);
+ Oid mapped_tables[4];
+
+ /* Zero out possible results from swapped_relation_files */
+ memset(mapped_tables, 0, sizeof(mapped_tables));
+
+ swap_relation_files(ind_old, ind_new,
+ (old_table_oid == RelationRelationId),
+ swap_toast_by_content,
+ true,
+ InvalidTransactionId,
+ InvalidMultiXactId,
+ mapped_tables);
+
+#ifdef USE_ASSERT_CHECKING
+ /*
+ * Concurrent processing is not supported for system relations, so
+ * there should be no mapped tables.
+ */
+ for (int i = 0; i < 4; i++)
+ Assert(mapped_tables[i] == 0);
+#endif
+ }
+
+ /* The new indexes must be visible for deletion. */
+ CommandCounterIncrement();
+
+ /* Close the old heap but keep lock until transaction commit. */
+ table_close(OldHeap, NoLock);
+ /* Close the new heap. (We didn't have to open its indexes). */
+ table_close(NewHeap, NoLock);
+
+ /* Cleanup what we don't need anymore. (And close the identity index.) */
+ pfree(ident_key);
+ free_index_insert_state(iistate);
+
+ /*
+ * Swap the relations and their TOAST relations and TOAST indexes. This
+ * also drops the new relation and its indexes.
+ *
+ * (System catalogs are currently not supported.)
+ */
+ Assert(!is_system_catalog);
+ finish_heap_swap(old_table_oid, new_table_oid,
+ is_system_catalog,
+ swap_toast_by_content,
+ false, true, false,
+ frozenXid, cutoffMulti,
+ relpersistence);
+
+ pfree(rri);
+}
+
+/*
+ * Build indexes on NewHeap according to those on OldHeap.
+ *
+ * OldIndexes is the list of index OIDs on OldHeap.
+ *
+ * A list of OIDs of the corresponding indexes created on NewHeap is
+ * returned. The order of items does match, so we can use these arrays to swap
+ * index storage.
+ */
+static List *
+build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes)
+{
+ StringInfo ind_name;
+ ListCell *lc;
+ List *result = NIL;
+
+ pgstat_progress_update_param(PROGRESS_CLUSTER_PHASE,
+ PROGRESS_CLUSTER_PHASE_REBUILD_INDEX);
+
+ ind_name = makeStringInfo();
+
+ foreach(lc, OldIndexes)
+ {
+ Oid ind_oid,
+ ind_oid_new,
+ tbsp_oid;
+ Relation ind;
+ IndexInfo *ind_info;
+ int i,
+ heap_col_id;
+ List *colnames;
+ int16 indnatts;
+ Oid *collations,
+ *opclasses;
+ HeapTuple tup;
+ bool isnull;
+ Datum d;
+ oidvector *oidvec;
+ int2vector *int2vec;
+ size_t oid_arr_size;
+ size_t int2_arr_size;
+ int16 *indoptions;
+ text *reloptions = NULL;
+ bits16 flags;
+ Datum *opclassOptions;
+ NullableDatum *stattargets;
+
+ ind_oid = lfirst_oid(lc);
+ ind = index_open(ind_oid, AccessShareLock);
+ ind_info = BuildIndexInfo(ind);
+
+ tbsp_oid = ind->rd_rel->reltablespace;
+ /*
+ * Index name really doesn't matter, we'll eventually use only their
+ * storage. Just make them unique within the table.
+ */
+ resetStringInfo(ind_name);
+ appendStringInfo(ind_name, "ind_%d",
+ list_cell_number(OldIndexes, lc));
+
+ flags = 0;
+ if (ind->rd_index->indisprimary)
+ flags |= INDEX_CREATE_IS_PRIMARY;
+
+ colnames = NIL;
+ indnatts = ind->rd_index->indnatts;
+ oid_arr_size = sizeof(Oid) * indnatts;
+ int2_arr_size = sizeof(int16) * indnatts;
+
+ collations = (Oid *) palloc(oid_arr_size);
+ for (i = 0; i < indnatts; i++)
+ {
+ char *colname;
+
+ heap_col_id = ind->rd_index->indkey.values[i];
+ if (heap_col_id > 0)
+ {
+ Form_pg_attribute att;
+
+ /* Normal attribute. */
+ att = TupleDescAttr(OldHeap->rd_att, heap_col_id - 1);
+ colname = pstrdup(NameStr(att->attname));
+ collations[i] = att->attcollation;
+ }
+ else if (heap_col_id == 0)
+ {
+ HeapTuple tuple;
+ Form_pg_attribute att;
+
+ /*
+ * Expression column is not present in relcache. What we need
+ * here is an attribute of the *index* relation.
+ */
+ tuple = SearchSysCache2(ATTNUM,
+ ObjectIdGetDatum(ind_oid),
+ Int16GetDatum(i + 1));
+ if (!HeapTupleIsValid(tuple))
+ elog(ERROR,
+ "cache lookup failed for attribute %d of relation %u",
+ i + 1, ind_oid);
+ att = (Form_pg_attribute) GETSTRUCT(tuple);
+ colname = pstrdup(NameStr(att->attname));
+ collations[i] = att->attcollation;
+ ReleaseSysCache(tuple);
+ }
+ else
+ elog(ERROR, "Unexpected column number: %d",
+ heap_col_id);
+
+ colnames = lappend(colnames, colname);
+ }
+
+ /*
+ * Special effort needed for variable length attributes of
+ * Form_pg_index.
+ */
+ tup = SearchSysCache1(INDEXRELID, ObjectIdGetDatum(ind_oid));
+ if (!HeapTupleIsValid(tup))
+ elog(ERROR, "cache lookup failed for index %u", ind_oid);
+ d = SysCacheGetAttr(INDEXRELID, tup, Anum_pg_index_indclass, &isnull);
+ Assert(!isnull);
+ oidvec = (oidvector *) DatumGetPointer(d);
+ opclasses = (Oid *) palloc(oid_arr_size);
+ memcpy(opclasses, oidvec->values, oid_arr_size);
+
+ d = SysCacheGetAttr(INDEXRELID, tup, Anum_pg_index_indoption,
+ &isnull);
+ Assert(!isnull);
+ int2vec = (int2vector *) DatumGetPointer(d);
+ indoptions = (int16 *) palloc(int2_arr_size);
+ memcpy(indoptions, int2vec->values, int2_arr_size);
+ ReleaseSysCache(tup);
+
+ tup = SearchSysCache1(RELOID, ObjectIdGetDatum(ind_oid));
+ if (!HeapTupleIsValid(tup))
+ elog(ERROR, "cache lookup failed for index relation %u", ind_oid);
+ d = SysCacheGetAttr(RELOID, tup, Anum_pg_class_reloptions, &isnull);
+ reloptions = !isnull ? DatumGetTextPCopy(d) : NULL;
+ ReleaseSysCache(tup);
+
+ opclassOptions = palloc0(sizeof(Datum) * ind_info->ii_NumIndexAttrs);
+ for (i = 0; i < ind_info->ii_NumIndexAttrs; i++)
+ opclassOptions[i] = get_attoptions(ind_oid, i + 1);
+
+ stattargets = get_index_stattargets(ind_oid, ind_info);
+
+ /*
+ * Neither parentIndexRelid nor parentConstraintId needs to be passed
+ * since the new catalog entries (pg_constraint, pg_inherits) would
+ * eventually be dropped. Therefore there's no need to record valid
+ * dependency on parents.
+ */
+ ind_oid_new = index_create(NewHeap,
+ ind_name->data,
+ InvalidOid,
+ InvalidOid, /* parentIndexRelid */
+ InvalidOid, /* parentConstraintId */
+ InvalidOid,
+ ind_info,
+ colnames,
+ ind->rd_rel->relam,
+ tbsp_oid,
+ collations,
+ opclasses,
+ opclassOptions,
+ indoptions,
+ stattargets,
+ PointerGetDatum(reloptions),
+ flags, /* flags */
+ 0, /* constr_flags */
+ false, /* allow_system_table_mods */
+ false, /* is_internal */
+ NULL /* constraintId */
+ );
+ result = lappend_oid(result, ind_oid_new);
+
+ index_close(ind, AccessShareLock);
+ list_free_deep(colnames);
+ pfree(collations);
+ pfree(opclasses);
+ pfree(indoptions);
+ if (reloptions)
+ pfree(reloptions);
+ }
+
+ return result;
+}
+
+static void
+init_rel_reopen_info(RelReopenInfo *rri, Relation *rel_p, Oid relid,
+ LOCKMODE lockmode_orig, LOCKMODE lockmode_new)
+{
+ rri->rel_p = rel_p;
+ rri->relid = relid;
+ rri->lockmode_orig = lockmode_orig;
+ rri->lockmode_new = lockmode_new;
+}
+
+/*
+ * Unlock and close relations specified by items of the 'rels' array. 'nrels'
+ * is the number of items.
+ *
+ * Information needed to (re)open the relations (or to issue meaningful ERROR)
+ * is added to the array items.
+ */
+static void
+unlock_and_close_relations(RelReopenInfo *rels, int nrel)
+{
+ int i;
+ RelReopenInfo *rri;
+
+ /*
+ * First, retrieve the information that we will need for re-opening.
+ *
+ * We could close (and unlock) each relation as soon as we have gathered
+ * the related information, but then we would have to be careful not to
+ * unlock the table until we have the info on all its indexes. (Once we
+ * unlock the table, any index can be dropped, and thus we can fail to get
+ * the name we want to report if re-opening fails.) It seem simpler to
+ * separate the work into two iterations.
+ */
+ for (i = 0; i < nrel; i++)
+ {
+ Relation rel;
+
+ rri = &rels[i];
+ rel = *rri->rel_p;
+
+ if (rel)
+ {
+ Assert(CheckRelationLockedByMe(rel, rri->lockmode_orig, false));
+ Assert(!OidIsValid(rri->relid));
+
+ rri->relid = RelationGetRelid(rel);
+ rri->relkind = rel->rd_rel->relkind;
+ rri->relname = pstrdup(RelationGetRelationName(rel));
+ }
+ else
+ {
+ Assert(OidIsValid(rri->relid));
+
+ rri->relname = get_rel_name(rri->relid);
+ rri->relkind = get_rel_relkind(rri->relid);
+ }
+ }
+
+ /* Second, close the relations. */
+ for (i = 0; i < nrel; i++)
+ {
+ Relation rel;
+
+ rri = &rels[i];
+ rel = *rri->rel_p;
+
+ /* Close the relation if the caller passed one. */
+ if (rel)
+ {
+ if (rri->relkind == RELKIND_RELATION)
+ table_close(rel, rri->lockmode_orig);
+ else
+ {
+ Assert(rri->relkind == RELKIND_INDEX);
+
+ index_close(rel, rri->lockmode_orig);
+ }
+ }
+ }
+}
+
+/*
+ * Re-open the relations closed previously by unlock_and_close_relations().
+ */
+static void
+reopen_relations(RelReopenInfo *rels, int nrel)
+{
+ for (int i = 0; i < nrel; i++)
+ {
+ RelReopenInfo *rri = &rels[i];
+ Relation rel;
+
+ if (rri->relkind == RELKIND_RELATION)
+ {
+ rel = try_table_open(rri->relid, rri->lockmode_new);
+ }
+ else
+ {
+ Assert(rri->relkind == RELKIND_INDEX);
+
+ rel = try_index_open(rri->relid, rri->lockmode_new);
+ }
+
+ if (rel == NULL)
+ {
+ const char *kind_str;
+
+ kind_str = (rri->relkind == RELKIND_RELATION) ? "table" : "index";
+ ereport(ERROR,
+ (errmsg("could not open \%s \"%s\"", kind_str,
+ rri->relname),
+ errhint("The %s could have been dropped by another transaction.",
+ kind_str)));
+ }
+ *rri->rel_p = rel;
+
+ pfree(rri->relname);
+ }
+}
diff --git a/src/backend/commands/matview.c b/src/backend/commands/matview.c
index c12817091e..8f0197378d 100644
--- a/src/backend/commands/matview.c
+++ b/src/backend/commands/matview.c
@@ -905,7 +905,7 @@ refresh_by_match_merge(Oid matviewOid, Oid tempOid, Oid relowner,
static void
refresh_by_heap_swap(Oid matviewOid, Oid OIDNewHeap, char relpersistence)
{
- finish_heap_swap(matviewOid, OIDNewHeap, false, false, true, true,
+ finish_heap_swap(matviewOid, OIDNewHeap, false, false, true, true, true,
RecentXmin, ReadNextMultiXactId(), relpersistence);
}
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 4fc54bd6eb..f070ffc960 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -4500,6 +4500,16 @@ AlterTableInternal(Oid relid, List *cmds, bool recurse)
rel = relation_open(relid, lockmode);
+ /*
+ * If lockmode allows, check if VACUUM FULL / CLUSTER CONCURRENTLY is in
+ * progress. If lockmode is too weak, cluster_rel() should detect
+ * incompatible DDLs executed by us.
+ *
+ * XXX We might skip the changes for DDLs which do not change the tuple
+ * descriptor.
+ */
+ check_for_concurrent_cluster(relid, lockmode);
+
EventTriggerAlterTableRelid(relid);
ATController(NULL, rel, cmds, recurse, lockmode, NULL);
@@ -5929,6 +5939,7 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode,
finish_heap_swap(tab->relid, OIDNewHeap,
false, false, true,
!OidIsValid(tab->newTableSpace),
+ true,
RecentXmin,
ReadNextMultiXactId(),
persistence);
diff --git a/src/backend/commands/vacuum.c b/src/backend/commands/vacuum.c
index e6745e6145..516a064fed 100644
--- a/src/backend/commands/vacuum.c
+++ b/src/backend/commands/vacuum.c
@@ -111,7 +111,7 @@ static void vac_truncate_clog(TransactionId frozenXID,
TransactionId lastSaneFrozenXid,
MultiXactId lastSaneMinMulti);
static bool vacuum_rel(Oid relid, RangeVar *relation, VacuumParams *params,
- BufferAccessStrategy bstrategy);
+ BufferAccessStrategy bstrategy, bool isTopLevel);
static double compute_parallel_delay(void);
static VacOptValue get_vacoptval_from_boolean(DefElem *def);
static bool vac_tid_reaped(ItemPointer itemptr, void *state);
@@ -153,6 +153,7 @@ ExecVacuum(ParseState *pstate, VacuumStmt *vacstmt, bool isTopLevel)
bool analyze = false;
bool freeze = false;
bool full = false;
+ bool concurrent = false;
bool disable_page_skipping = false;
bool process_main = true;
bool process_toast = true;
@@ -226,6 +227,8 @@ ExecVacuum(ParseState *pstate, VacuumStmt *vacstmt, bool isTopLevel)
freeze = defGetBoolean(opt);
else if (strcmp(opt->defname, "full") == 0)
full = defGetBoolean(opt);
+ else if (strcmp(opt->defname, "concurrently") == 0)
+ concurrent = defGetBoolean(opt);
else if (strcmp(opt->defname, "disable_page_skipping") == 0)
disable_page_skipping = defGetBoolean(opt);
else if (strcmp(opt->defname, "index_cleanup") == 0)
@@ -300,7 +303,7 @@ ExecVacuum(ParseState *pstate, VacuumStmt *vacstmt, bool isTopLevel)
(skip_locked ? VACOPT_SKIP_LOCKED : 0) |
(analyze ? VACOPT_ANALYZE : 0) |
(freeze ? VACOPT_FREEZE : 0) |
- (full ? VACOPT_FULL : 0) |
+ (full ? (concurrent ? VACOPT_FULL_CONCURRENT : VACOPT_FULL_EXCLUSIVE) : 0) |
(disable_page_skipping ? VACOPT_DISABLE_PAGE_SKIPPING : 0) |
(process_main ? VACOPT_PROCESS_MAIN : 0) |
(process_toast ? VACOPT_PROCESS_TOAST : 0) |
@@ -380,6 +383,12 @@ ExecVacuum(ParseState *pstate, VacuumStmt *vacstmt, bool isTopLevel)
errmsg("ONLY_DATABASE_STATS cannot be specified with other VACUUM options")));
}
+ /* This problem cannot be identified from the options. */
+ if (concurrent && !full)
+ ereport(ERROR,
+ (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
+ errmsg("CONCURRENTLY can only be specified with VACUUM FULL")));
+
/*
* All freeze ages are zero if the FREEZE option is given; otherwise pass
* them as -1 which means to use the default values.
@@ -543,7 +552,17 @@ vacuum(List *relations, VacuumParams *params, BufferAccessStrategy bstrategy,
relations = newrels;
}
else
+ {
+ /*
+ * Concurrent processing is currently considered rather special so it
+ * is not performed in bulk.
+ */
+ if (params->options & VACOPT_FULL_CONCURRENT)
+ ereport(ERROR,
+ (errmsg("VACUUM (CONCURRENTLY) requires explicit list of tables")));
+
relations = get_all_vacuum_rels(vac_context, params->options);
+ }
/*
* Decide whether we need to start/commit our own transactions.
@@ -616,7 +635,8 @@ vacuum(List *relations, VacuumParams *params, BufferAccessStrategy bstrategy,
if (params->options & VACOPT_VACUUM)
{
- if (!vacuum_rel(vrel->oid, vrel->relation, params, bstrategy))
+ if (!vacuum_rel(vrel->oid, vrel->relation, params, bstrategy,
+ isTopLevel))
continue;
}
@@ -960,6 +980,17 @@ expand_vacuum_rel(VacuumRelation *vrel, MemoryContext vac_context,
(errmsg("VACUUM ONLY of partitioned table \"%s\" has no effect",
vrel->relation->relname)));
+ /*
+ * Concurrent processing is currently considered rather special
+ * (e.g. in terms of resources consumed) so it is not performed in
+ * bulk.
+ */
+ if (is_partitioned_table && (options & VACOPT_FULL_CONCURRENT))
+ ereport(ERROR,
+ (errmsg("VACUUM (CONCURRENTLY) not supported for partitioned tables"),
+ errhint("Consider running the command for individual partitions.")));
+
+
ReleaseSysCache(tuple);
/*
@@ -1954,10 +1985,10 @@ vac_truncate_clog(TransactionId frozenXID,
/*
* vacuum_rel() -- vacuum one heap relation
*
- * relid identifies the relation to vacuum. If relation is supplied,
- * use the name therein for reporting any failure to open/lock the rel;
- * do not use it once we've successfully opened the rel, since it might
- * be stale.
+ * relid identifies the relation to vacuum. If relation is supplied, use
+ * the name therein for reporting any failure to open/lock the rel; do
+ * not use it once we've successfully opened the rel, since it might be
+ * stale.
*
* Returns true if it's okay to proceed with a requested ANALYZE
* operation on this table.
@@ -1972,7 +2003,7 @@ vac_truncate_clog(TransactionId frozenXID,
*/
static bool
vacuum_rel(Oid relid, RangeVar *relation, VacuumParams *params,
- BufferAccessStrategy bstrategy)
+ BufferAccessStrategy bstrategy, bool isTopLevel)
{
LOCKMODE lmode;
Relation rel;
@@ -2035,10 +2066,11 @@ vacuum_rel(Oid relid, RangeVar *relation, VacuumParams *params,
/*
* Determine the type of lock we want --- hard exclusive lock for a FULL
- * vacuum, but just ShareUpdateExclusiveLock for concurrent vacuum. Either
- * way, we can be sure that no other backend is vacuuming the same table.
+ * exclusive vacuum, but a weaker lock (ShareUpdateExclusiveLock) for
+ * concurrent vacuum. Either way, we can be sure that no other backend is
+ * vacuuming the same table.
*/
- lmode = (params->options & VACOPT_FULL) ?
+ lmode = (params->options & VACOPT_FULL_EXCLUSIVE) ?
AccessExclusiveLock : ShareUpdateExclusiveLock;
/* open the relation and get the appropriate lock on it */
@@ -2053,6 +2085,22 @@ vacuum_rel(Oid relid, RangeVar *relation, VacuumParams *params,
return false;
}
+ /*
+ * Skip the relation if VACUUM FULL / CLUSTER CONCURRENTLY is in progress
+ * as it will drop the current storage of the relation.
+ *
+ * This check should not take place until we have a lock that prevents
+ * another backend from starting VACUUM FULL / CLUSTER CONCURRENTLY later.
+ */
+ Assert(lmode >= ShareUpdateExclusiveLock);
+ if (is_concurrent_cluster_in_progress(relid))
+ {
+ relation_close(rel, lmode);
+ PopActiveSnapshot();
+ CommitTransactionCommand();
+ return false;
+ }
+
/*
* When recursing to a TOAST table, check privileges on the parent. NB:
* This is only safe to do because we hold a session lock on the main
@@ -2126,19 +2174,6 @@ vacuum_rel(Oid relid, RangeVar *relation, VacuumParams *params,
return true;
}
- /*
- * Get a session-level lock too. This will protect our access to the
- * relation across multiple transactions, so that we can vacuum the
- * relation's TOAST table (if any) secure in the knowledge that no one is
- * deleting the parent relation.
- *
- * NOTE: this cannot block, even if someone else is waiting for access,
- * because the lock manager knows that both lock requests are from the
- * same process.
- */
- lockrelid = rel->rd_lockInfo.lockRelId;
- LockRelationIdForSession(&lockrelid, lmode);
-
/*
* Set index_cleanup option based on index_cleanup reloption if it wasn't
* specified in VACUUM command, or when running in an autovacuum worker
@@ -2191,6 +2226,30 @@ vacuum_rel(Oid relid, RangeVar *relation, VacuumParams *params,
else
toast_relid = InvalidOid;
+ /*
+ * Get a session-level lock too. This will protect our access to the
+ * relation across multiple transactions, so that we can vacuum the
+ * relation's TOAST table (if any) secure in the knowledge that no one is
+ * deleting the parent relation.
+ *
+ * NOTE: this cannot block, even if someone else is waiting for access,
+ * because the lock manager knows that both lock requests are from the
+ * same process.
+ */
+ if (OidIsValid(toast_relid))
+ {
+ /*
+ * You might worry that, in the VACUUM (FULL, CONCURRENTLY) case,
+ * cluster_rel() needs to release all the locks on the relation at
+ * some point, but this session lock makes it impossible. In fact,
+ * cluster_rel() will will eventually be called for the TOAST relation
+ * and raise ERROR because, in the concurrent mode, it cannot process
+ * TOAST relation alone anyway.
+ */
+ lockrelid = rel->rd_lockInfo.lockRelId;
+ LockRelationIdForSession(&lockrelid, lmode);
+ }
+
/*
* Switch to the table owner's userid, so that any index functions are run
* as that user. Also lock down security-restricted operations and
@@ -2218,11 +2277,22 @@ vacuum_rel(Oid relid, RangeVar *relation, VacuumParams *params,
{
ClusterParams cluster_params = {0};
+ /*
+ * Invalid toast_relid means that there is no session lock on the
+ * relation. Such a lock would be a problem because it would
+ * prevent cluster_rel() from releasing all locks when it tries to
+ * get AccessExclusiveLock.
+ */
+ Assert(!OidIsValid(toast_relid));
+
if ((params->options & VACOPT_VERBOSE) != 0)
cluster_params.options |= CLUOPT_VERBOSE;
+ if ((params->options & VACOPT_FULL_CONCURRENT) != 0)
+ cluster_params.options |= CLUOPT_CONCURRENT;
+
/* VACUUM FULL is now a variant of CLUSTER; see cluster.c */
- cluster_rel(rel, InvalidOid, &cluster_params);
+ cluster_rel(rel, InvalidOid, &cluster_params, isTopLevel, true);
/* cluster_rel closes the relation, but keeps lock */
rel = NULL;
@@ -2268,13 +2338,15 @@ vacuum_rel(Oid relid, RangeVar *relation, VacuumParams *params,
toast_vacuum_params.options |= VACOPT_PROCESS_MAIN;
toast_vacuum_params.toast_parent = relid;
- vacuum_rel(toast_relid, NULL, &toast_vacuum_params, bstrategy);
+ vacuum_rel(toast_relid, NULL, &toast_vacuum_params, bstrategy,
+ isTopLevel);
}
/*
* Now release the session-level lock on the main table.
*/
- UnlockRelationIdForSession(&lockrelid, lmode);
+ if (OidIsValid(toast_relid))
+ UnlockRelationIdForSession(&lockrelid, lmode);
/* Report that we really did it. */
return true;
diff --git a/src/backend/meson.build b/src/backend/meson.build
index 2b0db21480..4b00a9b8c6 100644
--- a/src/backend/meson.build
+++ b/src/backend/meson.build
@@ -194,5 +194,6 @@ pg_test_mod_args = pg_mod_args + {
subdir('jit/llvm')
subdir('replication/libpqwalreceiver')
subdir('replication/pgoutput')
+subdir('replication/pgoutput_cluster')
subdir('snowball')
subdir('utils/mb/conversion_procs')
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index 0bff0f1065..8f45a7a168 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -33,6 +33,7 @@
#include "access/xlogreader.h"
#include "access/xlogrecord.h"
#include "catalog/pg_control.h"
+#include "commands/cluster.h"
#include "replication/decode.h"
#include "replication/logical.h"
#include "replication/message.h"
@@ -467,6 +468,29 @@ heap_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
TransactionId xid = XLogRecGetXid(buf->record);
SnapBuild *builder = ctx->snapshot_builder;
+ /*
+ * Check if CLUSTER CONCURRENTLY is being performed by this backend. If
+ * so, only decode data changes of the table that it is processing, and
+ * the changes of its TOAST relation.
+ *
+ * (TOAST locator should not be set unless the main is.)
+ */
+ Assert(!OidIsValid(clustered_rel_toast_locator.relNumber) ||
+ OidIsValid(clustered_rel_locator.relNumber));
+
+ if (OidIsValid(clustered_rel_locator.relNumber))
+ {
+ XLogReaderState *r = buf->record;
+ RelFileLocator locator;
+
+ /* Not all records contain the block. */
+ if (XLogRecGetBlockTagExtended(r, 0, &locator, NULL, NULL, NULL) &&
+ !RelFileLocatorEquals(locator, clustered_rel_locator) &&
+ (!OidIsValid(clustered_rel_toast_locator.relNumber) ||
+ !RelFileLocatorEquals(locator, clustered_rel_toast_locator)))
+ return;
+ }
+
ReorderBufferProcessXid(ctx->reorder, xid, buf->origptr);
/*
diff --git a/src/backend/replication/logical/snapbuild.c b/src/backend/replication/logical/snapbuild.c
index 84bf0503a5..86a9b0335a 100644
--- a/src/backend/replication/logical/snapbuild.c
+++ b/src/backend/replication/logical/snapbuild.c
@@ -486,6 +486,26 @@ SnapBuildInitialSnapshot(SnapBuild *builder)
return SnapBuildMVCCFromHistoric(snap, true);
}
+/*
+ * Build an MVCC snapshot for the initial data load performed by CLUSTER
+ * CONCURRENTLY command.
+ *
+ * The snapshot will only be used to scan one particular relation, which is
+ * treated like a catalog (therefore ->building_full_snapshot is not
+ * important), and the caller should already have a replication slot setup (so
+ * we do not set MyProc->xmin). XXX Do we yet need to add some restrictions?
+ */
+Snapshot
+SnapBuildInitialSnapshotForCluster(SnapBuild *builder)
+{
+ Snapshot snap;
+
+ Assert(builder->state == SNAPBUILD_CONSISTENT);
+
+ snap = SnapBuildBuildSnapshot(builder);
+ return SnapBuildMVCCFromHistoric(snap, false);
+}
+
/*
* Turn a historic MVCC snapshot into an ordinary MVCC snapshot.
*
diff --git a/src/backend/replication/pgoutput_cluster/Makefile b/src/backend/replication/pgoutput_cluster/Makefile
new file mode 100644
index 0000000000..31471bb546
--- /dev/null
+++ b/src/backend/replication/pgoutput_cluster/Makefile
@@ -0,0 +1,32 @@
+#-------------------------------------------------------------------------
+#
+# Makefile--
+# Makefile for src/backend/replication/pgoutput_cluster
+#
+# IDENTIFICATION
+# src/backend/replication/pgoutput_cluster
+#
+#-------------------------------------------------------------------------
+
+subdir = src/backend/replication/pgoutput_cluster
+top_builddir = ../../../..
+include $(top_builddir)/src/Makefile.global
+
+OBJS = \
+ $(WIN32RES) \
+ pgoutput_cluster.o
+PGFILEDESC = "pgoutput_cluster - logical replication output plugin for CLUSTER command"
+NAME = pgoutput_cluster
+
+all: all-shared-lib
+
+include $(top_srcdir)/src/Makefile.shlib
+
+install: all installdirs install-lib
+
+installdirs: installdirs-lib
+
+uninstall: uninstall-lib
+
+clean distclean: clean-lib
+ rm -f $(OBJS)
diff --git a/src/backend/replication/pgoutput_cluster/meson.build b/src/backend/replication/pgoutput_cluster/meson.build
new file mode 100644
index 0000000000..0f033064f2
--- /dev/null
+++ b/src/backend/replication/pgoutput_cluster/meson.build
@@ -0,0 +1,18 @@
+# Copyright (c) 2022-2024, PostgreSQL Global Development Group
+
+pgoutput_cluster_sources = files(
+ 'pgoutput_cluster.c',
+)
+
+if host_system == 'windows'
+ pgoutput_cluster_sources += rc_lib_gen.process(win32ver_rc, extra_args: [
+ '--NAME', 'pgoutput_cluster',
+ '--FILEDESC', 'pgoutput_cluster - logical replication output plugin for CLUSTER command',])
+endif
+
+pgoutput_cluster = shared_module('pgoutput_cluster',
+ pgoutput_cluster_sources,
+ kwargs: pg_mod_args,
+)
+
+backend_targets += pgoutput_cluster
diff --git a/src/backend/replication/pgoutput_cluster/pgoutput_cluster.c b/src/backend/replication/pgoutput_cluster/pgoutput_cluster.c
new file mode 100644
index 0000000000..43f7b34297
--- /dev/null
+++ b/src/backend/replication/pgoutput_cluster/pgoutput_cluster.c
@@ -0,0 +1,288 @@
+/* TODO Move into src/backend/cluster/ (and rename?) */
+/*-------------------------------------------------------------------------
+ *
+ * pgoutput_cluster.c
+ * Logical Replication output plugin for CLUSTER command
+ *
+ * Copyright (c) 2012-2024, PostgreSQL Global Development Group
+ *
+ * IDENTIFICATION
+ * src/backend/replication/pgoutput_cluster/pgoutput_cluster.c
+ *
+ *-------------------------------------------------------------------------
+ */
+#include "postgres.h"
+
+#include "access/heaptoast.h"
+#include "commands/cluster.h"
+#include "replication/snapbuild.h"
+
+PG_MODULE_MAGIC;
+
+static void plugin_startup(LogicalDecodingContext *ctx,
+ OutputPluginOptions *opt, bool is_init);
+static void plugin_shutdown(LogicalDecodingContext *ctx);
+static void plugin_begin_txn(LogicalDecodingContext *ctx,
+ ReorderBufferTXN *txn);
+static void plugin_commit_txn(LogicalDecodingContext *ctx,
+ ReorderBufferTXN *txn, XLogRecPtr commit_lsn);
+static void plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
+ Relation rel, ReorderBufferChange *change);
+static void plugin_truncate(struct LogicalDecodingContext *ctx,
+ ReorderBufferTXN *txn, int nrelations,
+ Relation relations[],
+ ReorderBufferChange *change);
+static void store_change(LogicalDecodingContext *ctx,
+ ConcurrentChangeKind kind, HeapTuple tuple);
+
+void
+_PG_output_plugin_init(OutputPluginCallbacks *cb)
+{
+ AssertVariableIsOfType(&_PG_output_plugin_init, LogicalOutputPluginInit);
+
+ cb->startup_cb = plugin_startup;
+ cb->begin_cb = plugin_begin_txn;
+ cb->change_cb = plugin_change;
+ cb->truncate_cb = plugin_truncate;
+ cb->commit_cb = plugin_commit_txn;
+ cb->shutdown_cb = plugin_shutdown;
+}
+
+
+/* initialize this plugin */
+static void
+plugin_startup(LogicalDecodingContext *ctx, OutputPluginOptions *opt,
+ bool is_init)
+{
+ ctx->output_plugin_private = NULL;
+
+ /* Probably unnecessary, as we don't use the SQL interface ... */
+ opt->output_type = OUTPUT_PLUGIN_BINARY_OUTPUT;
+
+ if (ctx->output_plugin_options != NIL)
+ {
+ ereport(ERROR,
+ (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
+ errmsg("This plugin does not expect any options")));
+ }
+}
+
+static void
+plugin_shutdown(LogicalDecodingContext *ctx)
+{
+}
+
+/*
+ * As we don't release the slot during processing of particular table, there's
+ * no room for SQL interface, even for debugging purposes. Therefore we need
+ * neither OutputPluginPrepareWrite() nor OutputPluginWrite() in the plugin
+ * callbacks. (Although we might want to write custom callbacks, this API
+ * seems to be unnecessarily generic for our purposes.)
+ */
+
+/* BEGIN callback */
+static void
+plugin_begin_txn(LogicalDecodingContext *ctx, ReorderBufferTXN *txn)
+{
+}
+
+/* COMMIT callback */
+static void
+plugin_commit_txn(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
+ XLogRecPtr commit_lsn)
+{
+}
+
+/*
+ * Callback for individual changed tuples
+ */
+static void
+plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
+ Relation relation, ReorderBufferChange *change)
+{
+ ClusterDecodingState *dstate;
+
+ dstate = (ClusterDecodingState *) ctx->output_writer_private;
+
+ /* Only interested in one particular relation. */
+ if (relation->rd_id != dstate->relid)
+ return;
+
+ /* Decode entry depending on its type */
+ switch (change->action)
+ {
+ case REORDER_BUFFER_CHANGE_INSERT:
+ {
+ HeapTuple newtuple;
+
+ newtuple = change->data.tp.newtuple != NULL ?
+ change->data.tp.newtuple : NULL;
+
+ /*
+ * Identity checks in the main function should have made this
+ * impossible.
+ */
+ if (newtuple == NULL)
+ elog(ERROR, "Incomplete insert info.");
+
+ store_change(ctx, CHANGE_INSERT, newtuple);
+ }
+ break;
+ case REORDER_BUFFER_CHANGE_UPDATE:
+ {
+ HeapTuple oldtuple,
+ newtuple;
+
+ oldtuple = change->data.tp.oldtuple != NULL ?
+ change->data.tp.oldtuple : NULL;
+ newtuple = change->data.tp.newtuple != NULL ?
+ change->data.tp.newtuple : NULL;
+
+ if (newtuple == NULL)
+ elog(ERROR, "Incomplete update info.");
+
+ if (oldtuple != NULL)
+ store_change(ctx, CHANGE_UPDATE_OLD, oldtuple);
+
+ store_change(ctx, CHANGE_UPDATE_NEW, newtuple);
+ }
+ break;
+ case REORDER_BUFFER_CHANGE_DELETE:
+ {
+ HeapTuple oldtuple;
+
+ oldtuple = change->data.tp.oldtuple ?
+ change->data.tp.oldtuple : NULL;
+
+ if (oldtuple == NULL)
+ elog(ERROR, "Incomplete delete info.");
+
+ store_change(ctx, CHANGE_DELETE, oldtuple);
+ }
+ break;
+ default:
+ /* Should not come here */
+ Assert(false);
+ break;
+ }
+}
+
+static void
+plugin_truncate(struct LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
+ int nrelations, Relation relations[],
+ ReorderBufferChange *change)
+{
+ ClusterDecodingState *dstate;
+ int i;
+ Relation relation = NULL;
+
+ dstate = (ClusterDecodingState *) ctx->output_writer_private;
+
+ /* Find the relation we are processing. */
+ for (i = 0; i < nrelations; i++)
+ {
+ relation = relations[i];
+
+ if (RelationGetRelid(relation) == dstate->relid)
+ break;
+ }
+
+ /* Is this truncation of another relation? */
+ if (i == nrelations)
+ return;
+
+ store_change(ctx, CHANGE_TRUNCATE, NULL);
+}
+
+/* Store concurrent data change. */
+static void
+store_change(LogicalDecodingContext *ctx, ConcurrentChangeKind kind,
+ HeapTuple tuple)
+{
+ ClusterDecodingState *dstate;
+ char *change_raw;
+ ConcurrentChange change;
+ bool flattened = false;
+ Size size;
+ Datum values[1];
+ bool isnull[1];
+ char *dst, *dst_start;
+
+ dstate = (ClusterDecodingState *) ctx->output_writer_private;
+
+ size = MAXALIGN(VARHDRSZ) + SizeOfConcurrentChange;
+
+ if (tuple)
+ {
+ /*
+ * ReorderBufferCommit() stores the TOAST chunks in its private memory
+ * context and frees them after having called
+ * apply_change(). Therefore we need flat copy (including TOAST) that
+ * we eventually copy into the memory context which is available to
+ * decode_concurrent_changes().
+ */
+ if (HeapTupleHasExternal(tuple))
+ {
+ /*
+ * toast_flatten_tuple_to_datum() might be more convenient but we
+ * don't want the decompression it does.
+ */
+ tuple = toast_flatten_tuple(tuple, dstate->tupdesc);
+ flattened = true;
+ }
+
+ size += tuple->t_len;
+ }
+
+ /* XXX Isn't there any function / macro to do this? */
+ if (size >= 0x3FFFFFFF)
+ elog(ERROR, "Change is too big.");
+
+ /* Construct the change. */
+ change_raw = (char *) palloc0(size);
+ SET_VARSIZE(change_raw, size);
+ /*
+ * Since the varlena alignment might not be sufficient for the structure,
+ * set the fields in a local instance and remember where it should
+ * eventually be copied.
+ */
+ change.kind = kind;
+ dst_start = (char *) VARDATA(change_raw);
+
+ /* No other information is needed for TRUNCATE. */
+ if (change.kind == CHANGE_TRUNCATE)
+ {
+ memcpy(dst_start, &change, SizeOfConcurrentChange);
+ goto store;
+ }
+
+ /*
+ * Copy the tuple.
+ *
+ * CAUTION: change->tup_data.t_data must be fixed on retrieval!
+ */
+ memcpy(&change.tup_data, tuple, sizeof(HeapTupleData));
+ dst = dst_start + SizeOfConcurrentChange;
+ memcpy(dst, tuple->t_data, tuple->t_len);
+
+ /* The data has been copied. */
+ if (flattened)
+ pfree(tuple);
+
+store:
+ /* Copy the structure so it can be stored. */
+ memcpy(dst_start, &change, SizeOfConcurrentChange);
+
+ /* Store as tuple of 1 bytea column. */
+ values[0] = PointerGetDatum(change_raw);
+ isnull[0] = false;
+ tuplestore_putvalues(dstate->tstore, dstate->tupdesc_change,
+ values, isnull);
+
+ /* Accounting. */
+ dstate->nchanges++;
+
+ /* Cleanup. */
+ pfree(change_raw);
+}
+
diff --git a/src/backend/storage/ipc/ipci.c b/src/backend/storage/ipc/ipci.c
index 174eed7036..56df243a8a 100644
--- a/src/backend/storage/ipc/ipci.c
+++ b/src/backend/storage/ipc/ipci.c
@@ -25,6 +25,7 @@
#include "access/xlogprefetcher.h"
#include "access/xlogrecovery.h"
#include "commands/async.h"
+#include "commands/cluster.h"
#include "miscadmin.h"
#include "pgstat.h"
#include "postmaster/autovacuum.h"
@@ -148,6 +149,7 @@ CalculateShmemSize(int *num_semaphores)
size = add_size(size, WaitEventCustomShmemSize());
size = add_size(size, InjectionPointShmemSize());
size = add_size(size, SlotSyncShmemSize());
+ size = add_size(size, ClusterShmemSize());
/* include additional requested shmem from preload libraries */
size = add_size(size, total_addin_request);
@@ -340,6 +342,7 @@ CreateOrAttachShmemStructs(void)
StatsShmemInit();
WaitEventCustomShmemInit();
InjectionPointShmemInit();
+ ClusterShmemInit();
}
/*
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 25fe3d5801..7e5484dae4 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1302,6 +1302,17 @@ ProcessUtilitySlow(ParseState *pstate,
lockmode = AlterTableGetLockLevel(atstmt->cmds);
relid = AlterTableLookupRelation(atstmt, lockmode);
+ /*
+ * If lockmode allows, check if VACUUM FULL / CLUSTER
+ * CONCURRENT is in progress. If lockmode is too weak,
+ * cluster_rel() should detect incompatible DDLs executed
+ * by us.
+ *
+ * XXX We might skip the changes for DDLs which do not
+ * change the tuple descriptor.
+ */
+ check_for_concurrent_cluster(relid, lockmode);
+
if (OidIsValid(relid))
{
AlterTableUtilityContext atcontext;
diff --git a/src/backend/utils/activity/backend_progress.c b/src/backend/utils/activity/backend_progress.c
index eebc968193..e2c84baba9 100644
--- a/src/backend/utils/activity/backend_progress.c
+++ b/src/backend/utils/activity/backend_progress.c
@@ -162,3 +162,19 @@ pgstat_progress_end_command(void)
beentry->st_progress.command_target = InvalidOid;
PGSTAT_END_WRITE_ACTIVITY(beentry);
}
+
+void
+pgstat_progress_restore_state(PgBackendProgress *backup)
+{
+ volatile PgBackendStatus *beentry = MyBEEntry;
+
+ if (!beentry || !pgstat_track_activities)
+ return;
+
+ PGSTAT_BEGIN_WRITE_ACTIVITY(beentry);
+ beentry->st_progress.command = backup->command;
+ beentry->st_progress.command_target = backup->command_target;
+ memcpy(MyBEEntry->st_progress.param, backup->param,
+ sizeof(beentry->st_progress.param));
+ PGSTAT_END_WRITE_ACTIVITY(beentry);
+}
diff --git a/src/backend/utils/activity/wait_event_names.txt b/src/backend/utils/activity/wait_event_names.txt
index 0b53cba807..a28fba18a7 100644
--- a/src/backend/utils/activity/wait_event_names.txt
+++ b/src/backend/utils/activity/wait_event_names.txt
@@ -345,6 +345,7 @@ WALSummarizer "Waiting to read or update WAL summarization state."
DSMRegistry "Waiting to read or update the dynamic shared memory registry."
InjectionPoint "Waiting to read or update information related to injection points."
SerialControl "Waiting to read or update shared <filename>pg_serial</filename> state."
+ClusteredRels "Waiting to read or update information on tables being clustered concurrently."
#
# END OF PREDEFINED LWLOCKS (DO NOT CHANGE THIS LINE)
diff --git a/src/backend/utils/cache/inval.c b/src/backend/utils/cache/inval.c
index 254947e943..e2d030e6a1 100644
--- a/src/backend/utils/cache/inval.c
+++ b/src/backend/utils/cache/inval.c
@@ -1565,6 +1565,28 @@ CacheInvalidateRelcache(Relation relation)
databaseId, relationId);
}
+/*
+ * CacheInvalidateRelcacheImmediate
+ * Send invalidation message for the specified relation's relcache entry.
+ *
+ * Currently this is used in VACUUM FULL/CLUSTER CONCURRENTLY, to make sure
+ * that other backends are aware that the command is being executed for the
+ * relation.
+ */
+void
+CacheInvalidateRelcacheImmediate(Relation relation)
+{
+ SharedInvalidationMessage msg;
+
+ msg.rc.id = SHAREDINVALRELCACHE_ID;
+ msg.rc.dbId = MyDatabaseId;
+ msg.rc.relId = RelationGetRelid(relation);
+ /* check AddCatcacheInvalidationMessage() for an explanation */
+ VALGRIND_MAKE_MEM_DEFINED(&msg, sizeof(msg));
+
+ SendSharedInvalidMessages(&msg, 1);
+}
+
/*
* CacheInvalidateRelcacheAll
* Register invalidation of the whole relcache at the end of command.
diff --git a/src/backend/utils/cache/relcache.c b/src/backend/utils/cache/relcache.c
index 43219a9629..55557c7589 100644
--- a/src/backend/utils/cache/relcache.c
+++ b/src/backend/utils/cache/relcache.c
@@ -64,6 +64,7 @@
#include "catalog/pg_type.h"
#include "catalog/schemapg.h"
#include "catalog/storage.h"
+#include "commands/cluster.h"
#include "commands/policy.h"
#include "commands/publicationcmds.h"
#include "commands/trigger.h"
@@ -1246,6 +1247,10 @@ retry:
/* make sure relation is marked as having no open file yet */
relation->rd_smgr = NULL;
+ /* Is CLUSTER CONCURRENTLY in progress? */
+ relation->rd_cluster_concurrent =
+ is_concurrent_cluster_in_progress(targetRelId);
+
/*
* now we can free the memory allocated for pg_class_tuple
*/
diff --git a/src/backend/utils/time/snapmgr.c b/src/backend/utils/time/snapmgr.c
index 42bded373b..103d1249bb 100644
--- a/src/backend/utils/time/snapmgr.c
+++ b/src/backend/utils/time/snapmgr.c
@@ -154,7 +154,6 @@ static List *exportedSnapshots = NIL;
/* Prototypes for local functions */
static void UnregisterSnapshotNoOwner(Snapshot snapshot);
-static void FreeSnapshot(Snapshot snapshot);
static void SnapshotResetXmin(void);
/* ResourceOwner callbacks to track snapshot references */
@@ -587,7 +586,7 @@ CopySnapshot(Snapshot snapshot)
* FreeSnapshot
* Free the memory associated with a snapshot.
*/
-static void
+void
FreeSnapshot(Snapshot snapshot)
{
Assert(snapshot->regd_count == 0);
diff --git a/src/bin/psql/tab-complete.in.c b/src/bin/psql/tab-complete.in.c
index 81cbf10aa2..14c2058997 100644
--- a/src/bin/psql/tab-complete.in.c
+++ b/src/bin/psql/tab-complete.in.c
@@ -3119,7 +3119,7 @@ match_previous_words(int pattern_id,
* one word, so the above test is correct.
*/
if (ends_with(prev_wd, '(') || ends_with(prev_wd, ','))
- COMPLETE_WITH("VERBOSE");
+ COMPLETE_WITH("VERBOSE", "CONCURRENTLY");
}
/* COMMENT */
@@ -5133,7 +5133,8 @@ match_previous_words(int pattern_id,
"DISABLE_PAGE_SKIPPING", "SKIP_LOCKED",
"INDEX_CLEANUP", "PROCESS_MAIN", "PROCESS_TOAST",
"TRUNCATE", "PARALLEL", "SKIP_DATABASE_STATS",
- "ONLY_DATABASE_STATS", "BUFFER_USAGE_LIMIT");
+ "ONLY_DATABASE_STATS", "BUFFER_USAGE_LIMIT",
+ "CONCURRENTLY");
else if (TailMatches("FULL|FREEZE|ANALYZE|VERBOSE|DISABLE_PAGE_SKIPPING|SKIP_LOCKED|PROCESS_MAIN|PROCESS_TOAST|TRUNCATE|SKIP_DATABASE_STATS|ONLY_DATABASE_STATS"))
COMPLETE_WITH("ON", "OFF");
else if (TailMatches("INDEX_CLEANUP"))
diff --git a/src/include/access/heapam.h b/src/include/access/heapam.h
index 7d06dad83f..9b1fb15d8c 100644
--- a/src/include/access/heapam.h
+++ b/src/include/access/heapam.h
@@ -413,6 +413,10 @@ extern HTSV_Result HeapTupleSatisfiesVacuumHorizon(HeapTuple htup, Buffer buffer
TransactionId *dead_after);
extern void HeapTupleSetHintBits(HeapTupleHeader tuple, Buffer buffer,
uint16 infomask, TransactionId xid);
+extern bool HeapTupleMVCCInserted(HeapTuple htup, Snapshot snapshot,
+ Buffer buffer);
+extern bool HeapTupleMVCCNotDeleted(HeapTuple htup, Snapshot snapshot,
+ Buffer buffer);
extern bool HeapTupleHeaderIsOnlyLocked(HeapTupleHeader tuple);
extern bool HeapTupleIsSurelyDead(HeapTuple htup,
struct GlobalVisState *vistest);
diff --git a/src/include/access/tableam.h b/src/include/access/tableam.h
index 09b9b394e0..8efe1c23de 100644
--- a/src/include/access/tableam.h
+++ b/src/include/access/tableam.h
@@ -21,6 +21,7 @@
#include "access/sdir.h"
#include "access/xact.h"
#include "executor/tuptable.h"
+#include "replication/logical.h"
#include "storage/read_stream.h"
#include "utils/rel.h"
#include "utils/snapshot.h"
@@ -629,6 +630,8 @@ typedef struct TableAmRoutine
Relation OldIndex,
bool use_sort,
TransactionId OldestXmin,
+ Snapshot snapshot,
+ LogicalDecodingContext *decoding_ctx,
TransactionId *xid_cutoff,
MultiXactId *multi_cutoff,
double *num_tuples,
@@ -1672,6 +1675,10 @@ table_relation_copy_data(Relation rel, const RelFileLocator *newrlocator)
* not needed for the relation's AM
* - *xid_cutoff - ditto
* - *multi_cutoff - ditto
+ * - snapshot - if != NULL, ignore data changes done by transactions that this
+ * (MVCC) snapshot considers still in-progress or in the future.
+ * - decoding_ctx - logical decoding context, to capture concurrent data
+ * changes.
*
* Output parameters:
* - *xid_cutoff - rel's new relfrozenxid value, may be invalid
@@ -1684,6 +1691,8 @@ table_relation_copy_for_cluster(Relation OldTable, Relation NewTable,
Relation OldIndex,
bool use_sort,
TransactionId OldestXmin,
+ Snapshot snapshot,
+ LogicalDecodingContext *decoding_ctx,
TransactionId *xid_cutoff,
MultiXactId *multi_cutoff,
double *num_tuples,
@@ -1692,6 +1701,7 @@ table_relation_copy_for_cluster(Relation OldTable, Relation NewTable,
{
OldTable->rd_tableam->relation_copy_for_cluster(OldTable, NewTable, OldIndex,
use_sort, OldestXmin,
+ snapshot, decoding_ctx,
xid_cutoff, multi_cutoff,
num_tuples, tups_vacuumed,
tups_recently_dead);
diff --git a/src/include/catalog/index.h b/src/include/catalog/index.h
index 4daa8bef5e..66431cc19e 100644
--- a/src/include/catalog/index.h
+++ b/src/include/catalog/index.h
@@ -100,6 +100,9 @@ extern Oid index_concurrently_create_copy(Relation heapRelation,
Oid tablespaceOid,
const char *newName);
+extern NullableDatum *get_index_stattargets(Oid indexid,
+ IndexInfo *indInfo);
+
extern void index_concurrently_build(Oid heapRelationId,
Oid indexRelationId);
diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h
index 60088a64cb..d420930d6b 100644
--- a/src/include/commands/cluster.h
+++ b/src/include/commands/cluster.h
@@ -13,10 +13,15 @@
#ifndef CLUSTER_H
#define CLUSTER_H
+#include "nodes/execnodes.h"
#include "nodes/parsenodes.h"
#include "parser/parse_node.h"
+#include "replication/logical.h"
#include "storage/lock.h"
+#include "storage/relfilelocator.h"
#include "utils/relcache.h"
+#include "utils/resowner.h"
+#include "utils/tuplestore.h"
/* flag bits for ClusterParams->options */
@@ -24,6 +29,7 @@
#define CLUOPT_RECHECK 0x02 /* recheck relation state */
#define CLUOPT_RECHECK_ISCLUSTERED 0x04 /* recheck relation state for
* indisclustered */
+#define CLUOPT_CONCURRENT 0x08 /* allow concurrent data changes */
/* options for CLUSTER */
typedef struct ClusterParams
@@ -31,12 +37,91 @@ typedef struct ClusterParams
bits32 options; /* bitmask of CLUOPT_* */
} ClusterParams;
+/*
+ * The following definitions are used for concurrent processing.
+ */
+
+extern RelFileLocator clustered_rel_locator;
+extern RelFileLocator clustered_rel_toast_locator;
+
+typedef enum
+{
+ CHANGE_INSERT,
+ CHANGE_UPDATE_OLD,
+ CHANGE_UPDATE_NEW,
+ CHANGE_DELETE,
+ CHANGE_TRUNCATE
+} ConcurrentChangeKind;
+
+typedef struct ConcurrentChange
+{
+ /* See the enum above. */
+ ConcurrentChangeKind kind;
+
+ /*
+ * The actual tuple.
+ *
+ * The tuple data follows the ConcurrentChange structure. Before use make
+ * sure the tuple is correctly aligned (ConcurrentChange can be stored as
+ * bytea) and that tuple->t_data is fixed.
+ */
+ HeapTupleData tup_data;
+} ConcurrentChange;
+
+#define SizeOfConcurrentChange (offsetof(ConcurrentChange, tup_data) + \
+ sizeof(HeapTupleData))
+
+/*
+ * Logical decoding state.
+ *
+ * Here we store the data changes that we decode from WAL while the table
+ * contents is being copied to a new storage. Also the necessary metadata
+ * needed to apply these changes to the table is stored here.
+ */
+typedef struct ClusterDecodingState
+{
+ /* The relation whose changes we're decoding. */
+ Oid relid;
+
+ /*
+ * Decoded changes are stored here. Although we try to avoid excessive
+ * batches, it can happen that the changes need to be stored to disk. The
+ * tuplestore does this transparently.
+ */
+ Tuplestorestate *tstore;
+
+ /* The current number of changes in tstore. */
+ double nchanges;
+
+ /*
+ * Descriptor to store the ConcurrentChange structure serialized (bytea).
+ * We can't store the tuple directly because tuplestore only supports
+ * minimum tuple and we may need to transfer OID system column from the
+ * output plugin. Also we need to transfer the change kind, so it's better
+ * to put everything in the structure than to use 2 tuplestores "in
+ * parallel".
+ */
+ TupleDesc tupdesc_change;
+
+ /* Tuple descriptor needed to update indexes. */
+ TupleDesc tupdesc;
+
+ /* Slot to retrieve data from tstore. */
+ TupleTableSlot *tsslot;
+
+ ResourceOwner resowner;
+} ClusterDecodingState;
+
extern void cluster(ParseState *pstate, ClusterStmt *stmt, bool isTopLevel);
-extern void cluster_rel(Relation OldHeap, Oid indexOid, ClusterParams *params);
+extern void cluster_rel(Relation OldHeap, Oid indexOid, ClusterParams *params,
+ bool isTopLevel, bool isVacuum);
extern void check_index_is_clusterable(Relation OldHeap, Oid indexOid,
LOCKMODE lockmode);
extern void mark_index_clustered(Relation rel, Oid indexOid, bool is_internal);
-
+extern void check_relation_is_clusterable_concurrently(Relation rel,
+ bool is_vacuum);
+extern void cluster_decode_concurrent_changes(LogicalDecodingContext *ctx,
+ XLogRecPtr end_of_wal);
extern Oid make_new_heap(Oid OIDOldHeap, Oid NewTableSpace, Oid NewAccessMethod,
char relpersistence, LOCKMODE lockmode);
extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap,
@@ -44,8 +129,13 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap,
bool swap_toast_by_content,
bool check_constraints,
bool is_internal,
+ bool reindex,
TransactionId frozenXid,
MultiXactId cutoffMulti,
char newrelpersistence);
+extern Size ClusterShmemSize(void);
+extern void ClusterShmemInit(void);
+extern bool is_concurrent_cluster_in_progress(Oid relid);
+extern void check_for_concurrent_cluster(Oid relid, LOCKMODE lockmode);
#endif /* CLUSTER_H */
diff --git a/src/include/commands/progress.h b/src/include/commands/progress.h
index 18e3179ef6..e1d53f2a0a 100644
--- a/src/include/commands/progress.h
+++ b/src/include/commands/progress.h
@@ -59,19 +59,22 @@
#define PROGRESS_CLUSTER_PHASE 1
#define PROGRESS_CLUSTER_INDEX_RELID 2
#define PROGRESS_CLUSTER_HEAP_TUPLES_SCANNED 3
-#define PROGRESS_CLUSTER_HEAP_TUPLES_WRITTEN 4
-#define PROGRESS_CLUSTER_TOTAL_HEAP_BLKS 5
-#define PROGRESS_CLUSTER_HEAP_BLKS_SCANNED 6
-#define PROGRESS_CLUSTER_INDEX_REBUILD_COUNT 7
+#define PROGRESS_CLUSTER_HEAP_TUPLES_INSERTED 4
+#define PROGRESS_CLUSTER_HEAP_TUPLES_UPDATED 5
+#define PROGRESS_CLUSTER_HEAP_TUPLES_DELETED 6
+#define PROGRESS_CLUSTER_TOTAL_HEAP_BLKS 7
+#define PROGRESS_CLUSTER_HEAP_BLKS_SCANNED 8
+#define PROGRESS_CLUSTER_INDEX_REBUILD_COUNT 9
/* Phases of cluster (as advertised via PROGRESS_CLUSTER_PHASE) */
#define PROGRESS_CLUSTER_PHASE_SEQ_SCAN_HEAP 1
#define PROGRESS_CLUSTER_PHASE_INDEX_SCAN_HEAP 2
#define PROGRESS_CLUSTER_PHASE_SORT_TUPLES 3
#define PROGRESS_CLUSTER_PHASE_WRITE_NEW_HEAP 4
-#define PROGRESS_CLUSTER_PHASE_SWAP_REL_FILES 5
-#define PROGRESS_CLUSTER_PHASE_REBUILD_INDEX 6
-#define PROGRESS_CLUSTER_PHASE_FINAL_CLEANUP 7
+#define PROGRESS_CLUSTER_PHASE_CATCH_UP 5
+#define PROGRESS_CLUSTER_PHASE_SWAP_REL_FILES 6
+#define PROGRESS_CLUSTER_PHASE_REBUILD_INDEX 7
+#define PROGRESS_CLUSTER_PHASE_FINAL_CLEANUP 8
/* Commands of PROGRESS_CLUSTER */
#define PROGRESS_CLUSTER_COMMAND_CLUSTER 1
diff --git a/src/include/commands/vacuum.h b/src/include/commands/vacuum.h
index 12d0b61950..05c1fce969 100644
--- a/src/include/commands/vacuum.h
+++ b/src/include/commands/vacuum.h
@@ -181,13 +181,16 @@ typedef struct VacAttrStats
#define VACOPT_ANALYZE 0x02 /* do ANALYZE */
#define VACOPT_VERBOSE 0x04 /* output INFO instrumentation messages */
#define VACOPT_FREEZE 0x08 /* FREEZE option */
-#define VACOPT_FULL 0x10 /* FULL (non-concurrent) vacuum */
-#define VACOPT_SKIP_LOCKED 0x20 /* skip if cannot get lock */
-#define VACOPT_PROCESS_MAIN 0x40 /* process main relation */
-#define VACOPT_PROCESS_TOAST 0x80 /* process the TOAST table, if any */
-#define VACOPT_DISABLE_PAGE_SKIPPING 0x100 /* don't skip any pages */
-#define VACOPT_SKIP_DATABASE_STATS 0x200 /* skip vac_update_datfrozenxid() */
-#define VACOPT_ONLY_DATABASE_STATS 0x400 /* only vac_update_datfrozenxid() */
+#define VACOPT_FULL_EXCLUSIVE 0x10 /* FULL (non-concurrent) vacuum */
+#define VACOPT_FULL_CONCURRENT 0x20 /* FULL (concurrent) vacuum */
+#define VACOPT_SKIP_LOCKED 0x40 /* skip if cannot get lock */
+#define VACOPT_PROCESS_MAIN 0x80 /* process main relation */
+#define VACOPT_PROCESS_TOAST 0x100 /* process the TOAST table, if any */
+#define VACOPT_DISABLE_PAGE_SKIPPING 0x200 /* don't skip any pages */
+#define VACOPT_SKIP_DATABASE_STATS 0x400 /* skip vac_update_datfrozenxid() */
+#define VACOPT_ONLY_DATABASE_STATS 0x800 /* only vac_update_datfrozenxid() */
+
+#define VACOPT_FULL (VACOPT_FULL_EXCLUSIVE | VACOPT_FULL_CONCURRENT)
/*
* Values used by index_cleanup and truncate params.
diff --git a/src/include/replication/snapbuild.h b/src/include/replication/snapbuild.h
index 6d4d2d1814..7c2e1f2ceb 100644
--- a/src/include/replication/snapbuild.h
+++ b/src/include/replication/snapbuild.h
@@ -73,6 +73,7 @@ extern void FreeSnapshotBuilder(SnapBuild *builder);
extern void SnapBuildSnapDecRefcount(Snapshot snap);
extern Snapshot SnapBuildInitialSnapshot(SnapBuild *builder);
+extern Snapshot SnapBuildInitialSnapshotForCluster(SnapBuild *builder);
extern Snapshot SnapBuildMVCCFromHistoric(Snapshot snapshot, bool in_place);
extern const char *SnapBuildExportSnapshot(SnapBuild *builder);
extern void SnapBuildClearExportedSnapshot(void);
diff --git a/src/include/storage/lockdefs.h b/src/include/storage/lockdefs.h
index 7f3ba0352f..b7b94dacf8 100644
--- a/src/include/storage/lockdefs.h
+++ b/src/include/storage/lockdefs.h
@@ -36,8 +36,9 @@ typedef int LOCKMODE;
#define AccessShareLock 1 /* SELECT */
#define RowShareLock 2 /* SELECT FOR UPDATE/FOR SHARE */
#define RowExclusiveLock 3 /* INSERT, UPDATE, DELETE */
-#define ShareUpdateExclusiveLock 4 /* VACUUM (non-FULL), ANALYZE, CREATE
- * INDEX CONCURRENTLY */
+#define ShareUpdateExclusiveLock 4 /* VACUUM (non-exclusive), ANALYZE, CREATE
+ * INDEX CONCURRENTLY, CLUSTER
+ * CONCURRENTLY */
#define ShareLock 5 /* CREATE INDEX (WITHOUT CONCURRENTLY) */
#define ShareRowExclusiveLock 6 /* like EXCLUSIVE MODE, but allows ROW
* SHARE */
diff --git a/src/include/storage/lwlocklist.h b/src/include/storage/lwlocklist.h
index cf56545238..9211124d10 100644
--- a/src/include/storage/lwlocklist.h
+++ b/src/include/storage/lwlocklist.h
@@ -83,3 +83,4 @@ PG_LWLOCK(49, WALSummarizer)
PG_LWLOCK(50, DSMRegistry)
PG_LWLOCK(51, InjectionPoint)
PG_LWLOCK(52, SerialControl)
+PG_LWLOCK(54, ClusteredRels)
diff --git a/src/include/utils/backend_progress.h b/src/include/utils/backend_progress.h
index 739629cb21..5b9903de08 100644
--- a/src/include/utils/backend_progress.h
+++ b/src/include/utils/backend_progress.h
@@ -35,7 +35,7 @@ typedef enum ProgressCommandType
/*
* Any command which wishes can advertise that it is running by setting
- * command, command_target, and param[]. command_target should be the OID of
+ * ommand, command_target, and param[]. command_target should be the OID of
* the relation which the command targets (we assume there's just one, as this
* is meant for utility commands), but the meaning of each element in the
* param array is command-specific.
@@ -55,6 +55,7 @@ extern void pgstat_progress_parallel_incr_param(int index, int64 incr);
extern void pgstat_progress_update_multi_param(int nparam, const int *index,
const int64 *val);
extern void pgstat_progress_end_command(void);
+extern void pgstat_progress_restore_state(PgBackendProgress *backup);
#endif /* BACKEND_PROGRESS_H */
diff --git a/src/include/utils/inval.h b/src/include/utils/inval.h
index 40658ba2ff..6b2faed672 100644
--- a/src/include/utils/inval.h
+++ b/src/include/utils/inval.h
@@ -49,6 +49,8 @@ extern void CacheInvalidateCatalog(Oid catalogId);
extern void CacheInvalidateRelcache(Relation relation);
+extern void CacheInvalidateRelcacheImmediate(Relation relation);
+
extern void CacheInvalidateRelcacheAll(void);
extern void CacheInvalidateRelcacheByTuple(HeapTuple classTuple);
diff --git a/src/include/utils/rel.h b/src/include/utils/rel.h
index 33d1e4a4e2..243b80d7fa 100644
--- a/src/include/utils/rel.h
+++ b/src/include/utils/rel.h
@@ -253,6 +253,9 @@ typedef struct RelationData
bool pgstat_enabled; /* should relation stats be counted */
/* use "struct" here to avoid needing to include pgstat.h: */
struct PgStat_TableStatus *pgstat_info; /* statistics collection area */
+
+ /* Is CLUSTER CONCURRENTLY being performed on this relation? */
+ bool rd_cluster_concurrent;
} RelationData;
@@ -684,7 +687,9 @@ RelationCloseSmgr(Relation relation)
#define RelationIsAccessibleInLogicalDecoding(relation) \
(XLogLogicalInfoActive() && \
RelationNeedsWAL(relation) && \
- (IsCatalogRelation(relation) || RelationIsUsedAsCatalogTable(relation)))
+ (IsCatalogRelation(relation) || \
+ RelationIsUsedAsCatalogTable(relation) || \
+ (relation)->rd_cluster_concurrent))
/*
* RelationIsLogicallyLogged
diff --git a/src/include/utils/snapmgr.h b/src/include/utils/snapmgr.h
index 147b190210..5eeabdc6c4 100644
--- a/src/include/utils/snapmgr.h
+++ b/src/include/utils/snapmgr.h
@@ -61,6 +61,8 @@ extern Snapshot GetLatestSnapshot(void);
extern void SnapshotSetCommandId(CommandId curcid);
extern Snapshot CopySnapshot(Snapshot snapshot);
+extern void FreeSnapshot(Snapshot snapshot);
+
extern Snapshot GetCatalogSnapshot(Oid relid);
extern Snapshot GetNonHistoricCatalogSnapshot(Oid relid);
extern void InvalidateCatalogSnapshot(void);
diff --git a/src/test/regress/expected/rules.out b/src/test/regress/expected/rules.out
index 3014d047fe..81300642a5 100644
--- a/src/test/regress/expected/rules.out
+++ b/src/test/regress/expected/rules.out
@@ -1962,17 +1962,20 @@ pg_stat_progress_cluster| SELECT s.pid,
WHEN 2 THEN 'index scanning heap'::text
WHEN 3 THEN 'sorting tuples'::text
WHEN 4 THEN 'writing new heap'::text
- WHEN 5 THEN 'swapping relation files'::text
- WHEN 6 THEN 'rebuilding index'::text
- WHEN 7 THEN 'performing final cleanup'::text
+ WHEN 5 THEN 'catch-up'::text
+ WHEN 6 THEN 'swapping relation files'::text
+ WHEN 7 THEN 'rebuilding index'::text
+ WHEN 8 THEN 'performing final cleanup'::text
ELSE NULL::text
END AS phase,
(s.param3)::oid AS cluster_index_relid,
s.param4 AS heap_tuples_scanned,
- s.param5 AS heap_tuples_written,
- s.param6 AS heap_blks_total,
- s.param7 AS heap_blks_scanned,
- s.param8 AS index_rebuild_count
+ s.param5 AS heap_tuples_inserted,
+ s.param6 AS heap_tuples_updated,
+ s.param7 AS heap_tuples_deleted,
+ s.param8 AS heap_blks_total,
+ s.param9 AS heap_blks_scanned,
+ s.param10 AS index_rebuild_count
FROM (pg_stat_get_progress_info('CLUSTER'::text) s(pid, datid, relid, param1, param2, param3, param4, param5, param6, param7, param8, param9, param10, param11, param12, param13, param14, param15, param16, param17, param18, param19, param20)
LEFT JOIN pg_database d ON ((s.datid = d.oid)));
pg_stat_progress_copy| SELECT s.pid,
--
2.45.2
[text/x-diff] v07-0005-Preserve-visibility-information-of-the-concurrent-da.patch (39.7K, ../6250.1736776111@antos/5-v07-0005-Preserve-visibility-information-of-the-concurrent-da.patch)
download | inline diff:
From 46a22614127c4cf1d45a0c47ccafbfac3243fee5 Mon Sep 17 00:00:00 2001
From: Antonin Houska <[email protected]>
Date: Mon, 13 Jan 2025 14:29:54 +0100
Subject: [PATCH 5/8] Preserve visibility information of the concurrent data
changes.
As explained in the commit message of the preceding patch of the series, the
data changes done by applications while VACUUM FULL / CLUSTER CONCURRENTLY is
copying the table contents to a new file are decoded from WAL and eventually
also applied to the new file. To reduce the complexity a little bit, the
preceding patch uses the current transaction (i.e. transaction opened by the
VACUUM FULL / CLUSTER command) to execute those INSERT, UPDATE and DELETE
commands.
However, neither VACUUM nor CLUSTER is expected to change visibility of
tuples. Therefore, this patch fixes the handling of the "concurrent data
changes". Now the tuples written into the new table storage have the same XID
and command ID (CID) as they had in the old storage.
Related change we do here is that the data changes (INSERT, UPDATE, DELETE) we
"replay" on the new storage are not logically decoded. First, the logical
decoding subsystem does not expect that already committed transaction is
decoded again. Second, repeated decoding would be just wasted effort.
---
src/backend/access/common/toast_internals.c | 3 +-
src/backend/access/heap/heapam.c | 73 ++++++++----
src/backend/access/heap/heapam_handler.c | 14 ++-
src/backend/access/transam/xact.c | 52 ++++++++
src/backend/commands/cluster.c | 111 ++++++++++++++++--
src/backend/replication/logical/decode.c | 76 ++++++++++--
src/backend/replication/logical/snapbuild.c | 22 ++--
.../pgoutput_cluster/pgoutput_cluster.c | 68 +++++++++--
src/include/access/heapam.h | 15 ++-
src/include/access/heapam_xlog.h | 2 +
src/include/access/xact.h | 2 +
src/include/commands/cluster.h | 18 +++
src/include/utils/snapshot.h | 3 +
13 files changed, 389 insertions(+), 70 deletions(-)
diff --git a/src/backend/access/common/toast_internals.c b/src/backend/access/common/toast_internals.c
index 7d8be8346c..75d889ec72 100644
--- a/src/backend/access/common/toast_internals.c
+++ b/src/backend/access/common/toast_internals.c
@@ -320,7 +320,8 @@ toast_save_datum(Relation rel, Datum value,
memcpy(VARDATA(&chunk_data), data_p, chunk_size);
toasttup = heap_form_tuple(toasttupDesc, t_values, t_isnull);
- heap_insert(toastrel, toasttup, mycid, options, NULL);
+ heap_insert(toastrel, toasttup, GetCurrentTransactionId(), mycid,
+ options, NULL);
/*
* Create the index entry. We cheat a little here by not using
diff --git a/src/backend/access/heap/heapam.c b/src/backend/access/heap/heapam.c
index 552993d4ef..e20634c030 100644
--- a/src/backend/access/heap/heapam.c
+++ b/src/backend/access/heap/heapam.c
@@ -58,7 +58,8 @@ static HeapTuple heap_prepare_insert(Relation relation, HeapTuple tup,
static XLogRecPtr log_heap_update(Relation reln, Buffer oldbuf,
Buffer newbuf, HeapTuple oldtup,
HeapTuple newtup, HeapTuple old_key_tuple,
- bool all_visible_cleared, bool new_all_visible_cleared);
+ bool all_visible_cleared, bool new_all_visible_cleared,
+ bool wal_logical);
#ifdef USE_ASSERT_CHECKING
static void check_lock_if_inplace_updateable_rel(Relation relation,
ItemPointer otid,
@@ -1969,7 +1970,7 @@ ReleaseBulkInsertStatePin(BulkInsertState bistate)
/*
* heap_insert - insert tuple into a heap
*
- * The new tuple is stamped with current transaction ID and the specified
+ * The new tuple is stamped with specified transaction ID and the specified
* command ID.
*
* See table_tuple_insert for comments about most of the input flags, except
@@ -1985,15 +1986,16 @@ ReleaseBulkInsertStatePin(BulkInsertState bistate)
* reflected into *tup.
*/
void
-heap_insert(Relation relation, HeapTuple tup, CommandId cid,
- int options, BulkInsertState bistate)
+heap_insert(Relation relation, HeapTuple tup, TransactionId xid,
+ CommandId cid, int options, BulkInsertState bistate)
{
- TransactionId xid = GetCurrentTransactionId();
HeapTuple heaptup;
Buffer buffer;
Buffer vmbuffer = InvalidBuffer;
bool all_visible_cleared = false;
+ Assert(TransactionIdIsValid(xid));
+
/* Cheap, simplistic check that the tuple matches the rel's rowtype. */
Assert(HeapTupleHeaderGetNatts(tup->t_data) <=
RelationGetNumberOfAttributes(relation));
@@ -2624,7 +2626,8 @@ heap_multi_insert(Relation relation, TupleTableSlot **slots, int ntuples,
void
simple_heap_insert(Relation relation, HeapTuple tup)
{
- heap_insert(relation, tup, GetCurrentCommandId(true), 0, NULL);
+ heap_insert(relation, tup, GetCurrentTransactionId(),
+ GetCurrentCommandId(true), 0, NULL);
}
/*
@@ -2681,11 +2684,11 @@ xmax_infomask_changed(uint16 new_infomask, uint16 old_infomask)
*/
TM_Result
heap_delete(Relation relation, ItemPointer tid,
- CommandId cid, Snapshot crosscheck, bool wait,
- TM_FailureData *tmfd, bool changingPart)
+ TransactionId xid, CommandId cid, Snapshot crosscheck, bool wait,
+ TM_FailureData *tmfd, bool changingPart,
+ bool wal_logical)
{
TM_Result result;
- TransactionId xid = GetCurrentTransactionId();
ItemId lp;
HeapTupleData tp;
Page page;
@@ -2702,6 +2705,7 @@ heap_delete(Relation relation, ItemPointer tid,
bool old_key_copied = false;
Assert(ItemPointerIsValid(tid));
+ Assert(TransactionIdIsValid(xid));
/*
* Forbid this during a parallel operation, lest it allocate a combo CID.
@@ -2927,7 +2931,8 @@ l1:
* Compute replica identity tuple before entering the critical section so
* we don't PANIC upon a memory allocation failure.
*/
- old_key_tuple = ExtractReplicaIdentity(relation, &tp, true, &old_key_copied);
+ old_key_tuple = wal_logical ?
+ ExtractReplicaIdentity(relation, &tp, true, &old_key_copied) : NULL;
/*
* If this is the first possibly-multixact-able operation in the current
@@ -2995,8 +3000,12 @@ l1:
/*
* For logical decode we need combo CIDs to properly decode the
* catalog
+ *
+ * Like in heap_insert(), visibility is unchanged when called from
+ * VACUUM FULL / CLUSTER.
*/
- if (RelationIsAccessibleInLogicalDecoding(relation))
+ if (wal_logical &&
+ RelationIsAccessibleInLogicalDecoding(relation))
log_heap_new_cid(relation, &tp);
xlrec.flags = 0;
@@ -3017,6 +3026,15 @@ l1:
xlrec.flags |= XLH_DELETE_CONTAINS_OLD_KEY;
}
+ /*
+ * Unlike UPDATE, DELETE is decoded even if there is no old key, so it
+ * does not help to clear both XLH_DELETE_CONTAINS_OLD_TUPLE and
+ * XLH_DELETE_CONTAINS_OLD_KEY. Thus we need an extra flag. TODO
+ * Consider not decoding tuples w/o the old tuple/key instead.
+ */
+ if (!wal_logical)
+ xlrec.flags |= XLH_DELETE_NO_LOGICAL;
+
XLogBeginInsert();
XLogRegisterData((char *) &xlrec, SizeOfHeapDelete);
@@ -3106,10 +3124,11 @@ simple_heap_delete(Relation relation, ItemPointer tid)
TM_Result result;
TM_FailureData tmfd;
- result = heap_delete(relation, tid,
+ result = heap_delete(relation, tid, GetCurrentTransactionId(),
GetCurrentCommandId(true), InvalidSnapshot,
true /* wait for commit */ ,
- &tmfd, false /* changingPart */ );
+ &tmfd, false, /* changingPart */
+ true /* wal_logical */);
switch (result)
{
case TM_SelfModified:
@@ -3148,12 +3167,11 @@ simple_heap_delete(Relation relation, ItemPointer tid)
*/
TM_Result
heap_update(Relation relation, ItemPointer otid, HeapTuple newtup,
- CommandId cid, Snapshot crosscheck, bool wait,
- TM_FailureData *tmfd, LockTupleMode *lockmode,
- TU_UpdateIndexes *update_indexes)
+ TransactionId xid, CommandId cid, Snapshot crosscheck,
+ bool wait, TM_FailureData *tmfd, LockTupleMode *lockmode,
+ TU_UpdateIndexes *update_indexes, bool wal_logical)
{
TM_Result result;
- TransactionId xid = GetCurrentTransactionId();
Bitmapset *hot_attrs;
Bitmapset *sum_attrs;
Bitmapset *key_attrs;
@@ -3193,6 +3211,7 @@ heap_update(Relation relation, ItemPointer otid, HeapTuple newtup,
infomask2_new_tuple;
Assert(ItemPointerIsValid(otid));
+ Assert(TransactionIdIsValid(xid));
/* Cheap, simplistic check that the tuple matches the rel's rowtype. */
Assert(HeapTupleHeaderGetNatts(newtup->t_data) <=
@@ -3985,8 +4004,12 @@ l2:
/*
* For logical decoding we need combo CIDs to properly decode the
* catalog.
+ *
+ * Like in heap_insert(), visibility is unchanged when called from
+ * VACUUM FULL / CLUSTER.
*/
- if (RelationIsAccessibleInLogicalDecoding(relation))
+ if (wal_logical &&
+ RelationIsAccessibleInLogicalDecoding(relation))
{
log_heap_new_cid(relation, &oldtup);
log_heap_new_cid(relation, heaptup);
@@ -3996,7 +4019,8 @@ l2:
newbuf, &oldtup, heaptup,
old_key_tuple,
all_visible_cleared,
- all_visible_cleared_new);
+ all_visible_cleared_new,
+ wal_logical);
if (newbuf != buffer)
{
PageSetLSN(BufferGetPage(newbuf), recptr);
@@ -4351,10 +4375,10 @@ simple_heap_update(Relation relation, ItemPointer otid, HeapTuple tup,
TM_FailureData tmfd;
LockTupleMode lockmode;
- result = heap_update(relation, otid, tup,
+ result = heap_update(relation, otid, tup, GetCurrentTransactionId(),
GetCurrentCommandId(true), InvalidSnapshot,
true /* wait for commit */ ,
- &tmfd, &lockmode, update_indexes);
+ &tmfd, &lockmode, update_indexes, true);
switch (result)
{
case TM_SelfModified:
@@ -8685,7 +8709,8 @@ static XLogRecPtr
log_heap_update(Relation reln, Buffer oldbuf,
Buffer newbuf, HeapTuple oldtup, HeapTuple newtup,
HeapTuple old_key_tuple,
- bool all_visible_cleared, bool new_all_visible_cleared)
+ bool all_visible_cleared, bool new_all_visible_cleared,
+ bool wal_logical)
{
xl_heap_update xlrec;
xl_heap_header xlhdr;
@@ -8696,10 +8721,12 @@ log_heap_update(Relation reln, Buffer oldbuf,
suffixlen = 0;
XLogRecPtr recptr;
Page page = BufferGetPage(newbuf);
- bool need_tuple_data = RelationIsLogicallyLogged(reln);
+ bool need_tuple_data;
bool init;
int bufflags;
+ need_tuple_data = RelationIsLogicallyLogged(reln) && wal_logical;
+
/* Caller should not call me on a non-WAL-logged relation */
Assert(RelationNeedsWAL(reln));
diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c
index c5ec21ca2f..c315abac02 100644
--- a/src/backend/access/heap/heapam_handler.c
+++ b/src/backend/access/heap/heapam_handler.c
@@ -252,7 +252,8 @@ heapam_tuple_insert(Relation relation, TupleTableSlot *slot, CommandId cid,
tuple->t_tableOid = slot->tts_tableOid;
/* Perform the insertion, and copy the resulting ItemPointer */
- heap_insert(relation, tuple, cid, options, bistate);
+ heap_insert(relation, tuple, GetCurrentTransactionId(), cid, options,
+ bistate);
ItemPointerCopy(&tuple->t_self, &slot->tts_tid);
if (shouldFree)
@@ -275,7 +276,8 @@ heapam_tuple_insert_speculative(Relation relation, TupleTableSlot *slot,
options |= HEAP_INSERT_SPECULATIVE;
/* Perform the insertion, and copy the resulting ItemPointer */
- heap_insert(relation, tuple, cid, options, bistate);
+ heap_insert(relation, tuple, GetCurrentTransactionId(), cid, options,
+ bistate);
ItemPointerCopy(&tuple->t_self, &slot->tts_tid);
if (shouldFree)
@@ -309,7 +311,8 @@ heapam_tuple_delete(Relation relation, ItemPointer tid, CommandId cid,
* the storage itself is cleaning the dead tuples by itself, it is the
* time to call the index tuple deletion also.
*/
- return heap_delete(relation, tid, cid, crosscheck, wait, tmfd, changingPart);
+ return heap_delete(relation, tid, GetCurrentTransactionId(), cid,
+ crosscheck, wait, tmfd, changingPart, true);
}
@@ -327,8 +330,9 @@ heapam_tuple_update(Relation relation, ItemPointer otid, TupleTableSlot *slot,
slot->tts_tableOid = RelationGetRelid(relation);
tuple->t_tableOid = slot->tts_tableOid;
- result = heap_update(relation, otid, tuple, cid, crosscheck, wait,
- tmfd, lockmode, update_indexes);
+ result = heap_update(relation, otid, tuple, GetCurrentTransactionId(),
+ cid, crosscheck, wait,
+ tmfd, lockmode, update_indexes, true);
ItemPointerCopy(&tuple->t_self, &slot->tts_tid);
/*
diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index d331ab90d7..e6a7414f9e 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -125,6 +125,18 @@ static FullTransactionId XactTopFullTransactionId = {InvalidTransactionId};
static int nParallelCurrentXids = 0;
static TransactionId *ParallelCurrentXids;
+/*
+ * Another case that requires TransactionIdIsCurrentTransactionId() to behave
+ * specially is when CLUSTER CONCURRENTLY is processing data changes made in
+ * the old storage of a table by other transactions. When applying the changes
+ * to the new storage, the backend executing the CLUSTER command needs to act
+ * on behalf on those other transactions. The transactions responsible for the
+ * changes in the old storage are stored in this array, sorted by
+ * xidComparator.
+ */
+static int nClusterCurrentXids = 0;
+static TransactionId *ClusterCurrentXids = NULL;
+
/*
* Miscellaneous flag bits to record events which occur on the top level
* transaction. These flags are only persisted in MyXactFlags and are intended
@@ -971,6 +983,8 @@ TransactionIdIsCurrentTransactionId(TransactionId xid)
int low,
high;
+ Assert(nClusterCurrentXids == 0);
+
low = 0;
high = nParallelCurrentXids - 1;
while (low <= high)
@@ -990,6 +1004,21 @@ TransactionIdIsCurrentTransactionId(TransactionId xid)
return false;
}
+ /*
+ * When executing CLUSTER CONCURRENTLY, the array of current transactions
+ * is given.
+ */
+ if (nClusterCurrentXids > 0)
+ {
+ Assert(nParallelCurrentXids == 0);
+
+ return bsearch(&xid,
+ ClusterCurrentXids,
+ nClusterCurrentXids,
+ sizeof(TransactionId),
+ xidComparator) != NULL;
+ }
+
/*
* We will return true for the Xid of the current subtransaction, any of
* its subcommitted children, any of its parents, or any of their
@@ -5628,6 +5657,29 @@ EndParallelWorkerTransaction(void)
CurrentTransactionState->blockState = TBLOCK_DEFAULT;
}
+/*
+ * SetClusterCurrentXids
+ * Set the XID array that TransactionIdIsCurrentTransactionId() should
+ * use.
+ */
+void
+SetClusterCurrentXids(TransactionId *xip, int xcnt)
+{
+ ClusterCurrentXids = xip;
+ nClusterCurrentXids = xcnt;
+}
+
+/*
+ * ResetClusterCurrentXids
+ * Undo the effect of SetClusterCurrentXids().
+ */
+void
+ResetClusterCurrentXids(void)
+{
+ ClusterCurrentXids = NULL;
+ nClusterCurrentXids = 0;
+}
+
/*
* ShowTransactionState
* Debug support
diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index c9cc061c45..3a1a51a56a 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -200,6 +200,7 @@ static void apply_concurrent_delete(Relation rel, HeapTuple tup_target,
ConcurrentChange *change);
static HeapTuple find_target_tuple(Relation rel, ScanKey key, int nkeys,
HeapTuple tup_key,
+ Snapshot snapshot,
IndexInsertState *iistate,
TupleTableSlot *ident_slot,
IndexScanDesc *scan_p);
@@ -2971,6 +2972,9 @@ setup_logical_decoding(Oid relid, const char *slotname, TupleDesc tupdesc)
dstate->relid = relid;
dstate->tstore = tuplestore_begin_heap(false, false,
maintenance_work_mem);
+#ifdef USE_ASSERT_CHECKING
+ dstate->last_change_xid = InvalidTransactionId;
+#endif
dstate->tupdesc = tupdesc;
/* Initialize the descriptor to store the changes ... */
@@ -3126,6 +3130,7 @@ apply_concurrent_changes(ClusterDecodingState *dstate, Relation rel,
tup_exist;
char *change_raw, *src;
ConcurrentChange change;
+ Snapshot snapshot;
bool isnull[1];
Datum values[1];
@@ -3194,8 +3199,30 @@ apply_concurrent_changes(ClusterDecodingState *dstate, Relation rel,
/*
* Find the tuple to be updated or deleted.
+ *
+ * As the table being CLUSTERed concurrently is considered an
+ * "user catalog", new CID is WAL-logged and decoded. And since we
+ * use the same XID that the original DMLs did, the snapshot used
+ * for the logical decoding (by now converted to a non-historic
+ * MVCC snapshot) should see the tuples inserted previously into
+ * the new heap and/or updated there.
+ */
+ snapshot = change.snapshot;
+
+ /*
+ * Set what should be considered current transaction (and
+ * subtransactions) during visibility check.
+ *
+ * Note that this snapshot was created from a historic snapshot
+ * using SnapBuildMVCCFromHistoric(), which does not touch
+ * 'subxip'. Thus, unlike in a regular MVCC snapshot, the array
+ * only contains the transactions whose data changes we are
+ * applying, and its subtransactions. That's exactly what we need
+ * to check if particular xact is a "current transaction:".
*/
- tup_exist = find_target_tuple(rel, key, nkeys, tup_key,
+ SetClusterCurrentXids(snapshot->subxip, snapshot->subxcnt);
+
+ tup_exist = find_target_tuple(rel, key, nkeys, tup_key, snapshot,
iistate, ident_slot, &ind_scan);
if (tup_exist == NULL)
elog(ERROR, "Failed to find target tuple");
@@ -3206,6 +3233,8 @@ apply_concurrent_changes(ClusterDecodingState *dstate, Relation rel,
else
apply_concurrent_delete(rel, tup_exist, &change);
+ ResetClusterCurrentXids();
+
if (tup_old != NULL)
{
pfree(tup_old);
@@ -3218,11 +3247,14 @@ apply_concurrent_changes(ClusterDecodingState *dstate, Relation rel,
else
elog(ERROR, "Unrecognized kind of change: %d", change.kind);
- /* If there's any change, make it visible to the next iteration. */
- if (change.kind != CHANGE_UPDATE_OLD)
+ /* Free the snapshot if this is the last change that needed it. */
+ Assert(change.snapshot->active_count > 0);
+ change.snapshot->active_count--;
+ if (change.snapshot->active_count == 0)
{
- CommandCounterIncrement();
- UpdateActiveSnapshotCommandId();
+ if (change.snapshot == dstate->snapshot)
+ dstate->snapshot = NULL;
+ FreeSnapshot(change.snapshot);
}
/* TTSOpsMinimalTuple has .get_heap_tuple==NULL. */
@@ -3242,10 +3274,30 @@ static void
apply_concurrent_insert(Relation rel, ConcurrentChange *change, HeapTuple tup,
IndexInsertState *iistate, TupleTableSlot *index_slot)
{
+ Snapshot snapshot = change->snapshot;
List *recheck;
+ /*
+ * For INSERT, the visibility information is not important, but we use the
+ * snapshot to get CID. Index functions might need the whole snapshot
+ * anyway.
+ */
+ SetClusterCurrentXids(snapshot->subxip, snapshot->subxcnt);
- heap_insert(rel, tup, GetCurrentCommandId(true), HEAP_INSERT_NO_LOGICAL, NULL);
+ /*
+ * Write the tuple into the new heap.
+ *
+ * The snapshot is the one we used to decode the insert (though converted
+ * to "non-historic" MVCC snapshot), i.e. the snapshot's curcid is the
+ * tuple CID incremented by one (due to the "new CID" WAL record that got
+ * written along with the INSERT record). Thus if we want to use the
+ * original CID, we need to subtract 1 from curcid.
+ */
+ Assert(snapshot->curcid != InvalidCommandId &&
+ snapshot->curcid > FirstCommandId);
+
+ heap_insert(rel, tup, change->xid, snapshot->curcid - 1,
+ HEAP_INSERT_NO_LOGICAL, NULL);
/*
* Update indexes.
@@ -3253,6 +3305,7 @@ apply_concurrent_insert(Relation rel, ConcurrentChange *change, HeapTuple tup,
* In case functions in the index need the active snapshot and caller
* hasn't set one.
*/
+ PushActiveSnapshot(snapshot);
ExecStoreHeapTuple(tup, index_slot, false);
recheck = ExecInsertIndexTuples(iistate->rri,
index_slot,
@@ -3263,6 +3316,8 @@ apply_concurrent_insert(Relation rel, ConcurrentChange *change, HeapTuple tup,
NIL, /* arbiterIndexes */
false /* onlySummarizing */
);
+ PopActiveSnapshot();
+ ResetClusterCurrentXids();
/*
* If recheck is required, it must have been preformed on the source
@@ -3280,18 +3335,36 @@ apply_concurrent_update(Relation rel, HeapTuple tup, HeapTuple tup_target,
TupleTableSlot *index_slot)
{
List *recheck;
+ LockTupleMode lockmode;
TU_UpdateIndexes update_indexes;
+ TM_Result res;
+ Snapshot snapshot = change->snapshot;
+ TM_FailureData tmfd;
/*
* Write the new tuple into the new heap. ('tup' gets the TID assigned
* here.)
+ *
+ * Regarding CID, see the comment in apply_concurrent_insert().
*/
- simple_heap_update(rel, &tup_target->t_self, tup, &update_indexes);
+ Assert(snapshot->curcid != InvalidCommandId &&
+ snapshot->curcid > FirstCommandId);
+
+ res = heap_update(rel, &tup_target->t_self, tup,
+ change->xid, snapshot->curcid - 1,
+ InvalidSnapshot,
+ false, /* no wait - only we are doing changes */
+ &tmfd, &lockmode, &update_indexes,
+ /* wal_logical */
+ false);
+ if (res != TM_Ok)
+ ereport(ERROR, (errmsg("failed to apply concurrent UPDATE")));
ExecStoreHeapTuple(tup, index_slot, false);
if (update_indexes != TU_None)
{
+ PushActiveSnapshot(snapshot);
recheck = ExecInsertIndexTuples(iistate->rri,
index_slot,
iistate->estate,
@@ -3301,6 +3374,7 @@ apply_concurrent_update(Relation rel, HeapTuple tup, HeapTuple tup_target,
NIL, /* arbiterIndexes */
/* onlySummarizing */
update_indexes == TU_Summarizing);
+ PopActiveSnapshot();
list_free(recheck);
}
@@ -3311,7 +3385,22 @@ static void
apply_concurrent_delete(Relation rel, HeapTuple tup_target,
ConcurrentChange *change)
{
- simple_heap_delete(rel, &tup_target->t_self);
+ TM_Result res;
+ TM_FailureData tmfd;
+ Snapshot snapshot = change->snapshot;
+
+ /* Regarding CID, see the comment in apply_concurrent_insert(). */
+ Assert(snapshot->curcid != InvalidCommandId &&
+ snapshot->curcid > FirstCommandId);
+
+ res = heap_delete(rel, &tup_target->t_self, change->xid,
+ snapshot->curcid - 1, InvalidSnapshot, false,
+ &tmfd, false,
+ /* wal_logical */
+ false);
+
+ if (res != TM_Ok)
+ ereport(ERROR, (errmsg("failed to apply concurrent DELETE")));
pgstat_progress_incr_param(PROGRESS_CLUSTER_HEAP_TUPLES_DELETED, 1);
}
@@ -3329,7 +3418,7 @@ apply_concurrent_delete(Relation rel, HeapTuple tup_target,
*/
static HeapTuple
find_target_tuple(Relation rel, ScanKey key, int nkeys, HeapTuple tup_key,
- IndexInsertState *iistate,
+ Snapshot snapshot, IndexInsertState *iistate,
TupleTableSlot *ident_slot, IndexScanDesc *scan_p)
{
IndexScanDesc scan;
@@ -3337,7 +3426,7 @@ find_target_tuple(Relation rel, ScanKey key, int nkeys, HeapTuple tup_key,
int2vector *ident_indkey;
HeapTuple result = NULL;
- scan = index_beginscan(rel, iistate->ident_index, GetActiveSnapshot(),
+ scan = index_beginscan(rel, iistate->ident_index, snapshot,
nkeys, 0);
*scan_p = scan;
index_rescan(scan, key, nkeys, NULL, 0);
@@ -3409,6 +3498,8 @@ process_concurrent_changes(LogicalDecodingContext *ctx, XLogRecPtr end_of_wal,
}
PG_FINALLY();
{
+ ResetClusterCurrentXids();
+
if (rel_src)
rel_dst->rd_toastoid = InvalidOid;
}
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index 8f45a7a168..23766ccfb6 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -469,9 +469,18 @@ heap_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
SnapBuild *builder = ctx->snapshot_builder;
/*
- * Check if CLUSTER CONCURRENTLY is being performed by this backend. If
- * so, only decode data changes of the table that it is processing, and
- * the changes of its TOAST relation.
+ * If the change is not intended for logical decoding, do not even
+ * establish transaction for it. This is particularly important if the
+ * record was generated by CLUSTER CONCURRENTLY because this command uses
+ * the original XID when doing changes in the new storage. The decoding
+ * subsystem probably does not expect to see the same transaction multiple
+ * times.
+ */
+
+ /*
+ * First, check if CLUSTER CONCURRENTLY is being performed by this
+ * backend. If so, only decode data changes of the table that it is
+ * processing, and the changes of its TOAST relation.
*
* (TOAST locator should not be set unless the main is.)
*/
@@ -491,6 +500,60 @@ heap_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
return;
}
+ /*
+ * Second, skip records which do not contain sufficient information for
+ * the decoding.
+ *
+ * The backend executing CLUSTER CONCURRENTLY should not return here
+ * because the records which passed the checks above should contain be
+ * eligible for decoding. However, CLUSTER CONCURRENTLY generates WAL when
+ * writing data into the new table, which should not be decoded by the
+ * other backends. This is where the other backends skip them.
+ */
+ switch (info)
+ {
+ case XLOG_HEAP_INSERT:
+ {
+ xl_heap_insert *rec;
+
+ rec = (xl_heap_insert *) XLogRecGetData(buf->record);
+ /*
+ * (Besides insertion into the main heap by CLUSTER CONCURRENTLY,
+ * this does happen when raw_heap_insert marks the TOAST record as
+ * HEAP_INSERT_NO_LOGICAL).
+ */
+ if ((rec->flags & XLH_INSERT_CONTAINS_NEW_TUPLE) == 0)
+ return;
+
+ break;
+ }
+
+ case XLOG_HEAP_HOT_UPDATE:
+ case XLOG_HEAP_UPDATE:
+ {
+ xl_heap_update *rec;
+
+ rec = (xl_heap_update *) XLogRecGetData(buf->record);
+ if ((rec->flags &
+ (XLH_UPDATE_CONTAINS_NEW_TUPLE |
+ XLH_UPDATE_CONTAINS_OLD_TUPLE |
+ XLH_UPDATE_CONTAINS_OLD_KEY)) == 0)
+ return;
+
+ break;
+ }
+
+ case XLOG_HEAP_DELETE:
+ {
+ xl_heap_delete *rec;
+
+ rec = (xl_heap_delete *) XLogRecGetData(buf->record);
+ if (rec->flags & XLH_DELETE_NO_LOGICAL)
+ return;
+ break;
+ }
+ }
+
ReorderBufferProcessXid(ctx->reorder, xid, buf->origptr);
/*
@@ -923,13 +986,6 @@ DecodeInsert(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
xlrec = (xl_heap_insert *) XLogRecGetData(r);
- /*
- * Ignore insert records without new tuples (this does happen when
- * raw_heap_insert marks the TOAST record as HEAP_INSERT_NO_LOGICAL).
- */
- if (!(xlrec->flags & XLH_INSERT_CONTAINS_NEW_TUPLE))
- return;
-
/* only interested in our database */
XLogRecGetBlockTag(r, 0, &target_locator, NULL, NULL);
if (target_locator.dbOid != ctx->slot->data.database)
diff --git a/src/backend/replication/logical/snapbuild.c b/src/backend/replication/logical/snapbuild.c
index 86a9b0335a..2d4ce8b37f 100644
--- a/src/backend/replication/logical/snapbuild.c
+++ b/src/backend/replication/logical/snapbuild.c
@@ -155,7 +155,7 @@ static bool ExportInProgress = false;
static void SnapBuildPurgeOlderTxn(SnapBuild *builder);
/* snapshot building/manipulation/distribution functions */
-static Snapshot SnapBuildBuildSnapshot(SnapBuild *builder);
+static Snapshot SnapBuildBuildSnapshot(SnapBuild *builder, XLogRecPtr lsn);
static void SnapBuildFreeSnapshot(Snapshot snap);
@@ -352,12 +352,17 @@ SnapBuildSnapDecRefcount(Snapshot snap)
* Build a new snapshot, based on currently committed catalog-modifying
* transactions.
*
+ * 'lsn' is the location of the commit record (of a catalog-changing
+ * transaction) that triggered creation of the snapshot. Pass
+ * InvalidXLogRecPtr for the transaction base snapshot or if it the user of
+ * the snapshot should not need the LSN.
+ *
* In-progress transactions with catalog access are *not* allowed to modify
* these snapshots; they have to copy them and fill in appropriate ->curcid
* and ->subxip/subxcnt values.
*/
static Snapshot
-SnapBuildBuildSnapshot(SnapBuild *builder)
+SnapBuildBuildSnapshot(SnapBuild *builder, XLogRecPtr lsn)
{
Snapshot snapshot;
Size ssize;
@@ -425,6 +430,7 @@ SnapBuildBuildSnapshot(SnapBuild *builder)
snapshot->active_count = 0;
snapshot->regd_count = 0;
snapshot->snapXactCompletionCount = 0;
+ snapshot->lsn = lsn;
return snapshot;
}
@@ -461,7 +467,7 @@ SnapBuildInitialSnapshot(SnapBuild *builder)
if (TransactionIdIsValid(MyProc->xmin))
elog(ERROR, "cannot build an initial slot snapshot when MyProc->xmin already is valid");
- snap = SnapBuildBuildSnapshot(builder);
+ snap = SnapBuildBuildSnapshot(builder, InvalidXLogRecPtr);
/*
* We know that snap->xmin is alive, enforced by the logical xmin
@@ -502,7 +508,7 @@ SnapBuildInitialSnapshotForCluster(SnapBuild *builder)
Assert(builder->state == SNAPBUILD_CONSISTENT);
- snap = SnapBuildBuildSnapshot(builder);
+ snap = SnapBuildBuildSnapshot(builder, InvalidXLogRecPtr);
return SnapBuildMVCCFromHistoric(snap, false);
}
@@ -636,7 +642,7 @@ SnapBuildGetOrBuildSnapshot(SnapBuild *builder)
/* only build a new snapshot if we don't have a prebuilt one */
if (builder->snapshot == NULL)
{
- builder->snapshot = SnapBuildBuildSnapshot(builder);
+ builder->snapshot = SnapBuildBuildSnapshot(builder, InvalidXLogRecPtr);
/* increase refcount for the snapshot builder */
SnapBuildSnapIncRefcount(builder->snapshot);
}
@@ -716,7 +722,7 @@ SnapBuildProcessChange(SnapBuild *builder, TransactionId xid, XLogRecPtr lsn)
/* only build a new snapshot if we don't have a prebuilt one */
if (builder->snapshot == NULL)
{
- builder->snapshot = SnapBuildBuildSnapshot(builder);
+ builder->snapshot = SnapBuildBuildSnapshot(builder, lsn);
/* increase refcount for the snapshot builder */
SnapBuildSnapIncRefcount(builder->snapshot);
}
@@ -1085,7 +1091,7 @@ SnapBuildCommitTxn(SnapBuild *builder, XLogRecPtr lsn, TransactionId xid,
if (builder->snapshot)
SnapBuildSnapDecRefcount(builder->snapshot);
- builder->snapshot = SnapBuildBuildSnapshot(builder);
+ builder->snapshot = SnapBuildBuildSnapshot(builder, lsn);
/* we might need to execute invalidations, add snapshot */
if (!ReorderBufferXidHasBaseSnapshot(builder->reorder, xid))
@@ -1910,7 +1916,7 @@ SnapBuildRestore(SnapBuild *builder, XLogRecPtr lsn)
{
SnapBuildSnapDecRefcount(builder->snapshot);
}
- builder->snapshot = SnapBuildBuildSnapshot(builder);
+ builder->snapshot = SnapBuildBuildSnapshot(builder, InvalidXLogRecPtr);
SnapBuildSnapIncRefcount(builder->snapshot);
ReorderBufferSetRestartPoint(builder->reorder, lsn);
diff --git a/src/backend/replication/pgoutput_cluster/pgoutput_cluster.c b/src/backend/replication/pgoutput_cluster/pgoutput_cluster.c
index 43f7b34297..8e915c55fb 100644
--- a/src/backend/replication/pgoutput_cluster/pgoutput_cluster.c
+++ b/src/backend/replication/pgoutput_cluster/pgoutput_cluster.c
@@ -33,7 +33,8 @@ static void plugin_truncate(struct LogicalDecodingContext *ctx,
Relation relations[],
ReorderBufferChange *change);
static void store_change(LogicalDecodingContext *ctx,
- ConcurrentChangeKind kind, HeapTuple tuple);
+ ConcurrentChangeKind kind, HeapTuple tuple,
+ TransactionId xid);
void
_PG_output_plugin_init(OutputPluginCallbacks *cb)
@@ -101,6 +102,7 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
Relation relation, ReorderBufferChange *change)
{
ClusterDecodingState *dstate;
+ Snapshot snapshot;
dstate = (ClusterDecodingState *) ctx->output_writer_private;
@@ -108,6 +110,48 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
if (relation->rd_id != dstate->relid)
return;
+ /*
+ * Catalog snapshot is fine because the table we are processing is
+ * temporarily considered a user catalog table.
+ */
+ snapshot = GetCatalogSnapshot(InvalidOid);
+ Assert(snapshot->snapshot_type == SNAPSHOT_HISTORIC_MVCC);
+ Assert(!snapshot->suboverflowed);
+
+ /*
+ * This should not happen, but if we don't have enough information to
+ * apply a new snapshot, the consequences would be bad. Thus prefer ERROR
+ * to Assert().
+ */
+ if (XLogRecPtrIsInvalid(snapshot->lsn))
+ ereport(ERROR, (errmsg("snapshot has invalid LSN")));
+
+ /*
+ * reorderbuffer.c changes the catalog snapshot as soon as it sees a new
+ * CID or a commit record of a catalog-changing transaction.
+ */
+ if (dstate->snapshot == NULL || snapshot->lsn != dstate->snapshot_lsn ||
+ snapshot->curcid != dstate->snapshot->curcid)
+ {
+ /* CID should not go backwards. */
+ Assert(dstate->snapshot == NULL ||
+ snapshot->curcid >= dstate->snapshot->curcid ||
+ change->txn->xid != dstate->last_change_xid);
+
+ /*
+ * XXX Is it a problem that the copy is created in
+ * TopTransactionContext?
+ *
+ * XXX Wouldn't it be o.k. for SnapBuildMVCCFromHistoric() to set xcnt
+ * to 0 instead of converting xip in this case? The point is that
+ * transactions which are still in progress from the perspective of
+ * reorderbuffer.c could not be replayed yet, so we do not need to
+ * examine their XIDs.
+ */
+ dstate->snapshot = SnapBuildMVCCFromHistoric(snapshot, false);
+ dstate->snapshot_lsn = snapshot->lsn;
+ }
+
/* Decode entry depending on its type */
switch (change->action)
{
@@ -125,7 +169,7 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
if (newtuple == NULL)
elog(ERROR, "Incomplete insert info.");
- store_change(ctx, CHANGE_INSERT, newtuple);
+ store_change(ctx, CHANGE_INSERT, newtuple, change->txn->xid);
}
break;
case REORDER_BUFFER_CHANGE_UPDATE:
@@ -142,9 +186,11 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
elog(ERROR, "Incomplete update info.");
if (oldtuple != NULL)
- store_change(ctx, CHANGE_UPDATE_OLD, oldtuple);
+ store_change(ctx, CHANGE_UPDATE_OLD, oldtuple,
+ change->txn->xid);
- store_change(ctx, CHANGE_UPDATE_NEW, newtuple);
+ store_change(ctx, CHANGE_UPDATE_NEW, newtuple,
+ change->txn->xid);
}
break;
case REORDER_BUFFER_CHANGE_DELETE:
@@ -157,7 +203,7 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
if (oldtuple == NULL)
elog(ERROR, "Incomplete delete info.");
- store_change(ctx, CHANGE_DELETE, oldtuple);
+ store_change(ctx, CHANGE_DELETE, oldtuple, change->txn->xid);
}
break;
default:
@@ -191,13 +237,13 @@ plugin_truncate(struct LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
if (i == nrelations)
return;
- store_change(ctx, CHANGE_TRUNCATE, NULL);
+ store_change(ctx, CHANGE_TRUNCATE, NULL, InvalidTransactionId);
}
/* Store concurrent data change. */
static void
store_change(LogicalDecodingContext *ctx, ConcurrentChangeKind kind,
- HeapTuple tuple)
+ HeapTuple tuple, TransactionId xid)
{
ClusterDecodingState *dstate;
char *change_raw;
@@ -265,6 +311,11 @@ store_change(LogicalDecodingContext *ctx, ConcurrentChangeKind kind,
dst = dst_start + SizeOfConcurrentChange;
memcpy(dst, tuple->t_data, tuple->t_len);
+ /* Initialize the other fields. */
+ change.xid = xid;
+ change.snapshot = dstate->snapshot;
+ dstate->snapshot->active_count++;
+
/* The data has been copied. */
if (flattened)
pfree(tuple);
@@ -278,6 +329,9 @@ store:
isnull[0] = false;
tuplestore_putvalues(dstate->tstore, dstate->tupdesc_change,
values, isnull);
+#ifdef USE_ASSERT_CHECKING
+ dstate->last_change_xid = xid;
+#endif
/* Accounting. */
dstate->nchanges++;
diff --git a/src/include/access/heapam.h b/src/include/access/heapam.h
index 9b1fb15d8c..5a6444e969 100644
--- a/src/include/access/heapam.h
+++ b/src/include/access/heapam.h
@@ -317,21 +317,24 @@ extern BulkInsertState GetBulkInsertState(void);
extern void FreeBulkInsertState(BulkInsertState);
extern void ReleaseBulkInsertStatePin(BulkInsertState bistate);
-extern void heap_insert(Relation relation, HeapTuple tup, CommandId cid,
- int options, BulkInsertState bistate);
+extern void heap_insert(Relation relation, HeapTuple tup, TransactionId xid,
+ CommandId cid, int options, BulkInsertState bistate);
extern void heap_multi_insert(Relation relation, struct TupleTableSlot **slots,
int ntuples, CommandId cid, int options,
BulkInsertState bistate);
extern TM_Result heap_delete(Relation relation, ItemPointer tid,
- CommandId cid, Snapshot crosscheck, bool wait,
- struct TM_FailureData *tmfd, bool changingPart);
+ TransactionId xid, CommandId cid,
+ Snapshot crosscheck, bool wait,
+ struct TM_FailureData *tmfd, bool changingPart,
+ bool wal_logical);
extern void heap_finish_speculative(Relation relation, ItemPointer tid);
extern void heap_abort_speculative(Relation relation, ItemPointer tid);
extern TM_Result heap_update(Relation relation, ItemPointer otid,
- HeapTuple newtup,
+ HeapTuple newtup, TransactionId xid,
CommandId cid, Snapshot crosscheck, bool wait,
struct TM_FailureData *tmfd, LockTupleMode *lockmode,
- TU_UpdateIndexes *update_indexes);
+ TU_UpdateIndexes *update_indexes,
+ bool wal_logical);
extern TM_Result heap_lock_tuple(Relation relation, HeapTuple tuple,
CommandId cid, LockTupleMode mode, LockWaitPolicy wait_policy,
bool follow_updates,
diff --git a/src/include/access/heapam_xlog.h b/src/include/access/heapam_xlog.h
index 277df6b3cf..8d4af07f84 100644
--- a/src/include/access/heapam_xlog.h
+++ b/src/include/access/heapam_xlog.h
@@ -104,6 +104,8 @@
#define XLH_DELETE_CONTAINS_OLD_KEY (1<<2)
#define XLH_DELETE_IS_SUPER (1<<3)
#define XLH_DELETE_IS_PARTITION_MOVE (1<<4)
+/* See heap_delete() */
+#define XLH_DELETE_NO_LOGICAL (1<<5)
/* convenience macro for checking whether any form of old tuple was logged */
#define XLH_DELETE_CONTAINS_OLD \
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index b2bc10ee04..1a0b173d48 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -482,6 +482,8 @@ extern Size EstimateTransactionStateSpace(void);
extern void SerializeTransactionState(Size maxsize, char *start_address);
extern void StartParallelWorkerTransaction(char *tstatespace);
extern void EndParallelWorkerTransaction(void);
+extern void SetClusterCurrentXids(TransactionId *xip, int xcnt);
+extern void ResetClusterCurrentXids(void);
extern bool IsTransactionBlock(void);
extern bool IsTransactionOrTransactionBlock(void);
extern char TransactionBlockStatusCode(void);
diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h
index d420930d6b..8945e46e64 100644
--- a/src/include/commands/cluster.h
+++ b/src/include/commands/cluster.h
@@ -58,6 +58,14 @@ typedef struct ConcurrentChange
/* See the enum above. */
ConcurrentChangeKind kind;
+ /* Transaction that changes the data. */
+ TransactionId xid;
+
+ /*
+ * Historic catalog snapshot that was used to decode this change.
+ */
+ Snapshot snapshot;
+
/*
* The actual tuple.
*
@@ -89,6 +97,8 @@ typedef struct ClusterDecodingState
* tuplestore does this transparently.
*/
Tuplestorestate *tstore;
+ /* XID of the last change added to tstore. */
+ TransactionId last_change_xid PG_USED_FOR_ASSERTS_ONLY;
/* The current number of changes in tstore. */
double nchanges;
@@ -109,6 +119,14 @@ typedef struct ClusterDecodingState
/* Slot to retrieve data from tstore. */
TupleTableSlot *tsslot;
+ /*
+ * Historic catalog snapshot that was used to decode the most recent
+ * change.
+ */
+ Snapshot snapshot;
+ /* LSN of the record */
+ XLogRecPtr snapshot_lsn;
+
ResourceOwner resowner;
} ClusterDecodingState;
diff --git a/src/include/utils/snapshot.h b/src/include/utils/snapshot.h
index 0e546ec149..014f27db7d 100644
--- a/src/include/utils/snapshot.h
+++ b/src/include/utils/snapshot.h
@@ -13,6 +13,7 @@
#ifndef SNAPSHOT_H
#define SNAPSHOT_H
+#include "access/xlogdefs.h"
#include "lib/pairingheap.h"
@@ -201,6 +202,8 @@ typedef struct SnapshotData
uint32 regd_count; /* refcount on RegisteredSnapshots */
pairingheap_node ph_node; /* link in the RegisteredSnapshots heap */
+ XLogRecPtr lsn; /* position in the WAL stream when taken */
+
/*
* The transaction completion count at the time GetSnapshotData() built
* this snapshot. Allows to avoid re-computing static snapshots when no
--
2.45.2
[text/x-diff] v07-0006-Add-regression-tests.patch (10.4K, ../6250.1736776111@antos/6-v07-0006-Add-regression-tests.patch)
download | inline diff:
From 4d1b029b59a33973cdcd475ed4e4748375557843 Mon Sep 17 00:00:00 2001
From: Antonin Houska <[email protected]>
Date: Mon, 13 Jan 2025 14:29:54 +0100
Subject: [PATCH 6/8] Add regression tests.
As this patch series adds the CONCURRENTLY option to the VACUUM FULL and
CLUSTER commands, it's appropriate to test that the "concurrent data changes"
(i.e. changes done by application while we are copying the table contents to
the new storage) are processed correctly.
Injection points are used to stop the data copying at some point. While the
backend in charge of the copying is waiting on the injection point, another
backend runs some INSERT, UPDATE and DELETE commands on the table. Then we
wake up the first backend and let the CLUSTER (CONCURRENTLY) command
finish. Finally we check that all the "concurrent data changes" are present in
the table and that they contain the correct visibility information.
---
src/backend/commands/cluster.c | 7 +
src/test/modules/injection_points/Makefile | 3 +-
.../injection_points/expected/cluster.out | 113 ++++++++++++++
.../modules/injection_points/logical.conf | 1 +
src/test/modules/injection_points/meson.build | 3 +
.../injection_points/specs/cluster.spec | 140 ++++++++++++++++++
6 files changed, 266 insertions(+), 1 deletion(-)
create mode 100644 src/test/modules/injection_points/expected/cluster.out
create mode 100644 src/test/modules/injection_points/logical.conf
create mode 100644 src/test/modules/injection_points/specs/cluster.spec
diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index 3a1a51a56a..8e73da73ee 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -58,6 +58,7 @@
#include "utils/acl.h"
#include "utils/fmgroids.h"
#include "utils/guc.h"
+#include "utils/injection_point.h"
#include "utils/inval.h"
#include "utils/lsyscache.h"
#include "utils/memutils.h"
@@ -3721,6 +3722,12 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
*/
ident_key = build_identity_key(ident_idx_new, OldHeap, &ident_key_nentries);
+ /*
+ * During testing, wait for another backend to perform concurrent data
+ * changes which we will process below.
+ */
+ INJECTION_POINT("cluster-concurrently-before-lock");
+
/*
* Flush all WAL records inserted so far (possibly except for the last
* incomplete page, see GetInsertRecPtr), to minimize the amount of data
diff --git a/src/test/modules/injection_points/Makefile b/src/test/modules/injection_points/Makefile
index 0753a9df58..e40ebec1bb 100644
--- a/src/test/modules/injection_points/Makefile
+++ b/src/test/modules/injection_points/Makefile
@@ -13,7 +13,8 @@ PGFILEDESC = "injection_points - facility for injection points"
REGRESS = injection_points reindex_conc
REGRESS_OPTS = --dlpath=$(top_builddir)/src/test/regress
-ISOLATION = basic inplace
+ISOLATION = basic inplace cluster
+ISOLATION_OPTS = --temp-config $(top_srcdir)/src/test/modules/injection_points/logical.conf
TAP_TESTS = 1
diff --git a/src/test/modules/injection_points/expected/cluster.out b/src/test/modules/injection_points/expected/cluster.out
new file mode 100644
index 0000000000..d84fff3693
--- /dev/null
+++ b/src/test/modules/injection_points/expected/cluster.out
@@ -0,0 +1,113 @@
+Parsed test spec with 2 sessions
+
+starting permutation: wait_before_lock change_existing change_new change_subxact1 change_subxact2 check2 wakeup_before_lock check1
+injection_points_attach
+-----------------------
+
+(1 row)
+
+step wait_before_lock:
+ CLUSTER (CONCURRENTLY) clstr_test USING clstr_test_pkey;
+ <waiting ...>
+step change_existing:
+ UPDATE clstr_test SET i=10 where i=1;
+ UPDATE clstr_test SET j=20 where i=2;
+ UPDATE clstr_test SET i=30 where i=3;
+ UPDATE clstr_test SET i=40 where i=30;
+ DELETE FROM clstr_test WHERE i=4;
+
+step change_new:
+ INSERT INTO clstr_test(i, j) VALUES (5, 5), (6, 6), (7, 7), (8, 8);
+ UPDATE clstr_test SET i=50 where i=5;
+ UPDATE clstr_test SET j=60 where i=6;
+ DELETE FROM clstr_test WHERE i=7;
+
+step change_subxact1:
+ BEGIN;
+ INSERT INTO clstr_test(i, j) VALUES (100, 100);
+ SAVEPOINT s1;
+ UPDATE clstr_test SET i=101 where i=100;
+ SAVEPOINT s2;
+ UPDATE clstr_test SET i=102 where i=101;
+ COMMIT;
+
+step change_subxact2:
+ BEGIN;
+ SAVEPOINT s1;
+ INSERT INTO clstr_test(i, j) VALUES (110, 110);
+ ROLLBACK TO SAVEPOINT s1;
+ INSERT INTO clstr_test(i, j) VALUES (110, 111);
+ COMMIT;
+
+step check2:
+ INSERT INTO relfilenodes(node)
+ SELECT relfilenode FROM pg_class WHERE relname='clstr_test';
+
+ SELECT i, j FROM clstr_test ORDER BY i, j;
+
+ INSERT INTO data_s2(_xmin, _cmin, i, j)
+ SELECT xmin, cmin, i, j FROM clstr_test;
+
+ i| j
+---+---
+ 2| 20
+ 6| 60
+ 8| 8
+ 10| 1
+ 40| 3
+ 50| 5
+102|100
+110|111
+(8 rows)
+
+step wakeup_before_lock:
+ SELECT injection_points_wakeup('cluster-concurrently-before-lock');
+
+injection_points_wakeup
+-----------------------
+
+(1 row)
+
+step wait_before_lock: <... completed>
+step check1:
+ INSERT INTO relfilenodes(node)
+ SELECT relfilenode FROM pg_class WHERE relname='clstr_test';
+
+ SELECT count(DISTINCT node) FROM relfilenodes;
+
+ SELECT i, j FROM clstr_test ORDER BY i, j;
+
+ INSERT INTO data_s1(_xmin, _cmin, i, j)
+ SELECT xmin, cmin, i, j FROM clstr_test;
+
+ SELECT count(*)
+ FROM data_s1 d1 FULL JOIN data_s2 d2 USING (_xmin, _cmin, i, j)
+ WHERE d1.i ISNULL OR d2.i ISNULL;
+
+count
+-----
+ 2
+(1 row)
+
+ i| j
+---+---
+ 2| 20
+ 6| 60
+ 8| 8
+ 10| 1
+ 40| 3
+ 50| 5
+102|100
+110|111
+(8 rows)
+
+count
+-----
+ 0
+(1 row)
+
+injection_points_detach
+-----------------------
+
+(1 row)
+
diff --git a/src/test/modules/injection_points/logical.conf b/src/test/modules/injection_points/logical.conf
new file mode 100644
index 0000000000..c8f264bc6c
--- /dev/null
+++ b/src/test/modules/injection_points/logical.conf
@@ -0,0 +1 @@
+wal_level = logical
\ No newline at end of file
diff --git a/src/test/modules/injection_points/meson.build b/src/test/modules/injection_points/meson.build
index 989b4db226..8c404ddd61 100644
--- a/src/test/modules/injection_points/meson.build
+++ b/src/test/modules/injection_points/meson.build
@@ -44,7 +44,10 @@ tests += {
'specs': [
'basic',
'inplace',
+ 'cluster',
],
+ # 'cluster' requires wal_level = 'logical'.
+ 'regress_args': ['--temp-config', files('logical.conf')],
},
'tap': {
'env': {
diff --git a/src/test/modules/injection_points/specs/cluster.spec b/src/test/modules/injection_points/specs/cluster.spec
new file mode 100644
index 0000000000..5f8404c5da
--- /dev/null
+++ b/src/test/modules/injection_points/specs/cluster.spec
@@ -0,0 +1,140 @@
+# Prefix the system columns with underscore as they are not allowed as column
+# names.
+setup
+{
+ CREATE EXTENSION injection_points;
+
+ CREATE TABLE clstr_test(i int PRIMARY KEY, j int);
+ INSERT INTO clstr_test(i, j) VALUES (1, 1), (2, 2), (3, 3), (4, 4);
+
+ CREATE TABLE relfilenodes(node oid);
+
+ CREATE TABLE data_s1(_xmin xid, _cmin cid, i int, j int);
+ CREATE TABLE data_s2(_xmin xid, _cmin cid, i int, j int);
+}
+
+teardown
+{
+ DROP TABLE clstr_test;
+ DROP EXTENSION injection_points;
+
+ DROP TABLE relfilenodes;
+ DROP TABLE data_s1;
+ DROP TABLE data_s2;
+}
+
+session s1
+setup
+{
+ SELECT injection_points_set_local();
+ SELECT injection_points_attach('cluster-concurrently-before-lock', 'wait');
+}
+# Perform the initial load and wait for s2 to do some data changes.
+step wait_before_lock
+{
+ CLUSTER (CONCURRENTLY) clstr_test USING clstr_test_pkey;
+}
+# Check the table from the perspective of s1.
+#
+# Besides the contents, we also check that relfilenode has changed.
+#
+# xmin and cmin columns are used to check that we do not change tuple
+# visibility information. Since we do not expect xmin to stay unchanged across
+# test runs, it cannot appear in the output text. Instead, have each session
+# write the contents into a table and use FULL JOIN to check if the outputs
+# are identical.
+step check1
+{
+ INSERT INTO relfilenodes(node)
+ SELECT relfilenode FROM pg_class WHERE relname='clstr_test';
+
+ SELECT count(DISTINCT node) FROM relfilenodes;
+
+ SELECT i, j FROM clstr_test ORDER BY i, j;
+
+ INSERT INTO data_s1(_xmin, _cmin, i, j)
+ SELECT xmin, cmin, i, j FROM clstr_test;
+
+ SELECT count(*)
+ FROM data_s1 d1 FULL JOIN data_s2 d2 USING (_xmin, _cmin, i, j)
+ WHERE d1.i ISNULL OR d2.i ISNULL;
+}
+teardown
+{
+ SELECT injection_points_detach('cluster-concurrently-before-lock');
+}
+
+session s2
+# Change the existing data. UPDATE changes both key and non-key columns. Also
+# update one row twice to test whether tuple version generated by this session
+# can be found.
+step change_existing
+{
+ UPDATE clstr_test SET i=10 where i=1;
+ UPDATE clstr_test SET j=20 where i=2;
+ UPDATE clstr_test SET i=30 where i=3;
+ UPDATE clstr_test SET i=40 where i=30;
+ DELETE FROM clstr_test WHERE i=4;
+}
+# Insert new rows and UPDATE / DELETE some of them. Again, update both key and
+# non-key column.
+step change_new
+{
+ INSERT INTO clstr_test(i, j) VALUES (5, 5), (6, 6), (7, 7), (8, 8);
+ UPDATE clstr_test SET i=50 where i=5;
+ UPDATE clstr_test SET j=60 where i=6;
+ DELETE FROM clstr_test WHERE i=7;
+}
+
+# When applying concurrent data changes, we should see the effects of an
+# in-progress subtransaction.
+step change_subxact1
+{
+ BEGIN;
+ INSERT INTO clstr_test(i, j) VALUES (100, 100);
+ SAVEPOINT s1;
+ UPDATE clstr_test SET i=101 where i=100;
+ SAVEPOINT s2;
+ UPDATE clstr_test SET i=102 where i=101;
+ COMMIT;
+}
+
+# When applying concurrent data changes, we should not see the effects of a
+# rolled back subtransaction.
+step change_subxact2
+{
+ BEGIN;
+ SAVEPOINT s1;
+ INSERT INTO clstr_test(i, j) VALUES (110, 110);
+ ROLLBACK TO SAVEPOINT s1;
+ INSERT INTO clstr_test(i, j) VALUES (110, 111);
+ COMMIT;
+}
+
+# Check the table from the perspective of s2.
+step check2
+{
+ INSERT INTO relfilenodes(node)
+ SELECT relfilenode FROM pg_class WHERE relname='clstr_test';
+
+ SELECT i, j FROM clstr_test ORDER BY i, j;
+
+ INSERT INTO data_s2(_xmin, _cmin, i, j)
+ SELECT xmin, cmin, i, j FROM clstr_test;
+}
+step wakeup_before_lock
+{
+ SELECT injection_points_wakeup('cluster-concurrently-before-lock');
+}
+
+# Test if data changes introduced while one session is performing CLUSTER
+# (CONCURRENTLY) find their way into the table.
+permutation
+ wait_before_lock
+ change_existing
+ change_new
+ change_subxact1
+ change_subxact2
+ check2
+ wakeup_before_lock
+ check1
--
2.45.2
[text/x-diff] v07-0007-Introduce-cluster_max_xlock_time-configuration-varia.patch (20.4K, ../6250.1736776111@antos/7-v07-0007-Introduce-cluster_max_xlock_time-configuration-varia.patch)
download | inline diff:
From 6c314f3e135938def75aa8b9efde86f782503d10 Mon Sep 17 00:00:00 2001
From: Antonin Houska <[email protected]>
Date: Mon, 13 Jan 2025 14:29:54 +0100
Subject: [PATCH 7/8] Introduce cluster_max_xlock_time configuration variable.
When executing VACUUM FULL / CLUSTER (CONCURRENTLY) we need the
AccessExclusiveLock to swap the relation files and that should require pretty
short time. However, on a busy system, other backends might change
non-negligible amount of data in the table while we are waiting for the
lock. Since these changes must be applied to the new storage before the swap,
the time we eventually hold the lock might become non-negligible too.
If the user is worried about this situation, he can set cluster_max_xlock_time
to the maximum time for which the exclusive lock may be held. If this amount
of time is not sufficient to complete the VACUUM FULL / CLUSTER (CONCURRENTLY)
command, ERROR is raised and the command is canceled.
---
doc/src/sgml/config.sgml | 32 +++++
doc/src/sgml/ref/cluster.sgml | 9 +-
src/backend/access/heap/heapam_handler.c | 3 +-
src/backend/commands/cluster.c | 133 +++++++++++++++---
src/backend/utils/misc/guc_tables.c | 14 ++
src/backend/utils/misc/postgresql.conf.sample | 1 +
src/include/commands/cluster.h | 5 +-
.../injection_points/expected/cluster.out | 74 +++++++++-
.../injection_points/specs/cluster.spec | 42 ++++++
9 files changed, 293 insertions(+), 20 deletions(-)
diff --git a/doc/src/sgml/config.sgml b/doc/src/sgml/config.sgml
index 3f41a17b1f..695d1fe2a4 100644
--- a/doc/src/sgml/config.sgml
+++ b/doc/src/sgml/config.sgml
@@ -10701,6 +10701,38 @@ dynamic_library_path = 'C:\tools\postgresql;H:\my_project\lib;$libdir'
</listitem>
</varlistentry>
+ <varlistentry id="guc-cluster-max-xclock-time" xreflabel="cluster_max_xlock_time">
+ <term><varname>cluster_max_xlock_time</varname> (<type>integer</type>)
+ <indexterm>
+ <primary><varname>cluster_max_xlock_time</varname> configuration parameter</primary>
+ </indexterm>
+ </term>
+ <listitem>
+ <para>
+ This is the maximum amount of time to hold an exclusive lock on a
+ table by commands <command>CLUSTER</command> and <command>VACUUM
+ FULL</command> with the <literal>CONCURRENTLY</literal>
+ option. Typically, these commands should not need the lock for longer
+ time than <command>TRUNCATE</command> does. However, additional time
+ might be needed if the system is too busy. (See
+ <xref linkend="sql-cluster"/> for explanation how
+ the <literal>CONCURRENTLY</literal> option works.)
+ </para>
+
+ <para>
+ If you want to restrict the lock time, set this variable to the
+ highest acceptable value. If it appears during the processing that
+ additional time is needed to release the lock, the command will be
+ cancelled.
+ </para>
+
+ <para>
+ The default value is 0, which means that the lock is not released
+ until the concurrent data changes are processed.
+ </para>
+ </listitem>
+ </varlistentry>
+
</variablelist>
</sect1>
diff --git a/doc/src/sgml/ref/cluster.sgml b/doc/src/sgml/ref/cluster.sgml
index 356b40e3fe..ebb85d9d47 100644
--- a/doc/src/sgml/ref/cluster.sgml
+++ b/doc/src/sgml/ref/cluster.sgml
@@ -141,7 +141,14 @@ CLUSTER [ ( <replaceable class="parameter">option</replaceable> [, ...] ) ] [ <r
(<xref linkend="logicaldecoding"/>) and applied before
the <literal>ACCESS EXCLUSIVE</literal> lock is requested. Thus the lock
is typically held only for the time needed to swap the files, which
- should be pretty short.
+ should be pretty short. However, the time might still be noticeable if
+ too many data changes have been done to the table while
+ <command>CLUSTER</command> was waiting for the lock: those changes must
+ be processed just before the files are swapped, while the
+ <literal>ACCESS EXCLUSIVE</literal> lock is being held. If you are
+ worried about this situation, set
+ the <link linkend="guc-cluster-max-xclock-time"><varname>cluster_max_xlock_time</varname></link>
+ configuration parameter to a value that your applications can tolerate.
</para>
<para>
diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c
index c315abac02..13c67f0e5f 100644
--- a/src/backend/access/heap/heapam_handler.c
+++ b/src/backend/access/heap/heapam_handler.c
@@ -1004,7 +1004,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap,
end_of_wal = GetFlushRecPtr(NULL);
if ((end_of_wal - end_of_wal_prev) > wal_segment_size)
{
- cluster_decode_concurrent_changes(decoding_ctx, end_of_wal);
+ cluster_decode_concurrent_changes(decoding_ctx, end_of_wal,
+ NULL);
end_of_wal_prev = end_of_wal;
}
}
diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index 8e73da73ee..dfae550123 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -17,6 +17,8 @@
*/
#include "postgres.h"
+#include <sys/time.h>
+
#include "access/amapi.h"
#include "access/heapam.h"
#include "access/multixact.h"
@@ -102,6 +104,15 @@ RelFileLocator clustered_rel_toast_locator = {.relNumber = InvalidOid};
#define CLUSTER_IN_PROGRESS_MESSAGE \
"relation \"%s\" is already being processed by CLUSTER CONCURRENTLY"
+/*
+ * The maximum time to hold AccessExclusiveLock during the final
+ * processing. Note that only the execution time of
+ * process_concurrent_changes() is included here. The very last steps like
+ * swap_relation_files() shouldn't get blocked and it'd be wrong to consider
+ * them a reason to abort otherwise completed processing.
+ */
+int cluster_max_xlock_time = 0;
+
/*
* Everything we need to call ExecInsertIndexTuples().
*/
@@ -188,7 +199,8 @@ static LogicalDecodingContext *setup_logical_decoding(Oid relid,
static HeapTuple get_changed_tuple(char *change);
static void apply_concurrent_changes(ClusterDecodingState *dstate,
Relation rel, ScanKey key, int nkeys,
- IndexInsertState *iistate);
+ IndexInsertState *iistate,
+ struct timeval *must_complete);
static void apply_concurrent_insert(Relation rel, ConcurrentChange *change,
HeapTuple tup, IndexInsertState *iistate,
TupleTableSlot *index_slot);
@@ -205,13 +217,15 @@ static HeapTuple find_target_tuple(Relation rel, ScanKey key, int nkeys,
IndexInsertState *iistate,
TupleTableSlot *ident_slot,
IndexScanDesc *scan_p);
-static void process_concurrent_changes(LogicalDecodingContext *ctx,
+static bool process_concurrent_changes(LogicalDecodingContext *ctx,
XLogRecPtr end_of_wal,
Relation rel_dst,
Relation rel_src,
ScanKey ident_key,
int ident_key_nentries,
- IndexInsertState *iistate);
+ IndexInsertState *iistate,
+ struct timeval *must_complete);
+static bool processing_time_elapsed(struct timeval *must_complete);
static IndexInsertState *get_index_insert_state(Relation relation,
Oid ident_index_id);
static ScanKey build_identity_key(Oid ident_idx_oid, Relation rel_src,
@@ -3027,7 +3041,8 @@ get_changed_tuple(char *change)
*/
void
cluster_decode_concurrent_changes(LogicalDecodingContext *ctx,
- XLogRecPtr end_of_wal)
+ XLogRecPtr end_of_wal,
+ struct timeval *must_complete)
{
ClusterDecodingState *dstate;
ResourceOwner resowner_old;
@@ -3065,6 +3080,9 @@ cluster_decode_concurrent_changes(LogicalDecodingContext *ctx,
if (record != NULL)
LogicalDecodingProcessRecord(ctx, ctx->reader);
+ if (processing_time_elapsed(must_complete))
+ break;
+
/*
* If WAL segment boundary has been crossed, inform the decoding
* system that the catalog_xmin can advance. (We can confirm more
@@ -3107,7 +3125,8 @@ cluster_decode_concurrent_changes(LogicalDecodingContext *ctx,
*/
static void
apply_concurrent_changes(ClusterDecodingState *dstate, Relation rel,
- ScanKey key, int nkeys, IndexInsertState *iistate)
+ ScanKey key, int nkeys, IndexInsertState *iistate,
+ struct timeval *must_complete)
{
TupleTableSlot *index_slot, *ident_slot;
HeapTuple tup_old = NULL;
@@ -3137,6 +3156,9 @@ apply_concurrent_changes(ClusterDecodingState *dstate, Relation rel,
CHECK_FOR_INTERRUPTS();
+ Assert(dstate->nchanges > 0);
+ dstate->nchanges--;
+
/* Get the change from the single-column tuple. */
tup_change = ExecFetchSlotHeapTuple(dstate->tsslot, false, &shouldFree);
heap_deform_tuple(tup_change, dstate->tupdesc_change, values, isnull);
@@ -3261,10 +3283,22 @@ apply_concurrent_changes(ClusterDecodingState *dstate, Relation rel,
/* TTSOpsMinimalTuple has .get_heap_tuple==NULL. */
Assert(shouldFree);
pfree(tup_change);
+
+ /*
+ * If there is a limit on the time of completion, check it
+ * now. However, make sure the loop does not break if tup_old was set
+ * in the previous iteration. In such a case we could not resume the
+ * processing in the next call.
+ */
+ if (must_complete && tup_old == NULL &&
+ processing_time_elapsed(must_complete))
+ /* The next call will process the remaining changes. */
+ break;
}
- tuplestore_clear(dstate->tstore);
- dstate->nchanges = 0;
+ /* If we could not apply all the changes, the next call will do. */
+ if (dstate->nchanges == 0)
+ tuplestore_clear(dstate->tstore);
/* Cleanup. */
ExecDropSingleTupleTableSlot(index_slot);
@@ -3467,11 +3501,15 @@ find_target_tuple(Relation rel, ScanKey key, int nkeys, HeapTuple tup_key,
* Decode and apply concurrent changes.
*
* Pass rel_src iff its reltoastrelid is needed.
+ *
+ * Returns true if must_complete is NULL or if managed to complete by the time
+ * *must_complete indicates.
*/
-static void
+static bool
process_concurrent_changes(LogicalDecodingContext *ctx, XLogRecPtr end_of_wal,
Relation rel_dst, Relation rel_src, ScanKey ident_key,
- int ident_key_nentries, IndexInsertState *iistate)
+ int ident_key_nentries, IndexInsertState *iistate,
+ struct timeval *must_complete)
{
ClusterDecodingState *dstate;
@@ -3480,10 +3518,19 @@ process_concurrent_changes(LogicalDecodingContext *ctx, XLogRecPtr end_of_wal,
dstate = (ClusterDecodingState *) ctx->output_writer_private;
- cluster_decode_concurrent_changes(ctx, end_of_wal);
+ cluster_decode_concurrent_changes(ctx, end_of_wal, must_complete);
+
+ if (processing_time_elapsed(must_complete))
+ /* Caller is responsible for applying the changes. */
+ return false;
+ /*
+ * *must_complete not reached, so there are really no changes. (It's
+ * possible to see no changes just because not enough time was left for
+ * the decoding.)
+ */
if (dstate->nchanges == 0)
- return;
+ return true;
PG_TRY();
{
@@ -3495,7 +3542,7 @@ process_concurrent_changes(LogicalDecodingContext *ctx, XLogRecPtr end_of_wal,
rel_dst->rd_toastoid = rel_src->rd_rel->reltoastrelid;
apply_concurrent_changes(dstate, rel_dst, ident_key,
- ident_key_nentries, iistate);
+ ident_key_nentries, iistate, must_complete);
}
PG_FINALLY();
{
@@ -3505,6 +3552,28 @@ process_concurrent_changes(LogicalDecodingContext *ctx, XLogRecPtr end_of_wal,
rel_dst->rd_toastoid = InvalidOid;
}
PG_END_TRY();
+
+ /*
+ * apply_concurrent_changes() does check the processing time, so if some
+ * changes are left, we ran out of time.
+ */
+ return dstate->nchanges == 0;
+}
+
+/*
+ * Check if the current time is beyond *must_complete.
+ */
+static bool
+processing_time_elapsed(struct timeval *must_complete)
+{
+ struct timeval now;
+
+ if (must_complete == NULL)
+ return false;
+
+ gettimeofday(&now, NULL);
+
+ return timercmp(&now, must_complete, >);
}
static IndexInsertState *
@@ -3665,6 +3734,8 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
RelReopenInfo *rri = NULL;
int nrel;
Relation *ind_refs_all, *ind_refs_p;
+ struct timeval t_end;
+ struct timeval *t_end_ptr = NULL;
/* Like in cluster_rel(). */
lockmode_old = ShareUpdateExclusiveLock;
@@ -3744,7 +3815,8 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
*/
process_concurrent_changes(ctx, end_of_wal, NewHeap,
swap_toast_by_content ? OldHeap : NULL,
- ident_key, ident_key_nentries, iistate);
+ ident_key, ident_key_nentries, iistate,
+ NULL);
/*
* Release the locks that allowed concurrent data changes, in order to
@@ -3866,9 +3938,38 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
end_of_wal = GetFlushRecPtr(NULL);
/* Apply the concurrent changes again. */
- process_concurrent_changes(ctx, end_of_wal, NewHeap,
- swap_toast_by_content ? OldHeap : NULL,
- ident_key, ident_key_nentries, iistate);
+ /*
+ * This time we have the exclusive lock on the table, so make sure that
+ * cluster_max_xlock_time is not exceeded.
+ */
+ if (cluster_max_xlock_time > 0)
+ {
+ int64 usec;
+ struct timeval t_start;
+
+ gettimeofday(&t_start, NULL);
+ /* Add the whole seconds. */
+ t_end.tv_sec = t_start.tv_sec + cluster_max_xlock_time / 1000;
+ /* Add the rest, expressed in microseconds. */
+ usec = t_start.tv_usec + 1000 * (cluster_max_xlock_time % 1000);
+ /* The number of microseconds could have overflown. */
+ t_end.tv_sec += usec / USECS_PER_SEC;
+ t_end.tv_usec = usec % USECS_PER_SEC;
+ t_end_ptr = &t_end;
+ }
+ /*
+ * During testing, stop here to simulate excessive processing time.
+ */
+ INJECTION_POINT("cluster-concurrently-after-lock");
+
+ if (!process_concurrent_changes(ctx, end_of_wal, NewHeap,
+ swap_toast_by_content ? OldHeap : NULL,
+ ident_key, ident_key_nentries, iistate,
+ t_end_ptr))
+ ereport(ERROR,
+ (errmsg("could not process concurrent data changes in time"),
+ errhint("Please consider adjusting \"cluster_max_xlock_time\".")));
+
/* Remember info about rel before closing OldHeap */
relpersistence = OldHeap->rd_rel->relpersistence;
diff --git a/src/backend/utils/misc/guc_tables.c b/src/backend/utils/misc/guc_tables.c
index c9d8cd796a..7f4686f31e 100644
--- a/src/backend/utils/misc/guc_tables.c
+++ b/src/backend/utils/misc/guc_tables.c
@@ -39,6 +39,7 @@
#include "catalog/namespace.h"
#include "catalog/storage.h"
#include "commands/async.h"
+#include "commands/cluster.h"
#include "commands/event_trigger.h"
#include "commands/tablespace.h"
#include "commands/trigger.h"
@@ -2791,6 +2792,19 @@ struct config_int ConfigureNamesInt[] =
1600000000, 0, 2100000000,
NULL, NULL, NULL
},
+ {
+ {"cluster_max_xlock_time", PGC_USERSET, LOCK_MANAGEMENT,
+ gettext_noop("Maximum time for VACUUM FULL / CLUSTER (CONCURRENTLY) to keep table locked."),
+ gettext_noop(
+ "The table is locked in exclusive mode during the final stage of processing. "
+ "If the lock time exceeds this value, error is raised and the lock is "
+ "released. Set to zero if you don't care how long the lock can be held."),
+ GUC_UNIT_MS
+ },
+ &cluster_max_xlock_time,
+ 0, 0, INT_MAX,
+ NULL, NULL, NULL
+ },
/*
* See also CheckRequiredParameterValues() if this parameter changes
diff --git a/src/backend/utils/misc/postgresql.conf.sample b/src/backend/utils/misc/postgresql.conf.sample
index b2bc43383d..eef7be70c5 100644
--- a/src/backend/utils/misc/postgresql.conf.sample
+++ b/src/backend/utils/misc/postgresql.conf.sample
@@ -728,6 +728,7 @@ autovacuum_worker_slots = 16 # autovacuum worker slots to allocate
#vacuum_multixact_freeze_table_age = 150000000
#vacuum_multixact_freeze_min_age = 5000000
#vacuum_multixact_failsafe_age = 1600000000
+#cluster_max_xlock_time = 0
#bytea_output = 'hex' # hex, escape
#xmlbinary = 'base64'
#xmloption = 'content'
diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h
index 8945e46e64..72221e71d5 100644
--- a/src/include/commands/cluster.h
+++ b/src/include/commands/cluster.h
@@ -44,6 +44,8 @@ typedef struct ClusterParams
extern RelFileLocator clustered_rel_locator;
extern RelFileLocator clustered_rel_toast_locator;
+extern PGDLLIMPORT int cluster_max_xlock_time;
+
typedef enum
{
CHANGE_INSERT,
@@ -139,7 +141,8 @@ extern void mark_index_clustered(Relation rel, Oid indexOid, bool is_internal);
extern void check_relation_is_clusterable_concurrently(Relation rel,
bool is_vacuum);
extern void cluster_decode_concurrent_changes(LogicalDecodingContext *ctx,
- XLogRecPtr end_of_wal);
+ XLogRecPtr end_of_wal,
+ struct timeval *must_complete);
extern Oid make_new_heap(Oid OIDOldHeap, Oid NewTableSpace, Oid NewAccessMethod,
char relpersistence, LOCKMODE lockmode);
extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap,
diff --git a/src/test/modules/injection_points/expected/cluster.out b/src/test/modules/injection_points/expected/cluster.out
index d84fff3693..646e31448f 100644
--- a/src/test/modules/injection_points/expected/cluster.out
+++ b/src/test/modules/injection_points/expected/cluster.out
@@ -1,4 +1,4 @@
-Parsed test spec with 2 sessions
+Parsed test spec with 4 sessions
starting permutation: wait_before_lock change_existing change_new change_subxact1 change_subxact2 check2 wakeup_before_lock check1
injection_points_attach
@@ -111,3 +111,75 @@ injection_points_detach
(1 row)
+injection_points_detach
+-----------------------
+
+(1 row)
+
+
+starting permutation: wait_after_lock wakeup_after_lock
+injection_points_attach
+-----------------------
+
+(1 row)
+
+step wait_after_lock:
+ CLUSTER (CONCURRENTLY) clstr_test USING clstr_test_pkey;
+ <waiting ...>
+step wakeup_after_lock:
+ SELECT injection_points_wakeup('cluster-concurrently-after-lock');
+
+injection_points_wakeup
+-----------------------
+
+(1 row)
+
+step wait_after_lock: <... completed>
+injection_points_detach
+-----------------------
+
+(1 row)
+
+injection_points_detach
+-----------------------
+
+(1 row)
+
+
+starting permutation: wait_after_lock after_lock_delay wakeup_after_lock
+injection_points_attach
+-----------------------
+
+(1 row)
+
+step wait_after_lock:
+ CLUSTER (CONCURRENTLY) clstr_test USING clstr_test_pkey;
+ <waiting ...>
+step after_lock_delay:
+ SELECT pg_sleep(1.5);
+
+pg_sleep
+--------
+
+(1 row)
+
+step wakeup_after_lock:
+ SELECT injection_points_wakeup('cluster-concurrently-after-lock');
+
+injection_points_wakeup
+-----------------------
+
+(1 row)
+
+step wait_after_lock: <... completed>
+ERROR: could not process concurrent data changes in time
+injection_points_detach
+-----------------------
+
+(1 row)
+
+injection_points_detach
+-----------------------
+
+(1 row)
+
diff --git a/src/test/modules/injection_points/specs/cluster.spec b/src/test/modules/injection_points/specs/cluster.spec
index 5f8404c5da..9af41bac6d 100644
--- a/src/test/modules/injection_points/specs/cluster.spec
+++ b/src/test/modules/injection_points/specs/cluster.spec
@@ -127,6 +127,34 @@ step wakeup_before_lock
SELECT injection_points_wakeup('cluster-concurrently-before-lock');
}
+session s3
+setup
+{
+ SELECT injection_points_set_local();
+ SELECT injection_points_attach('cluster-concurrently-after-lock', 'wait');
+ SET cluster_max_xlock_time TO '1s';
+}
+# Perform the initial load, lock the table in exclusive mode and wait. s4 will
+# cancel the waiting.
+step wait_after_lock
+{
+ CLUSTER (CONCURRENTLY) clstr_test USING clstr_test_pkey;
+}
+teardown
+{
+ SELECT injection_points_detach('cluster-concurrently-after-lock');
+}
+
+session s4
+step wakeup_after_lock
+{
+ SELECT injection_points_wakeup('cluster-concurrently-after-lock');
+}
+step after_lock_delay
+{
+ SELECT pg_sleep(1.5);
+}
+
# Test if data changes introduced while one session is performing CLUSTER
# (CONCURRENTLY) find their way into the table.
permutation
@@ -138,3 +166,17 @@ permutation
check2
wakeup_before_lock
check1
+
+# Test the cluster_max_xlock_time configuration variable.
+#
+# First, cancel waiting on the injection point immediately. That way, CLUSTER
+# should complete.
+permutation
+ wait_after_lock
+ wakeup_after_lock
+# Second, cancel the waiting with a delay that violates
+# cluster_max_xlock_time.
+permutation
+ wait_after_lock
+ after_lock_delay
+ wakeup_after_lock
--
2.45.2
view thread (7+ messages) latest in thread
reply
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Reply to all the recipients using the --to and --cc options:
reply via email
To: [email protected]
Cc: [email protected], [email protected], [email protected], [email protected], [email protected], [email protected]
Subject: Re: why there is not VACUUM FULL CONCURRENTLY?
In-Reply-To: <6250.1736776111@antos>
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
This inbox is served by agora; see mirroring instructions
for how to clone and mirror all data and code used for this inbox