Hi,
I have implemented the soft/hard limits on the MAXPROC parameter, as per
the maui manual (maui version 3.2.6p13, torque 2.0.0p2), but I am seeing
unexpected behavior:
- defined the parameter group-wise as follows
GROUPCFG[A] MAXPROC=100,200
GROUPCFG[B] MAXPROC=50,175
GROUPCFG[C] MAXPROC=40,175
GROUPCFG[DEFAULT] MAXPROC=30,175
- there are 5 cpus free and no jobs are in Q. Here is the group-wise
processor usage (active procs used):
group A 8
group B 10
group C 57
group other1 61
group other2 5
group X 88
- A user from group X then submits several more jobs. Each of these jobs
remains in Q state with checkjob reporting that "job 55059 violates
active SOFT MAXPROC limit of 30 for group X". There are still 5 cpus free.
This does not seem normal. I would have expected that group X would be
allowed to run jobs until they hit total procs=175, since there are no
other competing jobs. Infact, the way I read the manual, this setup
would allow groups 'other1', 'other2' and X to go upto max 175 procs
each. Right?
Also, the diagnostic message from maui is misleading. The soft-limit of
30 has long been crossed !
Can anyone suggest what am I missing in this picture? What's the
correlation between these numbers?
Thanks.
-Neel
_______________________________________________
mauiusers mailing list
[email protected]
http://www.supercluster.org/mailman/listinfo/mauiusers
