Re: pre9-4 OOM VM lockups

Date: Wed May 24 2000 - 02:09:12 EST

On 23 May, Jesse Pollard wrote:
> I think it really depends on the other load on the system. If the eatmem
> program is the only one running at the time the system goes OOM it will
> correctly identify it as the hog. The more other processes (non eatmem)
> that are active, the more likely that the other process will be killed.
> Mine survived, but I have hard limits available on a test system.
> It doesn't survive if I run it as root, or if I run it without limits.

No, I don't think that this is actually true... When testing with 2.3,
I knew I was going to blow my system, so I shut down as many daemons as
possible and unmounted as many filesystems as possible. When testing
with 2.2, all I shut down was X, but none of the daemons.
So with 2.2 the system was slightly more loaded and yet it consistently
behaved better.

I started this thread because of a runaway debconf perlscript which
caused my computer to lock up due to OOM -- in 2.2, the same program
just segfaults after using up all available memory.
That is repeatable for both the 2.3 and 2.2 kernels, however it's not a
very good test that others can use because I guess it depends too much
on the state of my Debian installation, and you'll need Debian in the
first place to try it.
Hence I undested 'eatmem', which is an artificial way to achieve a
similar effect.

> The only versions of Unix that would survive are those that cannot overcommit.

Ouch, let's not get started on that thread again -- those were rather
long & heated discussions :-)

> -------------------------------------------------------------------------
> Jesse I Pollard, II
> Email:
> Any opinions expressed are solely my own.

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to
Please read the FAQ at

This archive was generated by hypermail 2b29 : Wed May 31 2000 - 21:00:11 EST