On Thu, Feb 09, 2006 at 05:04:38PM -0500, Tom Lane wrote:
> Martijn van Oosterhout <kleptog@svana.org> writes:
> > When people talk about disabling the OOM killer, it doesn't stop the
> > SIGKILL behaviour,
>
> Yes it does, because the situation will never arise.
>
> > it just causes the kernel to return -ENOMEM for
> > malloc() much much earlier... (ie when you still actually have memory
> > available).
>
> Given the current price of disk, there is no sane reason not to have
> enough swap space configured to make this not-a-problem. The OOM kill
> mechanism was a reasonable solution for running systems that were not
> expected to be too reliable anyway on small hardware, but if you're
> trying to run a 24/7 server you're simply incompetent if you don't
> disable it.
BTW, I was shocked when I found out that FreeBSD actually has an OOM
killer itself. Yet I've never heard of anyone having problems with it.
Granted, the FreeBSD OOM could be better designed to pick the right
process to kill, but I'd bet that the real reason you never hear about
it is because FreeBSD admins are clued enough to a) setup a reasonable
amount of swap and b) do a better job of monitoring memory usage so that
you don't start swapping in the first place.
--
Jim C. Nasby, Sr. Engineering Consultant jnasby@pervasive.com
Pervasive Software http://pervasive.com work: 512-231-6117
vcard: http://jim.nasby.net/pervasive.vcf cell: 512-569-9461