Skip to content

Commit a745b93

Browse files
tglsfdctglsfdc
committed
Make pg_ctl stop/restart/promote recheck postmaster aliveness.
"pg_ctl stop/restart" checked that the postmaster PID is valid just once, as a side-effect of sending the stop signal, and then would wait-till-timeout for the postmaster.pid file to go away. This neglects the case wherein the postmaster dies uncleanly after we signal it. Similarly, once "pg_ctl promote" has sent the signal, it'd wait for the corresponding on-disk state change to occur even if the postmaster dies. I'm not sure how we've managed not to notice this problem, but it seems to explain slow execution of the 017_shm.pl test script on AIX since commit 4fdbf9a, which added a speculative "pg_ctl stop" with the idea of making real sure that the postmaster isn't there. In the test steps that kill-9 and then restart the postmaster, it's possible to get past the initial signal attempt before kill() stops working for the doomed postmaster. If that happens, pg_ctl waited till PGCTLTIMEOUT before giving up ... and the buildfarm's AIX members have that set very high. To fix, include a "kill(pid, 0)" test (similar to what postmaster_is_alive uses) in these wait loops, so that we'll give up immediately if the postmaster PID disappears. While here, I chose to refactor those loops out of where they were. do_stop() and do_restart() can perfectly well share one copy of the wait-for-stop loop, and it seems desirable to put a similar function beside that for wait-for-promote. Back-patch to all supported versions, since pg_ctl's wait logic is substantially identical in all, and we're seeing the slow test behavior in all branches. Discussion: https://postgr.es/m/20220210023537.GA3222837@rfd.leadboat.com
1 parent ea09a75 commit a745b93

File tree

1 file changed

+80
-48
lines changed

1 file changed

+80
-48
lines changed

‎src/bin/pg_ctl/pg_ctl.c

Lines changed: 80 additions & 48 deletions
Original file line numberDiff line numberDiff line change
@@ -155,7 +155,9 @@ static void free_readfile(char **optlines);
155155
static pgpid_t start_postmaster(void);
156156
static void read_post_opts(void);
157157

158-
static WaitPMResult wait_for_postmaster(pgpid_t pm_pid, bool do_checkpoint);
158+
static WaitPMResult wait_for_postmaster_start(pgpid_t pm_pid, bool do_checkpoint);
159+
static bool wait_for_postmaster_stop(void);
160+
static bool wait_for_postmaster_promote(void);
159161
static bool postmaster_is_alive(pid_t pid);
160162

161163
#if defined(HAVE_GETRLIMIT) && defined(RLIMIT_CORE)
@@ -594,7 +596,7 @@ start_postmaster(void)
594596
* manager checkpoint, it's got nothing to do with database checkpoints!!
595597
*/
596598
static WaitPMResult
597-
wait_for_postmaster(pgpid_t pm_pid, bool do_checkpoint)
599+
wait_for_postmaster_start(pgpid_t pm_pid, bool do_checkpoint)
598600
{
599601
int i;
600602

@@ -704,6 +706,76 @@ wait_for_postmaster(pgpid_t pm_pid, bool do_checkpoint)
704706
}
705707

706708

709+
/*
710+
* Wait for the postmaster to stop.
711+
*
712+
* Returns true if the postmaster stopped cleanly (i.e., removed its pidfile).
713+
* Returns false if the postmaster dies uncleanly, or if we time out.
714+
*/
715+
static bool
716+
wait_for_postmaster_stop(void)
717+
{
718+
int cnt;
719+
720+
for (cnt = 0; cnt < wait_seconds * WAITS_PER_SEC; cnt++)
721+
{
722+
pgpid_t pid;
723+
724+
if ((pid = get_pgpid(false)) == 0)
725+
return true; /* pid file is gone */
726+
727+
if (kill((pid_t) pid, 0) != 0)
728+
{
729+
/*
730+
* Postmaster seems to have died. Check the pid file once more to
731+
* avoid a race condition, but give up waiting.
732+
*/
733+
if (get_pgpid(false) == 0)
734+
return true; /* pid file is gone */
735+
return false; /* postmaster died untimely */
736+
}
737+
738+
if (cnt % WAITS_PER_SEC == 0)
739+
print_msg(".");
740+
pg_usleep(USEC_PER_SEC / WAITS_PER_SEC);
741+
}
742+
return false; /* timeout reached */
743+
}
744+
745+
746+
/*
747+
* Wait for the postmaster to promote.
748+
*
749+
* Returns true on success, else false.
750+
* To avoid waiting uselessly, we check for postmaster death here too.
751+
*/
752+
static bool
753+
wait_for_postmaster_promote(void)
754+
{
755+
int cnt;
756+
757+
for (cnt = 0; cnt < wait_seconds * WAITS_PER_SEC; cnt++)
758+
{
759+
pgpid_t pid;
760+
DBState state;
761+
762+
if ((pid = get_pgpid(false)) == 0)
763+
return false; /* pid file is gone */
764+
if (kill((pid_t) pid, 0) != 0)
765+
return false; /* postmaster died */
766+
767+
state = get_control_dbstate();
768+
if (state == DB_IN_PRODUCTION)
769+
return true; /* successful promotion */
770+
771+
if (cnt % WAITS_PER_SEC == 0)
772+
print_msg(".");
773+
pg_usleep(USEC_PER_SEC / WAITS_PER_SEC);
774+
}
775+
return false; /* timeout reached */
776+
}
777+
778+
707779
#if defined(HAVE_GETRLIMIT) && defined(RLIMIT_CORE)
708780
static void
709781
unlimit_core_size(void)
@@ -917,7 +989,7 @@ do_start(void)
917989

918990
print_msg(_("waiting for server to start..."));
919991

920-
switch (wait_for_postmaster(pm_pid, false))
992+
switch (wait_for_postmaster_start(pm_pid, false))
921993
{
922994
case POSTMASTER_READY:
923995
print_msg(_(" done\n"));
@@ -952,7 +1024,6 @@ do_start(void)
9521024
static void
9531025
do_stop(void)
9541026
{
955-
int cnt;
9561027
pgpid_t pid;
9571028
struct stat statbuf;
9581029

@@ -1003,19 +1074,7 @@ do_stop(void)
10031074

10041075
print_msg(_("waiting for server to shut down..."));
10051076

1006-
for (cnt = 0; cnt < wait_seconds * WAITS_PER_SEC; cnt++)
1007-
{
1008-
if ((pid = get_pgpid(false)) != 0)
1009-
{
1010-
if (cnt % WAITS_PER_SEC == 0)
1011-
print_msg(".");
1012-
pg_usleep(USEC_PER_SEC / WAITS_PER_SEC);
1013-
}
1014-
else
1015-
break;
1016-
}
1017-
1018-
if (pid != 0) /* pid file still exists */
1077+
if (!wait_for_postmaster_stop())
10191078
{
10201079
print_msg(_(" failed\n"));
10211080

@@ -1039,7 +1098,6 @@ do_stop(void)
10391098
static void
10401099
do_restart(void)
10411100
{
1042-
int cnt;
10431101
pgpid_t pid;
10441102
struct stat statbuf;
10451103

@@ -1093,20 +1151,7 @@ do_restart(void)
10931151
print_msg(_("waiting for server to shut down..."));
10941152

10951153
/* always wait for restart */
1096-
1097-
for (cnt = 0; cnt < wait_seconds * WAITS_PER_SEC; cnt++)
1098-
{
1099-
if ((pid = get_pgpid(false)) != 0)
1100-
{
1101-
if (cnt % WAITS_PER_SEC == 0)
1102-
print_msg(".");
1103-
pg_usleep(USEC_PER_SEC / WAITS_PER_SEC);
1104-
}
1105-
else
1106-
break;
1107-
}
1108-
1109-
if (pid != 0) /* pid file still exists */
1154+
if (!wait_for_postmaster_stop())
11101155
{
11111156
print_msg(_(" failed\n"));
11121157

@@ -1226,21 +1271,8 @@ do_promote(void)
12261271

12271272
if (do_wait)
12281273
{
1229-
DBState state = DB_STARTUP;
1230-
int cnt;
1231-
12321274
print_msg(_("waiting for server to promote..."));
1233-
for (cnt = 0; cnt < wait_seconds * WAITS_PER_SEC; cnt++)
1234-
{
1235-
state = get_control_dbstate();
1236-
if (state == DB_IN_PRODUCTION)
1237-
break;
1238-
1239-
if (cnt % WAITS_PER_SEC == 0)
1240-
print_msg(".");
1241-
pg_usleep(USEC_PER_SEC / WAITS_PER_SEC);
1242-
}
1243-
if (state == DB_IN_PRODUCTION)
1275+
if (wait_for_postmaster_promote())
12441276
{
12451277
print_msg(_(" done\n"));
12461278
print_msg(_("server promoted\n"));
@@ -1659,7 +1691,7 @@ pgwin32_ServiceMain(DWORD argc, LPTSTR *argv)
16591691
if (do_wait)
16601692
{
16611693
write_eventlog(EVENTLOG_INFORMATION_TYPE, _("Waiting for server startup...\n"));
1662-
if (wait_for_postmaster(postmasterPID, true) != POSTMASTER_READY)
1694+
if (wait_for_postmaster_start(postmasterPID, true) != POSTMASTER_READY)
16631695
{
16641696
write_eventlog(EVENTLOG_ERROR_TYPE, _("Timed out waiting for server startup\n"));
16651697
pgwin32_SetServiceStatus(SERVICE_STOPPED);
@@ -1680,7 +1712,7 @@ pgwin32_ServiceMain(DWORD argc, LPTSTR *argv)
16801712
{
16811713
/*
16821714
* status.dwCheckPoint can be incremented by
1683-
* wait_for_postmaster(), so it might not start from 0.
1715+
* wait_for_postmaster_start(), so it might not start from 0.
16841716
*/
16851717
int maxShutdownCheckPoint = status.dwCheckPoint + 12;
16861718

0 commit comments

Comments
 (0)