On Tue, 29 Mar 2005, Greg Sabino Mullane wrote:
> Not many data points yet, but here's all the occurances:
>
> $ grep "signal 9" 5810.log
> <2005-03-28 03:38:14 EST >LOG: server process (PID 29216) was terminated by signal 9
> <2005-03-28 10:15:45 EST >LOG: background writer process (PID 29271) was terminated by signal 9
> <2005-03-28 11:05:00 EST >LOG: background writer process (PID 2202) was terminated by signal 9
> <2005-03-28 12:22:26 EST >LOG: background writer process (PID 7748) was terminated by signal 9
> <2005-03-28 12:27:37 EST >LOG: background writer process (PID 17119) was terminated by signal 9
> <2005-03-28 12:27:48 EST >LOG: background writer process (PID 17404) was terminated by signal 9
> <2005-03-28 12:28:27 EST >LOG: background writer process (PID 17409) was terminated by signal 9
> <2005-03-28 12:28:41 EST >LOG: startup process (PID 17452) was terminated by signal 9
> <2005-03-28 13:52:00 EST >LOG: 00000: background writer process (PID 21456) was terminated by signal 9
> <2005-03-28 15:26:25 EST 21453>LOG: 00000: background writer process (PID 24526) was terminated by signal 9
> <2005-03-28 15:39:16 EST 21453>LOG: 00000: background writer process (PID 655) was terminated by signal 9
> <2005-03-28 16:05:02 EST 21453>LOG: 00000: background writer process (PID 1579) was terminated by signal 9
> <2005-03-28 16:05:09 EST 21453>LOG: 00000: background writer process (PID 4141) was terminated by signal 9
> <2005-03-28 16:15:06 EST 21453>LOG: 00000: background writer process (PID 4159) was terminated by signal 9
> <2005-03-28 16:15:20 EST 21453>LOG: 00000: background writer process (PID 5822) was terminated by signal 9
>
> > Any chance you have a "terminate with extreme prejudice" daemon or
> > cron job that kills processes meeting certain conditions?
>
> Nope. I've been happily running pg on this server for many years. I'd frankly suspect
> the computer before Postgres, however, but a better error message (e.g. pointing
> to a corrupt disk or something) would be helpful. The first one was a server process
> and not a background writer process, FWIW.
Well, it seems to be saying that it was terminated by SIGKILL which I
can't see a reason to be internally generated. Is there anything else
running on the system that might (for example) be taxing memory to cause
an OOM killing spree or some such?