I am using the latest snapshot of maui 3.2.6p16 and torque 2.1.3.
In the maui.cfg I am using the following options:
NODEMAXLOAD 4.00
NODEALLOCATIONPOLICY CPULOAD
NODELOADPOLICY ADJUSTSTATE
NODEAVAILABILITYPOLICY UTILIZED
DEFERTIME 0
JOBNODEMATCHPOLICY EXACTNODE
I am using cpuload as the node allocation policy as I want jobs
submitted on a load balanced basis.
If I submit a job and there are enough nodes on the cluster that have
no jobs running, to cover what is specified by the job submitted,
then everything is ok.
However if I submit a job that requires resources that have jobs
running, a checkjob on that job just says at the end:
cannot select job for partition DEFAULT (startdate in '00:00:01')
and the job just stays in the idle queue. (It can be run manually with qrun)
pbsnodes -a and checknode show that there are enough free resources
available as the load on nodes that have jobs already running is
below nodemaxload.
If I set NODEALLOCATIONPOLICY is set to minresource, for example,
this problem does not arise.
But then the jobs submitted are not load balanced.
Anybody have any ideas what could be going one here?
Thanks,
Regards,
Eugene.
_______________________________________________
mauiusers mailing list
[email protected]
http://www.supercluster.org/mailman/listinfo/mauiusers
