Re: [Condor-users] Memory usage limitation

[Stuart Anderson <anderson@xxxxxxxxxxxxxxxx>]

I have also found it useful to put a kernel level cap on memory, e.g.,
on Linux compute nodes limit all Condor spawned processes, including jobs
started by startd, individually to 1.3GByte of memory by adding to
/etc/init.d/condor: ulimit -v 1300000.


On Mon, Feb 19, 2007 at 11:54:09AM -0600, Burt Holzman wrote:
> Fabrice Charlier wrote:
>   > I would like to know if it's possible to limit the memory usage of a 
> job
> > submitted on a condor cluster.
> > For example, if my job takes more than 1Gb of memory, condor sends a 
> > SIGKILL to it.
> > 
> > I found a old post (2005) on the mailing list that says condor will 
> > never have this functionality.
> > It's always true?
> 
> Somewhere in late 6.7, condor added SYSTEM_PERIODIC_REMOVE.  You can add 
> a SYSTEM_PERIODIC_REMOVE expression such as
> 
> SYSTEM_PERIODIC_REMOVE = (ImageSize > 1000000)
> 
> which will terminate any job whose VSIZE exceeds 1 GB.
> 

-- 
Stuart Anderson  anderson@xxxxxxxxxxxxxxxx
http://www.ligo.caltech.edu/~anderson