On Wed, 19 Nov 2014 13:44:37 +0000 Kevin Taylor <[email protected]> wrote:
> I'm not sure if this is possible or not, but thought I'd ask it. > > We have a setup of consumable resources for our GPUs. If a system has two we > have a complex called gpu1_free and gpu2_free. They'll be equal to 1 if > they're free and zero if they're not. Typically we just request like this: > qsub -l gpu1_free=1 job.sh > > Is there a way though qsub to say > > qsub -l gpu1_free=1 OR gpu2_free=1 job.sh > > I know putting multiple -l's will ask for both, but we just want one or the > other, whichever. > > Univa has that nice RSMAP feature that would solve our issue, but we haven't > worked out finances on that yet so we're seeing if we can just make it work a > little easier with what we have. The solution we use on our cluster is as follows: we have a single integer complex for all GPUs attached to a node. The prolog then twiddles lock files to allocates one or both GPUs to the job as appropriate and changes permissions on the /dev/nv* files to ensure the job can only use the GPU it has been assigned (you have to pass a a special parameter to the nvidia module to stop it setting the permissions itself though). -- William Hay <[email protected]>
pgpljA6KTyUj7.pgp
Description: PGP signature
_______________________________________________ users mailing list [email protected] https://gridengine.org/mailman/listinfo/users
