Hi,

probably a beginners question ;

I'm running torque 2.0.0p7 and maui 3.2.6p14 on a dual-node cluster.
Jobs that are submitted with overly large ressource requirements (e.g. requesting more memory than available) go to a 'deferred' state and
'checkjob' reports something like :

 job is deferred.  Reason:  NoResources  (cannot create reservation for\
 job  'XXX' (intital reservation attempt)

That's ok.

However, when I adjust the ressource request (using 'qalter') I would expect the job to occasionally run (in particular after issuing a
'releasehold -a XXX' ). It doesn't and checkjob now says:

  Messages:  cannot create reservation for job '624' (intital reservation
  attempt)

  PE:  17.57  StartPriority:  100
  job cannot run in partition DEFAULT (idle procs do not meet requirements:
  0 of 16 procs found)

This looks as if maui is still calculating the PEs with the old ressource request (the job wanted 16 procs and ~10% more mem than available).

The job, when freshly submitted with the adjusted settings, runs fine.

So my question: Isn't maui informed about changed ressource requests ?
                If no, how does one tell maui ??

I do assume that it's not a torque problem since qstat displays everything
correctly.

Any help much appreciated,

Regards,
  Norbert
_______________________________________________
mauiusers mailing list
[email protected]
http://www.supercluster.org/mailman/listinfo/mauiusers

Reply via email to