Re: Avoiding OOM on overcommit...?

From: Jesse Pollard (pollard@tomcat.admin.navo.hpc.mil)
Date: Tue Mar 21 2000 - 12:43:02 EST


David Whysong <dwhysong@physics.ucsb.edu>:
> On Mon, 20 Mar 2000, Jesse Pollard wrote:
> >On Mon, 20 Mar 2000, David Whysong wrote:
> >>
> >>That's very misleading. In fact if you give the overcommitted system the
> >>same amount of VM, it will work just fine. In other words, turning off
> >>overcommit isn't what saves you. You added more memory!
> >
> >I guaranteed that the memory allocated could be used. I didn't just add
> >more memory. Just adding more memory will still allow the system to fail,
> >it may take longer, it may not happen as often. But it can still happen.
>
> If you properly handle system-wide OOM situations by killing tasks (see my
> other emails about doing this with the OOM killer patch + a userspace
> daemon), then you no longer have system crashes.
>
> Yes, your app can be killed, but that is also what happens with quotas. In
> fact, it happens earlier with quotas.

It happens before the OOM can kill the system. It happens before my process
causes other users to loose theirs.

You obviously haven't used a large multiuser system recently. Quotas are
applied in almost every location I've seen. Why - to prevent one users
job from interfereing with other users jobs.

If you were running a simulation/analysis and were supposedly given
resources to run, and I accessed the same system with the same goal in
mind, which user gets killed when I use up the resources given to me?
BTW, I have the same priority of access to the system that you have...

Who gets killed - your process or mine?
Which is the correct one?
How do you know it is the correct one?
If it happens again, are the answers the same?
-------------------------------------------------------------------------
Jesse I Pollard, II
Email: pollard@navo.hpc.mil

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 majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/



This archive was generated by hypermail 2b29 : Thu Mar 23 2000 - 21:00:33 EST