[Mauiusers] -W x=”NACCESSPOLICY:SINGLEJOB“ for nodes with different number of requested ppn

Craig West cwest at astro.umass.edu
Mon Jun 29 06:41:16 MDT 2009


Jana,

I just performed the exact same test on my system - running the same 
versions of Maui and Torque. It responded the same way as yours. The 
first two nodes were listed as job exclusive, while the third node was 
not. The third node that was chosen by maui actually already had some of 
the processors in use.
It should be noted that I really did get a total of 10 cpus available 
(3+3+4).

A further item to add to this discussion, is that the number of 
dedicated cpus on the system, actually exceeds the number of cpus 
available for the ppn=4 system in this case. Here are the three lines 
from "checknode"
Dedicated  Resources: PROCS: 16
Dedicated  Resources: PROCS: 16
Dedicated  Resources: PROCS: 24

Each of those nodes has 16 procs.

This looks like a bug.

Craig.

On 06/22/2009 09:20 AM, Jana Uhlirova wrote:
> Hello,
>
> we are running Torque-2.3.6 and Maui-3.2.6p21 at our cluster.
>
> We submitted a job with following requests:
>
> /qsub -q batch -W x=”NACCESSPOLICY:SINGLEJOB” -l nodes=2:ppn=3+1:ppn=4/
>
> /But only first 2 worker nodes (ppn=3) were “job-exclusive”. The third 
> one (ppn=4) wasn' t. Another job started unexpectedly at this node. In 
> the output of showq there was PROC = 20 instead of 24. (All nodes have 
> 8 CPUs)./
>
> /Does anybody have any suggestion?/
>
> /Thank you./
>
> /Best regards/
>
> Jana Uhlirova
>
> ------------------------------------------------------------------------
>
> _______________________________________________
> mauiusers mailing list
> mauiusers at supercluster.org
> http://www.supercluster.org/mailman/listinfo/mauiusers
>   


-- 
Craig West
HPCC Administrator
Astronomy Department
University of Massachusetts
Amherst, MA

413-545-3014



More information about the mauiusers mailing list