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

Commit985bd7d

Browse files
committed
Support clean switchover.
In replication, when we shutdown the master, walsender tries to sendall the outstanding WAL records to the standby, and then to exit. Thisbasically means that all the WAL records are fully synced betweentwo servers after the clean shutdown of the master. So, afterpromoting the standby to new master, we can restart the stoppedmaster as new standby without the need for a fresh backup fromnew master.But there was one problem so far: though walsender tries to send allthe outstanding WAL records, it doesn't wait for them to be replicatedto the standby. Then, before receiving all the WAL records,walreceiver can detect the closure of connection and exit. We cannotguarantee that there is no missing WAL in the standby after cleanshutdown of the master. In this case, backup from new master isrequired when restarting the stopped master as new standby.This patch fixes this problem. It just changes walsender so that itwaits for all the outstanding WAL records to be replicated to thestandby before closing the replication connection.Per discussion, this is a fix that needs to get backpatched rather thannew feature. So, back-patch to 9.1 where enough infrastructure forthis exists.Patch by me, reviewed by Andres Freund.
1 parent4f14c86 commit985bd7d

File tree

1 file changed

+8
-4
lines changed

1 file changed

+8
-4
lines changed

‎src/backend/replication/walsender.c

Lines changed: 8 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -27,7 +27,8 @@
2727
* If the server is shut down, postmaster sends us SIGUSR2 after all
2828
* regular backends have exited and the shutdown checkpoint has been written.
2929
* This instruct walsender to send any outstanding WAL, including the
30-
* shutdown checkpoint record, and then exit.
30+
* shutdown checkpoint record, wait for it to be replicated to the standby,
31+
* and then exit.
3132
*
3233
*
3334
* Portions Copyright (c) 2010-2013, PostgreSQL Global Development Group
@@ -1045,15 +1046,17 @@ WalSndLoop(void)
10451046

10461047
/*
10471048
* When SIGUSR2 arrives, we send any outstanding logs up to the
1048-
* shutdown checkpoint record (i.e., the latest record) and exit.
1049+
* shutdown checkpoint record (i.e., the latest record), wait
1050+
* for them to be replicated to the standby, and exit.
10491051
* This may be a normal termination at shutdown, or a promotion,
10501052
* the walsender is not sure which.
10511053
*/
10521054
if (walsender_ready_to_stop)
10531055
{
10541056
/* ... let's just be real sure we're caught up ... */
10551057
XLogSend(&caughtup);
1056-
if (caughtup&& !pq_is_send_pending())
1058+
if (caughtup&&sentPtr==MyWalSnd->flush&&
1059+
!pq_is_send_pending())
10571060
{
10581061
/* Inform the standby that XLOG streaming is done */
10591062
EndCommand("COPY 0",DestRemote);
@@ -1728,7 +1731,8 @@ WalSndLastCycleHandler(SIGNAL_ARGS)
17281731
/*
17291732
* If replication has not yet started, die like with SIGTERM. If
17301733
* replication is active, only set a flag and wake up the main loop. It
1731-
* will send any outstanding WAL, and then exit gracefully.
1734+
* will send any outstanding WAL, wait for it to be replicated to
1735+
* the standby, and then exit gracefully.
17321736
*/
17331737
if (!replication_active)
17341738
kill(MyProcPid,SIGTERM);

0 commit comments

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