Re: BUG #2225: Backend crash -- BIG table

From: Patrick Rotsaert <patrick(dot)rotsaert(at)arrowup(dot)be>
To: Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
Cc: pgsql-bugs(at)postgresql(dot)org
Subject: Re: BUG #2225: Backend crash -- BIG table
Date: 2006-01-31 15:47:12
Message-ID: 43DF8680.7020109@arrowup.be
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-bugs

on 31/01/2006 16:18 Tom Lane wrote :

>"Patrick Rotsaert" <patrick(dot)rotsaert(at)arrowup(dot)be> writes:
>
>
>>At the time of the crash, the server logs:
>> LOG: server process (PID 21815) was terminated by signal 9
>>
>>
>
>You're running on a Linux machine with memory overcommit enabled.
>Turn that off, or nothing will ever work very reliably --- the OOM
>killer is entirely capable of zapping innocent processes that have
>nothing to do with the one eating too much memory; and even when it
>kills the right process, "kill -9" is not IMHO an acceptable way for
>the system to tell a process it can't have any more memory. See
>http://www.postgresql.org/docs/8.1/static/kernel-resources.html#AEN18105
>
> regards, tom lane
>
>
Syslog indeed shows:
kernel: Out of Memory: Killed process 21815 (postmaster).

Looking at the kernel source mm/mmap.c, the function `int
vm_enough_memory(long pages)' does 1 simple test:
/* Sometimes we want to use more memory than we have. */
if (sysctl_overcommit_memory)
return 1;
But /proc/sys/vm/overcommit_memory reads `0', so my guess is that
overcommit is not enabled... right?
Any hints?

Thanks,
Patrick Rotsaert

In response to

Responses

Browse pgsql-bugs by date

  From Date Subject
Next Message Philipp Reisner 2006-01-31 15:47:30 Fwd: Re: Endless loop in ExecNestLoop
Previous Message Tom Lane 2006-01-31 15:39:58 Re: Endless loop in ExecNestLoop