Re: postgresql 8 abort with signal 10

On Tue, 2005-05-03 at 08:39, Alexandre Biancalana wrote:
> Hi list,
>
> I'm running postgresql 8.0.1 on FreeBSD 4.11-STABLE, the machine is
> and AMD Sempron 2.2, 1GB Ram..
>
> I use postgresql as database for dspam, an spam classification
> program. This database have and moderated use, on averange 10
> simultaneous conections executing relative big queries using "in"
> clausule.
>
> Watching postgresql logs I see the following messages ocurs a lot of
> times in a day:
>
> May 3 06:58:44 e-filter postgres[250]: [21-1] LOG: server process
> (PID 59608) was terminated by signal 10
> May 3 06:58:44 e-filter postgres[250]: [22-1] LOG: terminating any
> other active server processes

SNIP

> This is my postgresql.conf
>
> max_connections = 70
> superuser_reserved_connections = 2
> shared_buffers = 81920

Rather large, shared buffers for a machine with only 1 gig of ram. 640
Meg of RAM means the kernel is basically double buffering everything.
have you tested with smaller settings and this setting was the best?

You might want to look in your signal man page on BSD and see what
signal 10 means. On solaris it's a bus error. Not a clue what it is in
FreeBSD myself though.

> work_mem = 10240
> maintenance_work_mem = 51200
> fsync = true
> checkpoint_segments = 8
> effective_cache_size = 100000
> log_destination = 'syslog'
> silent_mode = true
> lc_messages = 'C'
> lc_monetary = 'C'
> lc_numeric = 'C'
> lc_time = 'C'
>
>
> and the shared memory configuration:
>
> kern.ipc.shmmax: 700000000
> kern.ipc.shmmin: 1
> kern.ipc.shmmni: 192
> kern.ipc.shmseg: 256
> kern.ipc.shmall: 700000000
>
>
> I have some configuration error that could result in this kind of problem ?
>
> Any ideas ? Any thoughts ?
>
> Best Regards,
> Alexandre
>
> ---------------------------(end of broadcast)---------------------------
> TIP 4: Don't 'kill -9' the postmaster