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

Commit3c0ef08

Browse files
committed
Prevent WAL corruption after a standby promotion.
When a PostgreSQL instance performing archive recovery but not usingstandby mode is promoted, and the last WAL segment that it attemptedto read ended in a partial record, the previous code would createinvalid WAL on the new timeline. The WAL from the previously timelinewould be copied to the new timeline up until the end of the last validrecord, but instead of beginning to write WAL at immediatelyafterwards, the promoted server would write an overwrite contrecord atthe beginning of the next segment. The end of the previous segmentwould be left as all-zeroes, resulting in failures if anything triedto read WAL from that file.The root of the issue is that ReadRecord() decides whether to setabortedRecPtr and missingContrecPtr based on the value of StandbyMode,but ReadRecord() switches to a new timeline based on the value ofArchiveRecoveryRequested. We shouldn't try to write an overwritecontrecord if we're switching to a new timeline, so change the test inReadRecod() to check ArchiveRecoveryRequested instead.Code fix by Dilip Kumar. Comments by me incorporating suggestedlanguage from Álvaro Herrera. Further review from Kyotaro Horiguchiand Sami Imseih.Discussion:http://postgr.es/m/CAFiTN-t7umki=PK8dT1tcPV=mOUe2vNhHML6b3T7W7qqvvajjg@mail.gmail.comDiscussion:http://postgr.es/m/FB0DEA0B-E14E-43A0-811F-C1AE93D00FF3%40amazon.com
1 parent25ddf59 commit3c0ef08

File tree

1 file changed

+19
-5
lines changed
  • src/backend/access/transam

1 file changed

+19
-5
lines changed

‎src/backend/access/transam/xlog.c

Lines changed: 19 additions & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -4321,12 +4321,18 @@ ReadRecord(XLogReaderState *xlogreader, XLogRecPtr RecPtr, int emode,
43214321
if (record==NULL)
43224322
{
43234323
/*
4324-
* When not in standby mode we find that WAL ends in an incomplete
4325-
* record, keep track of that record. After recovery is done,
4326-
* we'll write a record to indicate downstream WAL readers that
4327-
* that portion is to be ignored.
4324+
* When we find that WAL ends in an incomplete record, keep track
4325+
* of that record. After recovery is done, we'll write a record to
4326+
* indicate to downstream WAL readers that that portion is to be
4327+
* ignored.
4328+
*
4329+
* However, when ArchiveRecoveryRequested = true, we're going to
4330+
* switch to a new timeline at the end of recovery. We will only
4331+
* copy WAL over to the new timeline up to the end of the last
4332+
* complete record, so if we did this, we would later create an
4333+
* overwrite contrecord in the wrong place, breaking everything.
43284334
*/
4329-
if (!StandbyMode&&
4335+
if (!ArchiveRecoveryRequested&&
43304336
!XLogRecPtrIsInvalid(xlogreader->abortedRecPtr))
43314337
{
43324338
abortedRecPtr=xlogreader->abortedRecPtr;
@@ -7612,6 +7618,14 @@ StartupXLOG(void)
76127618
*/
76137619
if (!XLogRecPtrIsInvalid(missingContrecPtr))
76147620
{
7621+
/*
7622+
* We should only have a missingContrecPtr if we're not switching to
7623+
* a new timeline. When a timeline switch occurs, WAL is copied from
7624+
* the old timeline to the new only up to the end of the last complete
7625+
* record, so there can't be an incomplete WAL record that we need to
7626+
* disregard.
7627+
*/
7628+
Assert(ThisTimeLineID==PrevTimeLineID);
76157629
Assert(!XLogRecPtrIsInvalid(abortedRecPtr));
76167630
EndOfLog=missingContrecPtr;
76177631
}

0 commit comments

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