[torqueusers] Submitting jobs to use multiprocessors.

Gus Correa gus at ldeo.columbia.edu
Wed Mar 12 17:21:22 MDT 2014


Hi Hitesh

Have you tried an alternative syntax with nodes and ppn?
For instance (assuming one node has >=4 cores):

#PBS -l nodes=1:ppn=4

Did you enable scheduling on the pbs_server?

Which scheduler are you using, pbs_sched, Maui, Moab, other?

FWIW, I was never lucky with --enable-numa-support with Torque+Maui
on Linux (with Opteron processors, which are legitimate NUMA processors).
Jobs would hang in Q state, or flip between Q and R, but never really run.
Hence, I gave it up, reconfigured without that option.
My guess is that the --enable-numa-support option is
tailored for SGI computers.
Do you have SGI computers?
In any case, it requires a specific torque numa configuration
in each node on $TORQUE/mom_priv.
Details here:

http://docs.adaptivecomputing.com/torque/4-0-2/Content/topics/1-installConfig/buildingWithNUMA.htm

I hope it helps,
Gus Correa

On 03/12/2014 06:26 PM, hitesh chugani wrote:
> Hi all,
>
>
> I am trying to submit a job with to use multiprocessors(Added #PBS -l
> procs=4 in the job script) but the job is remaining queued forever. I am
> using 2 computes nodes (ncpus=8 and 2). Any idea why is it not running?
> Please help.
>
> I have installed torque using this configuration option.
> *./configure --enable-unixsockets --enable-cpuset
> --enable-geometry-requests --enable-numa-support *
>
>
>
>
> Thanks,
> Hitesh Chugani.
> Student Linux specialist
> University of North Carolina, Charlotte
>
>
> _______________________________________________
> torqueusers mailing list
> torqueusers at supercluster.org
> http://www.supercluster.org/mailman/listinfo/torqueusers
>



More information about the torqueusers mailing list