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

Commit9e403c2

Browse files
committed
Add recovery_end_command option to recovery.conf. recovery_end_command
is run at the end of archive recovery, providing a chance to do externalcleanup. Modify pg_standby so that it no longer removes the trigger file,that is to be done using the recovery_end_command now.Provide a "smart" failover mode in pg_standby, where we don't fail overimmediately, but only after recovering all unapplied WAL from the archive.That gives you zero data loss assuming all WAL was archived beforefailover, which is what most users of pg_standby actually want.recovery_end_command by Simon Riggs, pg_standby changes by Fujii Masao andmyself.
1 parenta710713 commit9e403c2

File tree

4 files changed

+351
-76
lines changed

4 files changed

+351
-76
lines changed

‎contrib/pg_standby/pg_standby.c

Lines changed: 162 additions & 68 deletions
Original file line numberDiff line numberDiff line change
@@ -1,5 +1,5 @@
11
/*
2-
* $PostgreSQL: pgsql/contrib/pg_standby/pg_standby.c,v 1.21 2009/03/26 22:29:13 tgl Exp $
2+
* $PostgreSQL: pgsql/contrib/pg_standby/pg_standby.c,v 1.22 2009/05/14 20:31:09 heikki Exp $
33
*
44
*
55
* pg_standby.c
@@ -26,6 +26,7 @@
2626
#include<ctype.h>
2727
#include<dirent.h>
2828
#include<sys/stat.h>
29+
#include<fcntl.h>
2930
#include<signal.h>
3031

3132
#ifdefWIN32
@@ -52,7 +53,6 @@ intmaxwaittime = 0;/* how long are we prepared to wait for? */
5253
intkeepfiles=0;/* number of WAL files to keep, 0 keep all */
5354
intmaxretries=3;/* number of retries on restore command */
5455
booldebug= false;/* are we debugging? */
55-
booltriggered= false;/* have we been triggered? */
5656
boolneed_cleanup= false;/* do we need to remove files from
5757
* archive? */
5858

@@ -69,6 +69,30 @@ charrestoreCommand[MAXPGPATH];/* run this to restore */
6969
charexclusiveCleanupFileName[MAXPGPATH];/* the file we need to
7070
* get from archive */
7171

72+
/*
73+
* Two types of failover are supported (smart and fast failover).
74+
*
75+
* The content of the trigger file determines the type of failover. If the
76+
* trigger file contains the word "smart" (or the file is empty), smart
77+
* failover is chosen: pg_standby acts as cp or ln command itself, on
78+
* successful completion all the available WAL records will be applied
79+
* resulting in zero data loss. But, it might take a long time to finish
80+
* recovery if there's a lot of unapplied WAL.
81+
*
82+
* On the other hand, if the trigger file contains the word "fast", the
83+
* recovery is finished immediately even if unapplied WAL files remain. Any
84+
* transactions in the unapplied WAL files are lost.
85+
*
86+
* An empty trigger file performs smart failover. SIGUSR or SIGINT triggers
87+
* fast failover. A timeout causes fast failover (smart failover would have
88+
* the same effect, since if the timeout is reached there is no unapplied WAL).
89+
*/
90+
#defineNoFailover0
91+
#defineSmartFailover1
92+
#defineFastFailover2
93+
94+
staticintFailover=NoFailover;
95+
7296
#defineRESTORE_COMMAND_COPY 0
7397
#defineRESTORE_COMMAND_LINK 1
7498
intrestoreCommandType;
@@ -108,7 +132,6 @@ struct stat stat_buf;
108132
*
109133
*As an example, and probably the common case, we use either
110134
*cp/ln commands on *nix, or copy/move command on Windows.
111-
*
112135
*/
113136
staticvoid
114137
CustomizableInitialize(void)
@@ -352,41 +375,96 @@ SetWALFileNameForCleanup(void)
352375
/*
353376
* CheckForExternalTrigger()
354377
*
355-
* Is there a trigger file?
378+
* Is there a trigger file? Sets global 'Failover' variable to indicate
379+
* what kind of a trigger file it was. A "fast" trigger file is turned
380+
* into a "smart" file as a side-effect.
356381
*/
357-
staticbool
382+
staticvoid
358383
CheckForExternalTrigger(void)
359384
{
360-
intrc;
385+
charbuf[32];
386+
intfd;
387+
intlen;
361388

362389
/*
363390
* Look for a trigger file, if that option has been selected
364391
*
365392
* We use stat() here because triggerPath is always a file rather than
366393
* potentially being in an archive
367394
*/
368-
if (triggerPath&&stat(triggerPath,&stat_buf)==0)
395+
if (!triggerPath||stat(triggerPath,&stat_buf)!=0)
396+
return;
397+
398+
/*
399+
* An empty trigger file performs smart failover. There's a little race
400+
* condition here: if the writer of the trigger file has just created
401+
* the file, but not yet written anything to it, we'll treat that as
402+
* smart shutdown even if the other process was just about to write "fast"
403+
* to it. But that's fine: we'll restore one more WAL file, and when we're
404+
* invoked next time, we'll see the word "fast" and fail over immediately.
405+
*/
406+
if (stat_buf.st_size==0)
369407
{
370-
fprintf(stderr,"trigger file found\n");
408+
Failover=SmartFailover;
409+
fprintf(stderr,"trigger file found: smart failover\n");
410+
fflush(stderr);
411+
return;
412+
}
413+
414+
if ((fd=open(triggerPath,O_RDWR,0))<0)
415+
{
416+
fprintf(stderr,"WARNING: could not open \"%s\": %s\n",
417+
triggerPath,strerror(errno));
418+
fflush(stderr);
419+
return;
420+
}
421+
422+
if ((len=read(fd,buf,sizeof(buf)))<0)
423+
{
424+
fprintf(stderr,"WARNING: could not read \"%s\": %s\n",
425+
triggerPath,strerror(errno));
426+
fflush(stderr);
427+
close(fd);
428+
return;
429+
}
430+
buf[len]='\0';
431+
432+
if (strncmp(buf,"smart",5)==0)
433+
{
434+
Failover=SmartFailover;
435+
fprintf(stderr,"trigger file found: smart failover\n");
436+
fflush(stderr);
437+
close(fd);
438+
return;
439+
}
440+
441+
if (strncmp(buf,"fast",4)==0)
442+
{
443+
Failover=FastFailover;
444+
445+
fprintf(stderr,"trigger file found: fast failover\n");
371446
fflush(stderr);
372447

373448
/*
374-
* If trigger file found, we *must* delete it. Here's why: When
375-
* recovery completes, we will be asked again for the same file from
376-
* the archive using pg_standby so must remove trigger file so we can
377-
* reload file again and come up correctly.
449+
* Turn it into a "smart" trigger by truncating the file. Otherwise
450+
* if the server asks us again to restore a segment that was restored
451+
* restored already, we would return "not found" and upset the server.
378452
*/
379-
rc=unlink(triggerPath);
380-
if (rc!=0)
453+
if (ftruncate(fd,0)<0)
381454
{
382-
fprintf(stderr,"\n ERROR: could not remove \"%s\": %s",triggerPath,strerror(errno));
455+
fprintf(stderr,"WARNING: could not read \"%s\": %s\n",
456+
triggerPath,strerror(errno));
383457
fflush(stderr);
384-
exit(rc);
385458
}
386-
return true;
387-
}
459+
close(fd);
388460

389-
return false;
461+
return;
462+
}
463+
close(fd);
464+
465+
fprintf(stderr,"WARNING: invalid content in \"%s\"\n",triggerPath);
466+
fflush(stderr);
467+
return;
390468
}
391469

392470
/*
@@ -402,7 +480,7 @@ RestoreWALFileForRecovery(void)
402480

403481
if (debug)
404482
{
405-
fprintf(stderr,"\nrunning restore:");
483+
fprintf(stderr,"running restore:");
406484
fflush(stderr);
407485
}
408486

@@ -413,7 +491,7 @@ RestoreWALFileForRecovery(void)
413491
{
414492
if (debug)
415493
{
416-
fprintf(stderr," OK");
494+
fprintf(stderr," OK\n");
417495
fflush(stderr);
418496
}
419497
return true;
@@ -425,7 +503,7 @@ RestoreWALFileForRecovery(void)
425503
* Allow caller to add additional info
426504
*/
427505
if (debug)
428-
fprintf(stderr,"not restored:");
506+
fprintf(stderr,"not restored\n");
429507
return false;
430508
}
431509

@@ -552,8 +630,6 @@ main(int argc, char **argv)
552630
break;
553631
case't':/* Trigger file */
554632
triggerPath=optarg;
555-
if (CheckForExternalTrigger())
556-
exit(1);/* Normal exit, with non-zero */
557633
break;
558634
case'w':/* Max wait time */
559635
maxwaittime=atoi(optarg);
@@ -633,20 +709,20 @@ main(int argc, char **argv)
633709

634710
if (debug)
635711
{
636-
fprintf(stderr,"\nTrigger file : %s",triggerPath ?triggerPath :"<not set>");
637-
fprintf(stderr,"\nWaiting for WAL file: %s",nextWALFileName);
638-
fprintf(stderr,"\nWAL file path: %s",WALFilePath);
639-
fprintf(stderr,"\nRestoring to...: %s",xlogFilePath);
640-
fprintf(stderr,"\nSleep interval: %d second%s",
712+
fprintf(stderr,"Trigger file : %s\n",triggerPath ?triggerPath :"<not set>");
713+
fprintf(stderr,"Waiting for WAL file: %s\n",nextWALFileName);
714+
fprintf(stderr,"WAL file path: %s\n",WALFilePath);
715+
fprintf(stderr,"Restoring to: %s\n",xlogFilePath);
716+
fprintf(stderr,"Sleep interval: %d second%s\n",
641717
sleeptime, (sleeptime>1 ?"s" :" "));
642-
fprintf(stderr,"\nMax wait interval: %d %s",
718+
fprintf(stderr,"Max wait interval: %d %s\n",
643719
maxwaittime, (maxwaittime>0 ?"seconds" :"forever"));
644-
fprintf(stderr,"\nCommand for restore: %s",restoreCommand);
645-
fprintf(stderr,"\nKeep archive history: ");
720+
fprintf(stderr,"Command for restore: %s\n",restoreCommand);
721+
fprintf(stderr,"Keep archive history: ");
646722
if (need_cleanup)
647-
fprintf(stderr,"%s and later",exclusiveCleanupFileName);
723+
fprintf(stderr,"%s and later\n",exclusiveCleanupFileName);
648724
else
649-
fprintf(stderr,"No cleanup required");
725+
fprintf(stderr,"No cleanup required\n");
650726
fflush(stderr);
651727
}
652728

@@ -676,56 +752,74 @@ main(int argc, char **argv)
676752
/*
677753
* Main wait loop
678754
*/
679-
while (!CustomizableNextWALFileReady()&& !triggered)
755+
for (;;)
680756
{
681-
if (sleeptime <=60)
682-
pg_usleep(sleeptime*1000000L);
683-
757+
/* Check for trigger file or signal first */
758+
CheckForExternalTrigger();
684759
if (signaled)
685760
{
686-
triggered=true;
761+
Failover=FastFailover;
687762
if (debug)
688763
{
689-
fprintf(stderr,"\nsignaled to exit\n");
764+
fprintf(stderr,"signaled to exit: fast failover\n");
690765
fflush(stderr);
691766
}
692767
}
693-
else
768+
769+
/*
770+
* Check for fast failover immediately, before checking if the
771+
* requested WAL file is available
772+
*/
773+
if (Failover==FastFailover)
774+
exit(1);
775+
776+
if (CustomizableNextWALFileReady())
694777
{
778+
/*
779+
* Once we have restored this file successfully we can remove some
780+
* prior WAL files. If this restore fails we musn't remove any file
781+
* because some of them will be requested again immediately after
782+
* the failed restore, or when we restart recovery.
783+
*/
784+
if (RestoreWALFileForRecovery())
785+
{
786+
if (need_cleanup)
787+
CustomizableCleanupPriorWALFiles();
695788

696-
if (debug)
789+
exit(0);
790+
}
791+
else
697792
{
698-
fprintf(stderr,"\nWAL file not present yet.");
699-
if (triggerPath)
700-
fprintf(stderr," Checking for trigger file...");
701-
fflush(stderr);
793+
/* Something went wrong in copying the file */
794+
exit(1);
702795
}
796+
}
797+
798+
/* Check for smart failover if the next WAL file was not available */
799+
if (Failover==SmartFailover)
800+
exit(1);
703801

704-
waittime+=sleeptime;
802+
if (sleeptime <=60)
803+
pg_usleep(sleeptime*1000000L);
705804

706-
if (!triggered&& (CheckForExternalTrigger()|| (waittime >=maxwaittime&&maxwaittime>0)))
805+
waittime+=sleeptime;
806+
if (waittime >=maxwaittime&&maxwaittime>0)
807+
{
808+
Failover=FastFailover;
809+
if (debug)
707810
{
708-
triggered= true;
709-
if (debug&&waittime >=maxwaittime&&maxwaittime>0)
710-
fprintf(stderr,"\nTimed out after %d seconds\n",waittime);
811+
fprintf(stderr,"Timed out after %d seconds: fast failover\n",
812+
waittime);
813+
fflush(stderr);
711814
}
712815
}
816+
if (debug)
817+
{
818+
fprintf(stderr,"WAL file not present yet.");
819+
if (triggerPath)
820+
fprintf(stderr," Checking for trigger file...");
821+
fprintf(stderr,"\n");
822+
fflush(stderr);
823+
}
713824
}
714-
715-
/*
716-
* Action on exit
717-
*/
718-
if (triggered)
719-
exit(1);/* Normal exit, with non-zero */
720-
721-
/*
722-
* Once we have restored this file successfully we can remove some prior
723-
* WAL files. If this restore fails we musn't remove any file because some
724-
* of them will be requested again immediately after the failed restore,
725-
* or when we restart recovery.
726-
*/
727-
if (RestoreWALFileForRecovery()&&need_cleanup)
728-
CustomizableCleanupPriorWALFiles();
729-
730-
return0;
731825
}

‎doc/src/sgml/backup.sgml

Lines changed: 24 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1,4 +1,4 @@
1-
<!-- $PostgreSQL: pgsql/doc/src/sgml/backup.sgml,v 2.125 2009/04/27 16:27:35 momjian Exp $ -->
1+
<!-- $PostgreSQL: pgsql/doc/src/sgml/backup.sgml,v 2.126 2009/05/14 20:31:09 heikki Exp $ -->
22

33
<chapter id="backup">
44
<title>Backup and Restore</title>
@@ -1126,6 +1126,29 @@ restore_command = 'copy "C:\\server\\archivedir\\%f" "%p"' # Windows
11261126
</listitem>
11271127
</varlistentry>
11281128

1129+
<varlistentry id="recovery-end-command" xreflabel="recovery_end_command">
1130+
<term><varname>recovery_end_command</varname> (<type>string</type>)</term>
1131+
<listitem>
1132+
<para>
1133+
This parameter specifies a shell command that will be executed once only
1134+
at the end of recovery. This parameter is optional. The purpose of the
1135+
recovery_end_command is to provide a mechanism for cleanup following
1136+
replication or recovery.
1137+
Any <literal>%r</> is replaced by the name of the file
1138+
containing the last valid restart point. That is the earliest file that
1139+
must be kept to allow a restore to be restartable, so this information
1140+
can be used to truncate the archive to just the minimum required to
1141+
support restart of the current restore. <literal>%r</> would only be
1142+
used in a warm-standby configuration (see <xref linkend="warm-standby">).
1143+
Write <literal>%%</> to embed an actual <literal>%</> character
1144+
in the command.
1145+
If the command returns a non-zero exit status then a WARNING log
1146+
message will be written, unless signalled in which case we return
1147+
a FATAL error.
1148+
</para>
1149+
</listitem>
1150+
</varlistentry>
1151+
11291152
<varlistentry id="recovery-target-time" xreflabel="recovery_target_time">
11301153
<term><varname>recovery_target_time</varname>
11311154
(<type>timestamp</type>)

0 commit comments

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