8.1.2 postmaster died
One of our 8.1.2 postmasters on HPUX 11.23 ia64 just received a
SIGKILL signal from unknown origins. After reviewing all
command history files for the DBA and root, I do not believe
anyone manually sent it, and we have no scripts etc that would
do that, at least that we can find or imagine. The machine had
ample RAM available.
Any ideas where could this SIGKILL would have come from?
Thanks,
Ed
Hi Ed,
You need to look through the log files for details.
Regards
Talha Khan
Show quoted text
On 11/8/06, Ed L. <pgsql@bluepolka.net> wrote:
One of our 8.1.2 postmasters on HPUX 11.23 ia64 just received a
SIGKILL signal from unknown origins. After reviewing all
command history files for the DBA and root, I do not believe
anyone manually sent it, and we have no scripts etc that would
do that, at least that we can find or imagine. The machine had
ample RAM available.Any ideas where could this SIGKILL would have come from?
Thanks,
Ed---------------------------(end of broadcast)---------------------------
TIP 4: Have you searched our list archives?
To be specific you need to look at the syslogs to see if it was actually
some OOM killer or not.
Thanks,
-------
Shoaib Mir
EnterpriseDB (www.enterprisedb.com)
Show quoted text
On 11/9/06, Talha Khan <talha.amjad@gmail.com> wrote:
Hi Ed,
You need to look through the log files for details.
Regards
Talha KhanOn 11/8/06, Ed L. < pgsql@bluepolka.net> wrote:
One of our 8.1.2 postmasters on HPUX 11.23 ia64 just received a
SIGKILL signal from unknown origins. After reviewing all
command history files for the DBA and root, I do not believe
anyone manually sent it, and we have no scripts etc that would
do that, at least that we can find or imagine. The machine had
ample RAM available.Any ideas where could this SIGKILL would have come from?
Thanks,
Ed---------------------------(end of broadcast)---------------------------
TIP 4: Have you searched our list archives?
On Wednesday November 8 2006 12:30 pm, Shoaib Mir wrote:
To be specific you need to look at the syslogs to see if it
was actually some OOM killer or not.
I checked all db logs + syslog, of course. All they show is the
process receiving SIGKILL (followed by all others being shutdown
as a result).
Is there any circumstance at all under which pgsql itself would
issue a sigkill to one of the backends?
Ed
Hi Ed,
I don't think its pgsql thats sending the SIGKILL its being done by the OS.
The reason can only be confirmed after looking at your log files but the
probability is that your OS ran out of memory and in order to protect itself
it had to kill the processes that were taking a lot of memory thus ended up
killing pgsql processes. Its just a hunch that i have but the reason can
only be confirmed after seeing your logs.
Regards
Talha Khan
Show quoted text
On 11/9/06, Ed Loehr <ed@loehrtech.com> wrote:
On Wednesday November 8 2006 12:30 pm, Shoaib Mir wrote:
To be specific you need to look at the syslogs to see if it
was actually some OOM killer or not.I checked all db logs + syslog, of course. All they show is the
process receiving SIGKILL (followed by all others being shutdown
as a result).Is there any circumstance at all under which pgsql itself would
issue a sigkill to one of the backends?Ed
Have a look at "16.4.3. Linux Memory Overcommit" on
http://www.postgresql.org/docs/8.1/interactive/kernel-resources.html
Thanks,
------
Shoaib Mir
EnterpriseDB (www.enterprisedb.com)
Show quoted text
On 11/9/06, Talha Khan <talha.amjad@gmail.com> wrote:
Hi Ed,
I don't think its pgsql thats sending the SIGKILL its being done by the
OS. The reason can only be confirmed after looking at your log files but the
probability is that your OS ran out of memory and in order to protect itself
it had to kill the processes that were taking a lot of memory thus ended up
killing pgsql processes. Its just a hunch that i have but the reason can
only be confirmed after seeing your logs.Regards
Talha KhanOn 11/9/06, Ed Loehr <ed@loehrtech.com> wrote:
On Wednesday November 8 2006 12:30 pm, Shoaib Mir wrote:
To be specific you need to look at the syslogs to see if it
was actually some OOM killer or not.I checked all db logs + syslog, of course. All they show is the
process receiving SIGKILL (followed by all others being shutdown
as a result).Is there any circumstance at all under which pgsql itself would
issue a sigkill to one of the backends?Ed
On Thu, Nov 09, 2006 at 11:52:48AM +0500, Shoaib Mir wrote:
Have a look at "16.4.3. Linux Memory Overcommit" on
http://www.postgresql.org/docs/8.1/interactive/kernel-resources.html
If you look at the beginning of the thread you'dve seen we're talking
about HPUX here...
Maybe they have an OOM killer too? Although maybe it's quotas?
Have a nice day,
--
Martijn van Oosterhout <kleptog@svana.org> http://svana.org/kleptog/
Show quoted text
From each according to his ability. To each according to his ability to litigate.
Look at the database server logs and see what actually was happening just
before the server crashed... do you have the auto vacuuming running at
backend? or was there some client making connection just when the server did
crash? .... your db logs can actually help you here.
Thank you,
-----------
Shoaib Mir
EnterpriseDB (www.enterprisedb.com)
Show quoted text
On 11/9/06, Martijn van Oosterhout <kleptog@svana.org> wrote:
On Thu, Nov 09, 2006 at 11:52:48AM +0500, Shoaib Mir wrote:
Have a look at "16.4.3. Linux Memory Overcommit" on
http://www.postgresql.org/docs/8.1/interactive/kernel-resources.htmlIf you look at the beginning of the thread you'dve seen we're talking
about HPUX here...Maybe they have an OOM killer too? Although maybe it's quotas?
Have a nice day,
--
Martijn van Oosterhout <kleptog@svana.org> http://svana.org/kleptog/From each according to his ability. To each according to his ability to
litigate.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (GNU/Linux)iD8DBQFFU0RoIB7bNG8LQkwRAjRIAJwJig0SLNyUtfthMw0rPM5olfXuVwCfYoX5
OiFfQVVyp55mDo4MDfD88v8=
=7Enb
-----END PGP SIGNATURE-----
