Movatterモバイル変換

[0]ホーム
URL:

Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly

 Sign up
Appearance settings

Commit0668c84

Browse files
committed
Fix bug in cancellation of non-exclusive backup to avoid assertion failure.
Previously an assertion failure occurred when pg_stop_backup() fornon-exclusive backup was aborted while it's waiting for WAL files tobe archived. This assertion failure happened in do_pg_abort_backup()which was called when a non-exclusive backup was canceled.do_pg_abort_backup() assumes that there is at least one non-exclusivebackup running when it's called. But pg_stop_backup() can be canceledeven after it marks the end of non-exclusive backup (e.g.,during waiting for WAL archiving). This broke the assumption thatdo_pg_abort_backup() relies on, and which caused an assertion failure.This commit changes do_pg_abort_backup() so that it does nothingwhen non-exclusive backup has been already marked as completed.That is, the asssumption is also changed, and do_pg_abort_backup()now can handle even the case where it's called when there isno running backup.Backpatch to 9.6 where SQL-callable non-exclusive backup was added.Author: Masahiko Sawada and Michael PaquierReviewed-By: Robert Haas and Fujii MasaoDiscussion:https://www.postgresql.org/message-id/CAD21AoD2L1Fu2c==gnVASMyFAAaq3y-AQ2uEVj-zTCGFFjvmDg@mail.gmail.com
1 parent986a915 commit0668c84

File tree

2 files changed

+35
-6
lines changed

2 files changed

+35
-6
lines changed

‎src/backend/access/transam/xlog.c

Lines changed: 32 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -10275,13 +10275,20 @@ do_pg_start_backup(const char *backupidstr, bool fast, TimeLineID *starttli_p,
1027510275
/*
1027610276
* Mark that start phase has correctly finished for an exclusive backup.
1027710277
* Session-level locks are updated as well to reflect that state.
10278+
*
10279+
* Note that CHECK_FOR_INTERRUPTS() must not occur while updating
10280+
* backup counters and session-level lock. Otherwise they can be
10281+
* updated inconsistently, and which might cause do_pg_abort_backup()
10282+
* to fail.
1027810283
*/
1027910284
if (exclusive)
1028010285
{
1028110286
WALInsertLockAcquireExclusive();
1028210287
XLogCtl->Insert.exclusiveBackupState=EXCLUSIVE_BACKUP_IN_PROGRESS;
10283-
WALInsertLockRelease();
10288+
10289+
/* Set session-level lock */
1028410290
sessionBackupState=SESSION_BACKUP_EXCLUSIVE;
10291+
WALInsertLockRelease();
1028510292
}
1028610293
else
1028710294
sessionBackupState=SESSION_BACKUP_NON_EXCLUSIVE;
@@ -10489,7 +10496,11 @@ do_pg_stop_backup(char *labelfile, bool waitforarchive, TimeLineID *stoptli_p)
1048910496
}
1049010497

1049110498
/*
10492-
* OK to update backup counters and forcePageWrites
10499+
* OK to update backup counters, forcePageWrites and session-level lock.
10500+
*
10501+
* Note that CHECK_FOR_INTERRUPTS() must not occur while updating them.
10502+
* Otherwise they can be updated inconsistently, and which might cause
10503+
* do_pg_abort_backup() to fail.
1049310504
*/
1049410505
WALInsertLockAcquireExclusive();
1049510506
if (exclusive)
@@ -10513,11 +10524,20 @@ do_pg_stop_backup(char *labelfile, bool waitforarchive, TimeLineID *stoptli_p)
1051310524
{
1051410525
XLogCtl->Insert.forcePageWrites= false;
1051510526
}
10516-
WALInsertLockRelease();
1051710527

10518-
/* Clean up session-level lock */
10528+
/*
10529+
* Clean up session-level lock.
10530+
*
10531+
* You might think that WALInsertLockRelease() can be called
10532+
* before cleaning up session-level lock because session-level
10533+
* lock doesn't need to be protected with WAL insertion lock.
10534+
* But since CHECK_FOR_INTERRUPTS() can occur in it,
10535+
* session-level lock must be cleaned up before it.
10536+
*/
1051910537
sessionBackupState=SESSION_BACKUP_NONE;
1052010538

10539+
WALInsertLockRelease();
10540+
1052110541
/*
1052210542
* Read and parse the START WAL LOCATION line (this code is pretty crude,
1052310543
* but we are not expecting any variability in the file format).
@@ -10750,8 +10770,16 @@ do_pg_stop_backup(char *labelfile, bool waitforarchive, TimeLineID *stoptli_p)
1075010770
void
1075110771
do_pg_abort_backup(void)
1075210772
{
10773+
/*
10774+
* Quick exit if session is not keeping around a non-exclusive backup
10775+
* already started.
10776+
*/
10777+
if (sessionBackupState==SESSION_BACKUP_NONE)
10778+
return;
10779+
1075310780
WALInsertLockAcquireExclusive();
1075410781
Assert(XLogCtl->Insert.nonExclusiveBackups>0);
10782+
Assert(sessionBackupState==SESSION_BACKUP_NON_EXCLUSIVE);
1075510783
XLogCtl->Insert.nonExclusiveBackups--;
1075610784

1075710785
if (XLogCtl->Insert.exclusiveBackupState==EXCLUSIVE_BACKUP_NONE&&

‎src/backend/replication/basebackup.c

Lines changed: 3 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -140,7 +140,7 @@ perform_base_backup(basebackup_options *opt, DIR *tblspcdir)
140140
* Once do_pg_start_backup has been called, ensure that any failure causes
141141
* us to abort the backup so we don't "leak" a backup counter. For this
142142
* reason, *all* functionality between do_pg_start_backup() and
143-
* do_pg_stop_backup() should be inside the error cleanup block!
143+
*the end ofdo_pg_stop_backup() should be inside the error cleanup block!
144144
*/
145145

146146
PG_ENSURE_ERROR_CLEANUP(base_backup_cleanup, (Datum)0);
@@ -249,10 +249,11 @@ perform_base_backup(basebackup_options *opt, DIR *tblspcdir)
249249
else
250250
pq_putemptymessage('c');/* CopyDone */
251251
}
252+
253+
endptr=do_pg_stop_backup(labelfile->data, !opt->nowait,&endtli);
252254
}
253255
PG_END_ENSURE_ERROR_CLEANUP(base_backup_cleanup, (Datum)0);
254256

255-
endptr=do_pg_stop_backup(labelfile->data, !opt->nowait,&endtli);
256257

257258
if (opt->includewal)
258259
{

0 commit comments

Comments
 (0)
[8]ページ先頭
©2009-2025 Movatter.jp