Skip to content

Commit 63631ee

Browse files
nmischnmisch
committed
Revert "Skip WAL for new relfilenodes, under wal_level=minimal."
This reverts commit cb2fd7e. Per numerous buildfarm members, it was incompatible with parallel query, and a test case assumed LP64. Back-patch to 9.5 (all supported versions). Discussion: https://postgr.es/m/20200321224920.GB1763544@rfd.leadboat.com
1 parent e4b0a02 commit 63631ee

Some content is hidden

Large Commits have some content hidden by default. Use the searchbox below for content that may be hidden.

51 files changed

+332
-1447
lines changed

‎contrib/pg_visibility/expected/pg_visibility.out

Lines changed: 0 additions & 35 deletions
Original file line numberDiff line numberDiff line change
@@ -1,40 +1,5 @@
11
CREATE EXTENSION pg_visibility;
22
--
3-
-- recently-dropped table
4-
--
5-
\set VERBOSITY sqlstate
6-
BEGIN;
7-
CREATE TABLE droppedtest (c int);
8-
SELECT 'droppedtest'::regclass::oid AS oid \gset
9-
SAVEPOINT q; DROP TABLE droppedtest; RELEASE q;
10-
SAVEPOINT q; SELECT * FROM pg_visibility_map(:oid); ROLLBACK TO q;
11-
ERROR: XX000
12-
-- ERROR: could not open relation with OID 16xxx
13-
SAVEPOINT q; SELECT 1; ROLLBACK TO q;
14-
?column?
15-
----------
16-
1
17-
(1 row)
18-
19-
SAVEPOINT q; SELECT 1; ROLLBACK TO q;
20-
?column?
21-
----------
22-
1
23-
(1 row)
24-
25-
SELECT pg_relation_size(:oid), pg_relation_filepath(:oid),
26-
has_table_privilege(:oid, 'SELECT');
27-
pg_relation_size | pg_relation_filepath | has_table_privilege
28-
------------------+----------------------+---------------------
29-
| |
30-
(1 row)
31-
32-
SELECT * FROM pg_visibility_map(:oid);
33-
ERROR: XX000
34-
-- ERROR: could not open relation with OID 16xxx
35-
ROLLBACK;
36-
\set VERBOSITY default
37-
--
383
-- check that using the module's functions with unsupported relations will fail
394
--
405
-- partitioned tables (the parent ones) don't have visibility maps

‎contrib/pg_visibility/sql/pg_visibility.sql

Lines changed: 0 additions & 19 deletions
Original file line numberDiff line numberDiff line change
@@ -1,24 +1,5 @@
11
CREATE EXTENSION pg_visibility;
22

3-
--
4-
-- recently-dropped table
5-
--
6-
\set VERBOSITY sqlstate
7-
BEGIN;
8-
CREATE TABLE droppedtest (c int);
9-
SELECT 'droppedtest'::regclass::oid AS oid \gset
10-
SAVEPOINT q; DROP TABLE droppedtest; RELEASE q;
11-
SAVEPOINT q; SELECT * FROM pg_visibility_map(:oid); ROLLBACK TO q;
12-
-- ERROR: could not open relation with OID 16xxx
13-
SAVEPOINT q; SELECT 1; ROLLBACK TO q;
14-
SAVEPOINT q; SELECT 1; ROLLBACK TO q;
15-
SELECT pg_relation_size(:oid), pg_relation_filepath(:oid),
16-
has_table_privilege(:oid, 'SELECT');
17-
SELECT * FROM pg_visibility_map(:oid);
18-
-- ERROR: could not open relation with OID 16xxx
19-
ROLLBACK;
20-
\set VERBOSITY default
21-
223
--
234
-- check that using the module's functions with unsupported relations will fail
245
--

‎doc/src/sgml/config.sgml

Lines changed: 8 additions & 31 deletions
Original file line numberDiff line numberDiff line change
@@ -2462,19 +2462,16 @@ include_dir 'conf.d'
24622462
levels. This parameter can only be set at server start.
24632463
</para>
24642464
<para>
2465-
In <literal>minimal</literal> level, no information is logged for
2466-
permanent relations for the remainder of a transaction that creates or
2467-
rewrites them. This can make operations much faster (see
2468-
<xref linkend="populate-pitr"/>). Operations that initiate this
2469-
optimization include:
2465+
In <literal>minimal</literal> level, WAL-logging of some bulk
2466+
operations can be safely skipped, which can make those
2467+
operations much faster (see <xref linkend="populate-pitr"/>).
2468+
Operations in which this optimization can be applied include:
24702469
<simplelist>
2471-
<member><command>ALTER ... SET TABLESPACE</command></member>
2470+
<member><command>CREATE TABLE AS</command></member>
2471+
<member><command>CREATE INDEX</command></member>
24722472
<member><command>CLUSTER</command></member>
2473-
<member><command>CREATE TABLE</command></member>
2474-
<member><command>REFRESH MATERIALIZED VIEW</command>
2475-
(without <option>CONCURRENTLY</option>)</member>
2476-
<member><command>REINDEX</command></member>
2477-
<member><command>TRUNCATE</command></member>
2473+
<member><command>COPY</command> into tables that were created or truncated in the same
2474+
transaction</member>
24782475
</simplelist>
24792476
But minimal WAL does not contain enough information to reconstruct the
24802477
data from a base backup and the WAL logs, so <literal>replica</literal> or
@@ -2871,26 +2868,6 @@ include_dir 'conf.d'
28712868
</listitem>
28722869
</varlistentry>
28732870

2874-
<varlistentry id="guc-wal-skip-threshold" xreflabel="wal_skip_threshold">
2875-
<term><varname>wal_skip_threshold</varname> (<type>integer</type>)
2876-
<indexterm>
2877-
<primary><varname>wal_skip_threshold</varname> configuration parameter</primary>
2878-
</indexterm>
2879-
</term>
2880-
<listitem>
2881-
<para>
2882-
When <varname>wal_level</varname> is <literal>minimal</literal> and a
2883-
transaction commits after creating or rewriting a permanent relation,
2884-
this setting determines how to persist the new data. If the data is
2885-
smaller than this setting, write it to the WAL log; otherwise, use an
2886-
fsync of affected files. Depending on the properties of your storage,
2887-
raising or lowering this value might help if such commits are slowing
2888-
concurrent transactions. The default is two megabytes
2889-
(<literal>2MB</literal>).
2890-
</para>
2891-
</listitem>
2892-
</varlistentry>
2893-
28942871
<varlistentry id="guc-commit-delay" xreflabel="commit_delay">
28952872
<term><varname>commit_delay</varname> (<type>integer</type>)
28962873
<indexterm>

‎doc/src/sgml/perform.sgml

Lines changed: 38 additions & 9 deletions
Original file line numberDiff line numberDiff line change
@@ -1606,8 +1606,8 @@ SELECT * FROM x, y, a, b, c WHERE something AND somethingelse;
16061606
needs to be written, because in case of an error, the files
16071607
containing the newly loaded data will be removed anyway.
16081608
However, this consideration only applies when
1609-
<xref linkend="guc-wal-level"/> is <literal>minimal</literal>
1610-
as all commands must write WAL otherwise.
1609+
<xref linkend="guc-wal-level"/> is <literal>minimal</literal> for
1610+
non-partitioned tables as all commands must write WAL otherwise.
16111611
</para>
16121612

16131613
</sect2>
@@ -1707,13 +1707,42 @@ SELECT * FROM x, y, a, b, c WHERE something AND somethingelse;
17071707
</para>
17081708

17091709
<para>
1710-
Aside from avoiding the time for the archiver or WAL sender to process the
1711-
WAL data, doing this will actually make certain commands faster, because
1712-
they do not to write WAL at all if <varname>wal_level</varname>
1713-
is <literal>minimal</literal> and the current subtransaction (or top-level
1714-
transaction) created or truncated the table or index they change. (They
1715-
can guarantee crash safety more cheaply by doing
1716-
an <function>fsync</function> at the end than by writing WAL.)
1710+
Aside from avoiding the time for the archiver or WAL sender to
1711+
process the WAL data,
1712+
doing this will actually make certain commands faster, because they
1713+
are designed not to write WAL at all if <varname>wal_level</varname>
1714+
is <literal>minimal</literal>. (They can guarantee crash safety more cheaply
1715+
by doing an <function>fsync</function> at the end than by writing WAL.)
1716+
This applies to the following commands:
1717+
<itemizedlist>
1718+
<listitem>
1719+
<para>
1720+
<command>CREATE TABLE AS SELECT</command>
1721+
</para>
1722+
</listitem>
1723+
<listitem>
1724+
<para>
1725+
<command>CREATE INDEX</command> (and variants such as
1726+
<command>ALTER TABLE ADD PRIMARY KEY</command>)
1727+
</para>
1728+
</listitem>
1729+
<listitem>
1730+
<para>
1731+
<command>ALTER TABLE SET TABLESPACE</command>
1732+
</para>
1733+
</listitem>
1734+
<listitem>
1735+
<para>
1736+
<command>CLUSTER</command>
1737+
</para>
1738+
</listitem>
1739+
<listitem>
1740+
<para>
1741+
<command>COPY FROM</command>, when the target table has been
1742+
created or truncated earlier in the same transaction
1743+
</para>
1744+
</listitem>
1745+
</itemizedlist>
17171746
</para>
17181747
</sect2>
17191748

‎src/backend/access/gist/gistutil.c

Lines changed: 5 additions & 26 deletions
Original file line numberDiff line numberDiff line change
@@ -1019,44 +1019,23 @@ gistproperty(Oid index_oid, int attno,
10191019
}
10201020

10211021
/*
1022-
* Some indexes are not WAL-logged, but we need LSNs to detect concurrent page
1023-
* splits anyway. This function provides a fake sequence of LSNs for that
1024-
* purpose.
1022+
* Temporary and unlogged GiST indexes are not WAL-logged, but we need LSNs
1023+
* to detect concurrent page splits anyway. This function provides a fake
1024+
* sequence of LSNs for that purpose.
10251025
*/
10261026
XLogRecPtr
10271027
gistGetFakeLSN(Relation rel)
10281028
{
1029+
static XLogRecPtr counter = FirstNormalUnloggedLSN;
1030+
10291031
if (rel->rd_rel->relpersistence == RELPERSISTENCE_TEMP)
10301032
{
10311033
/*
10321034
* Temporary relations are only accessible in our session, so a simple
10331035
* backend-local counter will do.
10341036
*/
1035-
static XLogRecPtr counter = FirstNormalUnloggedLSN;
1036-
10371037
return counter++;
10381038
}
1039-
else if (rel->rd_rel->relpersistence == RELPERSISTENCE_PERMANENT)
1040-
{
1041-
/*
1042-
* WAL-logging on this relation will start after commit, so its LSNs
1043-
* must be distinct numbers smaller than the LSN at the next commit.
1044-
* Emit a dummy WAL record if insert-LSN hasn't advanced after the
1045-
* last call.
1046-
*/
1047-
static XLogRecPtr lastlsn = InvalidXLogRecPtr;
1048-
XLogRecPtr currlsn = GetXLogInsertRecPtr();
1049-
1050-
/* Shouldn't be called for WAL-logging relations */
1051-
Assert(!RelationNeedsWAL(rel));
1052-
1053-
/* No need for an actual record if we already have a distinct LSN */
1054-
if (!XLogRecPtrIsInvalid(lastlsn) && lastlsn == currlsn)
1055-
currlsn = gistXLogAssignLSN();
1056-
1057-
lastlsn = currlsn;
1058-
return currlsn;
1059-
}
10601039
else
10611040
{
10621041
/*

‎src/backend/access/gist/gistxlog.c

Lines changed: 0 additions & 21 deletions
Original file line numberDiff line numberDiff line change
@@ -449,9 +449,6 @@ gist_redo(XLogReaderState *record)
449449
case XLOG_GIST_PAGE_DELETE:
450450
gistRedoPageDelete(record);
451451
break;
452-
case XLOG_GIST_ASSIGN_LSN:
453-
/* nop. See gistGetFakeLSN(). */
454-
break;
455452
default:
456453
elog(PANIC, "gist_redo: unknown op code %u", info);
457454
}
@@ -595,24 +592,6 @@ gistXLogPageDelete(Buffer buffer, FullTransactionId xid,
595592
return recptr;
596593
}
597594

598-
/*
599-
* Write an empty XLOG record to assign a distinct LSN.
600-
*/
601-
XLogRecPtr
602-
gistXLogAssignLSN(void)
603-
{
604-
int dummy = 0;
605-
606-
/*
607-
* Records other than SWITCH_WAL must have content. We use an integer 0 to
608-
* follow the restriction.
609-
*/
610-
XLogBeginInsert();
611-
XLogSetRecordFlags(XLOG_MARK_UNIMPORTANT);
612-
XLogRegisterData((char *) &dummy, sizeof(dummy));
613-
return XLogInsert(RM_GIST_ID, XLOG_GIST_ASSIGN_LSN);
614-
}
615-
616595
/*
617596
* Write XLOG record about reuse of a deleted page.
618597
*/

‎src/backend/access/heap/heapam.c

Lines changed: 14 additions & 8 deletions
Original file line numberDiff line numberDiff line change
@@ -21,6 +21,7 @@
2121
* heap_multi_insert - insert multiple tuples into a relation
2222
* heap_delete - delete a tuple from a relation
2323
* heap_update - replace a tuple in a relation with another tuple
24+
* heap_sync - sync heap, for when no WAL has been written
2425
*
2526
* NOTES
2627
* This file contains the heap_ routines which implement
@@ -1935,7 +1936,7 @@ heap_insert(Relation relation, HeapTuple tup, CommandId cid,
19351936
MarkBufferDirty(buffer);
19361937

19371938
/* XLOG stuff */
1938-
if (RelationNeedsWAL(relation))
1939+
if (!(options & HEAP_INSERT_SKIP_WAL) && RelationNeedsWAL(relation))
19391940
{
19401941
xl_heap_insert xlrec;
19411942
xl_heap_header xlhdr;
@@ -2118,7 +2119,7 @@ heap_multi_insert(Relation relation, TupleTableSlot **slots, int ntuples,
21182119
/* currently not needed (thus unsupported) for heap_multi_insert() */
21192120
AssertArg(!(options & HEAP_INSERT_NO_LOGICAL));
21202121

2121-
needwal = RelationNeedsWAL(relation);
2122+
needwal = !(options & HEAP_INSERT_SKIP_WAL) && RelationNeedsWAL(relation);
21222123
saveFreeSpace = RelationGetTargetPageFreeSpace(relation,
21232124
HEAP_DEFAULT_FILLFACTOR);
21242125

@@ -8920,13 +8921,18 @@ heap2_redo(XLogReaderState *record)
89208921
}
89218922

89228923
/*
8923-
* heap_sync - for binary compatibility
8924+
* heap_sync - sync a heap, for use when no WAL has been written
8925+
*
8926+
* This forces the heap contents (including TOAST heap if any) down to disk.
8927+
* If we skipped using WAL, and WAL is otherwise needed, we must force the
8928+
* relation down to disk before it's safe to commit the transaction. This
8929+
* requires writing out any dirty buffers and then doing a forced fsync.
89248930
*
8925-
* A newer PostgreSQL version removes this function. It exists here just in
8926-
* case an extension calls it. See "Skipping WAL for New RelFileNode" in
8927-
* src/backend/access/transam/README for the system that superseded it,
8928-
* allowing removal of most calls. Cases like RelationCopyStorage() should
8929-
* call smgrimmedsync() directly.
8931+
* Indexes are not touched. (Currently, index operations associated with
8932+
* the commands that use this are WAL-logged and so do not need fsync.
8933+
* That behavior might change someday, but in any case it's likely that
8934+
* any fsync decisions required would be per-index and hence not appropriate
8935+
* to be done here.)
89308936
*/
89318937
void
89328938
heap_sync(Relation rel)

‎src/backend/access/heap/heapam_handler.c

Lines changed: 19 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -559,6 +559,17 @@ heapam_tuple_lock(Relation relation, ItemPointer tid, Snapshot snapshot,
559559
return result;
560560
}
561561

562+
static void
563+
heapam_finish_bulk_insert(Relation relation, int options)
564+
{
565+
/*
566+
* If we skipped writing WAL, then we need to sync the heap (but not
567+
* indexes since those use WAL anyway / don't go through tableam)
568+
*/
569+
if (options & HEAP_INSERT_SKIP_WAL)
570+
heap_sync(relation);
571+
}
572+
562573

563574
/* ------------------------------------------------------------------------
564575
* DDL related callbacks for heap AM.
@@ -691,6 +702,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap,
691702
IndexScanDesc indexScan;
692703
TableScanDesc tableScan;
693704
HeapScanDesc heapScan;
705+
bool use_wal;
694706
bool is_system_catalog;
695707
Tuplesortstate *tuplesort;
696708
TupleDesc oldTupDesc = RelationGetDescr(OldHeap);
@@ -705,9 +717,12 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap,
705717
is_system_catalog = IsSystemRelation(OldHeap);
706718

707719
/*
708-
* Valid smgr_targblock implies something already wrote to the relation.
709-
* This may be harmless, but this function hasn't planned for it.
720+
* We need to log the copied data in WAL iff WAL archiving/streaming is
721+
* enabled AND it's a WAL-logged rel.
710722
*/
723+
use_wal = XLogIsNeeded() && RelationNeedsWAL(NewHeap);
724+
725+
/* use_wal off requires smgr_targblock be initially invalid */
711726
Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber);
712727

713728
/* Preallocate values/isnull arrays */
@@ -717,7 +732,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap,
717732

718733
/* Initialize the rewrite operation */
719734
rwstate = begin_heap_rewrite(OldHeap, NewHeap, OldestXmin, *xid_cutoff,
720-
*multi_cutoff);
735+
*multi_cutoff, use_wal);
721736

722737

723738
/* Set up sorting if wanted */
@@ -2611,6 +2626,7 @@ static const TableAmRoutine heapam_methods = {
26112626
.tuple_delete = heapam_tuple_delete,
26122627
.tuple_update = heapam_tuple_update,
26132628
.tuple_lock = heapam_tuple_lock,
2629+
.finish_bulk_insert = heapam_finish_bulk_insert,
26142630

26152631
.tuple_fetch_row_version = heapam_fetch_row_version,
26162632
.tuple_get_latest_tid = heap_get_latest_tid,

0 commit comments

Comments
 (0)