[Beowulf] Torque Error: multi-req PBS jobs not allowed
Rahul Nabar
rpnabar at gmail.com
Mon Apr 20 14:59:31 PDT 2009
Often I need to request 9 cpus on our architecture with 8 cpus/server.
In such situations it makes sense to ask for a full server (8 cpus)
plus 1 more cpu wherever I can get it.
What is the best request to do this? [Scheduler: PBS-Torque;
Super-scheduler: Maui]
I am trying
#PBS -l nodes=1:ppn=8+1:ppn=1
The job gets queued but checkjob reveals an error due to which it won't run:
Torque Error: multi-req PBS jobs not allowed
On the other hand jobs with the following request run without trouble:
#PBS -l nodes=node01:ppn=8+node02:ppn=1
Why is one form (more restrictive) ok and the other (less restrictive)
not? Is this just non-intuitive behavior or a PBS bug?
Some googling shows that perhaps "ENABLEMULTIREQJOBS" might be my
solution (http://www.clusterresources.com/products/maui/docs/a.fparameters.shtml).
But just wanted to get an opinion if others have this enabled or not.
What, if any, are the downsides?
Why would PBS-Torque not allow this and my previous threads
"JOBNODEMATCHPOLICY EXACTNODE" by default? Are there any reasons not
to use them? The compromise is not obvious to me.
--
Rahul
More information about the Beowulf
mailing list