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

Commit6d8096e

Browse files
committed
When two base backups are started at the same time with pg_basebackup,
ensure that they use different checkpoints as the starting point. We usethe checkpoint redo location as a unique identifier for the base backup inthe end-of-backup record, and in the backup history file name.Bug spotted by Fujii Masao.
1 parent62bf9c2 commit6d8096e

File tree

1 file changed

+61
-34
lines changed
  • src/backend/access/transam

1 file changed

+61
-34
lines changed

‎src/backend/access/transam/xlog.c‎

Lines changed: 61 additions & 34 deletions
Original file line numberDiff line numberDiff line change
@@ -355,10 +355,13 @@ typedef struct XLogCtlInsert
355355
* exclusiveBackup is true if a backup started with pg_start_backup() is
356356
* in progress, and nonExclusiveBackups is a counter indicating the number
357357
* of streaming base backups currently in progress. forcePageWrites is
358-
* set to true when either of these is non-zero.
358+
* set to true when either of these is non-zero. lastBackupStart is the
359+
* latest checkpoint redo location used as a starting point for an online
360+
* backup.
359361
*/
360362
boolexclusiveBackup;
361363
intnonExclusiveBackups;
364+
XLogRecPtrlastBackupStart;
362365
}XLogCtlInsert;
363366

364367
/*
@@ -8808,6 +8811,19 @@ do_pg_start_backup(const char *backupidstr, bool fast, char **labelfile)
88088811
errmsg("backup label too long (max %d bytes)",
88098812
MAXPGPATH)));
88108813

8814+
/*
8815+
* Force an XLOG file switch before the checkpoint, to ensure that the WAL
8816+
* segment the checkpoint is written to doesn't contain pages with old
8817+
* timeline IDs. That would otherwise happen if you called
8818+
* pg_start_backup() right after restoring from a PITR archive: the first
8819+
* WAL segment containing the startup checkpoint has pages in the
8820+
* beginning with the old timeline ID. That can cause trouble at recovery:
8821+
* we won't have a history file covering the old timeline if pg_xlog
8822+
* directory was not included in the base backup and the WAL archive was
8823+
* cleared too before starting the backup.
8824+
*/
8825+
RequestXLogSwitch();
8826+
88118827
/*
88128828
* Mark backup active in shared memory. We must do full-page WAL writes
88138829
* during an on-line backup even if not doing so at other times, because
@@ -8843,43 +8859,54 @@ do_pg_start_backup(const char *backupidstr, bool fast, char **labelfile)
88438859
XLogCtl->Insert.forcePageWrites= true;
88448860
LWLockRelease(WALInsertLock);
88458861

8846-
/*
8847-
* Force an XLOG file switch before the checkpoint, to ensure that the WAL
8848-
* segment the checkpoint is written to doesn't contain pages with old
8849-
* timeline IDs. That would otherwise happen if you called
8850-
* pg_start_backup() right after restoring from a PITR archive: the first
8851-
* WAL segment containing the startup checkpoint has pages in the
8852-
* beginning with the old timeline ID. That can cause trouble at recovery:
8853-
* we won't have a history file covering the old timeline if pg_xlog
8854-
* directory was not included in the base backup and the WAL archive was
8855-
* cleared too before starting the backup.
8856-
*/
8857-
RequestXLogSwitch();
8858-
88598862
/* Ensure we release forcePageWrites if fail below */
88608863
PG_ENSURE_ERROR_CLEANUP(pg_start_backup_callback, (Datum)BoolGetDatum(exclusive));
88618864
{
8862-
/*
8863-
* Force a CHECKPOINT.Aside from being necessary to prevent torn
8864-
* page problems, this guarantees that two successive backup runs will
8865-
* have different checkpoint positions and hence different history
8866-
* file names, even if nothing happened in between.
8867-
*
8868-
* We use CHECKPOINT_IMMEDIATE only if requested by user (via passing
8869-
* fast = true). Otherwise this can take awhile.
8870-
*/
8871-
RequestCheckpoint(CHECKPOINT_FORCE |CHECKPOINT_WAIT |
8872-
(fast ?CHECKPOINT_IMMEDIATE :0));
8865+
boolgotUniqueStartpoint= false;
8866+
do
8867+
{
8868+
/*
8869+
* Force a CHECKPOINT.Aside from being necessary to prevent torn
8870+
* page problems, this guarantees that two successive backup runs will
8871+
* have different checkpoint positions and hence different history
8872+
* file names, even if nothing happened in between.
8873+
*
8874+
* We use CHECKPOINT_IMMEDIATE only if requested by user (via passing
8875+
* fast = true). Otherwise this can take awhile.
8876+
*/
8877+
RequestCheckpoint(CHECKPOINT_FORCE |CHECKPOINT_WAIT |
8878+
(fast ?CHECKPOINT_IMMEDIATE :0));
88738879

8874-
/*
8875-
* Now we need to fetch the checkpoint record location, and also its
8876-
* REDO pointer. The oldest point in WAL that would be needed to
8877-
* restore starting from the checkpoint is precisely the REDO pointer.
8878-
*/
8879-
LWLockAcquire(ControlFileLock,LW_SHARED);
8880-
checkpointloc=ControlFile->checkPoint;
8881-
startpoint=ControlFile->checkPointCopy.redo;
8882-
LWLockRelease(ControlFileLock);
8880+
/*
8881+
* Now we need to fetch the checkpoint record location, and also its
8882+
* REDO pointer. The oldest point in WAL that would be needed to
8883+
* restore starting from the checkpoint is precisely the REDO pointer.
8884+
*/
8885+
LWLockAcquire(ControlFileLock,LW_SHARED);
8886+
checkpointloc=ControlFile->checkPoint;
8887+
startpoint=ControlFile->checkPointCopy.redo;
8888+
LWLockRelease(ControlFileLock);
8889+
8890+
/*
8891+
* If two base backups are started at the same time (in WAL
8892+
* sender processes), we need to make sure that they use
8893+
* different checkpoints as starting locations, because we use
8894+
* the starting WAL location as a unique identifier for the base
8895+
* backup in the end-of-backup WAL record and when we write the
8896+
* backup history file. Perhaps it would be better generate a
8897+
* separate unique ID for each backup instead of forcing another
8898+
* checkpoint, but taking a checkpoint right after another is
8899+
* not that expensive either because only few buffers have been
8900+
* dirtied yet.
8901+
*/
8902+
LWLockAcquire(WALInsertLock,LW_SHARED);
8903+
if (XLByteLT(XLogCtl->Insert.lastBackupStart,startpoint))
8904+
{
8905+
XLogCtl->Insert.lastBackupStart=startpoint;
8906+
gotUniqueStartpoint= true;
8907+
}
8908+
LWLockRelease(WALInsertLock);
8909+
}while(!gotUniqueStartpoint);
88838910

88848911
XLByteToSeg(startpoint,_logId,_logSeg);
88858912
XLogFileName(xlogfilename,ThisTimeLineID,_logId,_logSeg);

0 commit comments

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