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

Commita1b8b3b

Browse files
committed
Fix unconditional WAL receiver shutdown during stream-archive transition
Commitb4f584f (affecting v15~, later backpatched down to 13 as of3635a0a) introduced an unconditional WAL receiver shutdown whenswitching from streaming to archive WAL sources. This causes problemsduring a timeline switch, when a WAL receiver enters WALRCV_WAITINGstate but remains alive, waiting for instructions.The unconditional shutdown can break some monitoring scenarios as theWAL receiver gets repeatedly terminated and re-spawned, causingpg_stat_wal_receiver.status to show a "streaming" instead of "waiting"status, masking the fact that the WAL receiver is waiting for a new TLIand a new LSN to be able to continue streaming.This commit changes the WAL receiver behavior so as the shutdown becomesconditional, with InstallXLogFileSegmentActive being always reset toprevent the regression fixed byb4f584f: only terminate the WALreceiver when it is actively streaming (WALRCV_STREAMING,WALRCV_STARTING, or WALRCV_RESTARTING). When in WALRCV_WAITING state,just reset InstallXLogFileSegmentActive flag to allow archiverestoration without killing the process. WALRCV_STOPPED andWALRCV_STOPPING are not reachable states in this code path. For thelatter, the startup process is the one in charge of settingWALRCV_STOPPING via ShutdownWalRcv(), waiting for the WAL receiver toreach a WALRCV_STOPPED state after switching walRcvState, soWaitForWALToBecomeAvailable() cannot be reached while a WAL receiver isin a WALRCV_STOPPING state.A regression test is added to check that a WAL receiver is not stoppedon timeline jump, that fails when the fix of this commit is reverted.Reported-by: Ryan Bird <ryanzxg@gmail.com>Author: Xuneng Zhou <xunengzhou@gmail.com>Reviewed-by: Noah Misch <noah@leadboat.com>Reviewed-by: Michael Paquier <michael@paquier.xyz>Discussion:https://postgr.es/m/19093-c4fff49a608f82a0@postgresql.orgBackpatch-through: 13
1 parentfaba259 commita1b8b3b

File tree

2 files changed

+31
-6
lines changed

2 files changed

+31
-6
lines changed

‎src/backend/access/transam/xlog.c‎

Lines changed: 22 additions & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -946,6 +946,7 @@ static intXLogPageRead(XLogReaderState *xlogreader, XLogRecPtr targetPagePtr,
946946
intreqLen,XLogRecPtrtargetRecPtr,char*readBuf);
947947
staticboolWaitForWALToBecomeAvailable(XLogRecPtrRecPtr,boolrandAccess,
948948
boolfetching_ckpt,XLogRecPtrtliRecPtr);
949+
staticvoidResetInstallXLogFileSegmentActive(void);
949950
staticvoidXLogShutdownWalRcv(void);
950951
staticintemode_for_corrupt_record(intemode,XLogRecPtrRecPtr);
951952
staticvoidXLogFileClose(void);
@@ -12837,8 +12838,18 @@ WaitForWALToBecomeAvailable(XLogRecPtr RecPtr, bool randAccess,
1283712838
* Before we leave XLOG_FROM_STREAM state, make sure that
1283812839
* walreceiver is not active, so that it won't overwrite
1283912840
* WAL that we restore from archive.
12841+
* If walreceiver is actively streaming (or attempting to
12842+
* connect), we must shut it down. However, if it's
12843+
* already in WAITING state (e.g., due to timeline
12844+
* divergence), we only need to reset the install flag to
12845+
* allow archive restoration.
1284012846
*/
12841-
XLogShutdownWalRcv();
12847+
if (WalRcvStreaming())
12848+
XLogShutdownWalRcv();
12849+
else
12850+
{
12851+
ResetInstallXLogFileSegmentActive();
12852+
}
1284212853

1284312854
/*
1284412855
* Before we sleep, re-scan for possible new timelines if
@@ -13191,17 +13202,23 @@ StartupRequestWalReceiverRestart(void)
1319113202
}
1319213203
}
1319313204

13194-
/*Thin wrapper around ShutdownWalRcv(). */
13205+
/*Disable WAL file recycling and preallocation. */
1319513206
staticvoid
13196-
XLogShutdownWalRcv(void)
13207+
ResetInstallXLogFileSegmentActive(void)
1319713208
{
13198-
ShutdownWalRcv();
13199-
1320013209
LWLockAcquire(ControlFileLock,LW_EXCLUSIVE);
1320113210
XLogCtl->InstallXLogFileSegmentActive= false;
1320213211
LWLockRelease(ControlFileLock);
1320313212
}
1320413213

13214+
/* Thin wrapper around ShutdownWalRcv(). */
13215+
staticvoid
13216+
XLogShutdownWalRcv(void)
13217+
{
13218+
ShutdownWalRcv();
13219+
ResetInstallXLogFileSegmentActive();
13220+
}
13221+
1320513222
/*
1320613223
* Determine what log level should be used to report a corrupt WAL record
1320713224
* in the current WAL page, previously read by XLogPageRead().

‎src/test/recovery/t/004_timeline_switch.pl‎

Lines changed: 9 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -7,7 +7,7 @@
77
use File::Pathqw(rmtree);
88
use PostgresNode;
99
use TestLib;
10-
use Test::Moretests=>3;
10+
use Test::Moretests=>4;
1111

1212
$ENV{PGDATABASE} ='postgres';
1313

@@ -71,6 +71,14 @@
7171
$node_standby_2->safe_psql('postgres',"SELECT count(*) FROM tab_int");
7272
is($result,qq(2000),'check content of standby 2');
7373

74+
# Check the logs, WAL receiver should not have been stopped while
75+
# transitioning to its new timeline. There is no need to rely on an
76+
# offset in this check of the server logs: a new log file is used on
77+
# node restart when primary_conninfo is updated above.
78+
ok( !$node_standby_2->log_contains(
79+
"FATAL: .* terminating walreceiver process due to administrator command"
80+
),
81+
'WAL receiver should not be stopped across timeline jumps');
7482

7583
# Ensure that a standby is able to follow a primary on a newer timeline
7684
# when WAL archiving is enabled.

0 commit comments

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