OOM killer "Out of Memory: Killed process" SOLUTIONS / SUMMARY

Eric Sisler esisler at westminster.lib.co.us
Mon Aug 20 18:52:33 UTC 2007


On Mon, 2007-08-20 at 12:06 -0400, Cristian Silva wrote:

> Sorry, forgot to mention that we are getting the oom-killer with another
> application, we are not using VMware.

The actual application is pretty much irrelevant once the oom-killer is
involved.

> The total low memory reported by the PAE and the normal kernel are the same
> 
> PAE
> # free -lm
>              total       used       free     shared    buffers     cached
> Mem:          7994        348       7646          0         70        179
> Low:           827        102        724
> High:         7167        245       6922
> -/+ buffers/cache:         98       7896
> Swap:         4094          0       4094
> #
>
> 
> Normal
> # free -lm
>              total       used       free     shared    buffers     cached
> Mem:          3930        309       3621          0         17        196
> Low:           859         47        811
> High:         3071        261       2810
> -/+ buffers/cache:         95       3835
> Swap:         4094          0       4094
> #
> 
> Both commands were executed just after a reboot of the system and it's
> interesting to see that the used low memory in the PAE kernel is the double
> of the  used low memory in a  "normal"  kernel

There's also slightly less low memory when running PAE, but obviously
way more high memory since the regular kernel is limited to 4Gb.  My
understanding is that the PAE kernel *should* be roughly equivalent to
the RHEL4 hugemem kernel, but unfortunately this doesn't seem to be the
case.  Per my earlier post, low memory on a box running RHEL4 jumped
significantly when running the hugemem kernel.  I wonder why this isn't
the case when using the RHEL5 PAE kernel?  Unfortunately I don't have
any speare hardware to experiment with at the moment.

> We already tried the parameters that Ray suggested and they didnt make any
> difference.
> 
> vm.lower_zone_protection is not available in the rhel5, so we are going to
> try with less memory and see what it happens.
> 
> Thanks all for the information/suggestions, this thread has been excellent.

You mentioned 64-bit wasn't an option at the moment...any chance it migh
be?  When I installed 64-bit RHEL4, it also installed the 32-bit
versions of many packages and thus far I haven't had any issues.

-Eric

-- 
Eric Sisler <esisler at westminster.lib.co.us>
Library Network Specialist
Westminster Public Library
Westminster, CO USA

Linux - Don't fear the Penguin.
Want to know what we use Linux for?
Visit http://wallace.westminster.lib.co.us/linux




More information about the redhat-list mailing list