[Mauiusers] maui + torque job start rate

Jones de Andrade johannesrs at gmail.com
Mon Sep 14 17:48:49 MDT 2009


Hi all.

I'm having a hard time here setting up the limits for our queue system.

We have the following line in our /usr/local/maui/maui.cfg file:

CLASSCFG[cluster]  MAXPROC[GROUP]=16 MAXPROC[USER]=8

The rest of the file is pretty much the standard one. Shouldn't it alone be
enough to constrain any user to not be using more than 8 processors? Also,
neither group to use more than 16?

My problem is that, despite that line, we have users getting easilly 12
processors, and group summing up to 20. For example, user "gullit" bellow is
using 12 processors at this moment, and group "hubert" sums up to 20
processors (sorry, that was the simplest command line and output I could
come up with):

#############
*$ showq*
ACTIVE JOBS--------------------
JOBNAME            USERNAME      STATE  PROC   REMAINING
STARTTIME

323                  gullit    Running     4 96:07:23:19  Fri Sep 11
03:50:54
325                  gullit    Running     4 97:06:23:48  Sat Sep 12
02:51:23
327                 msegala    Running     4 98:03:23:32  Sat Sep 12
23:51:07
328                  gullit    Running     4 98:14:41:59  Sun Sep 13
11:09:34
330                   elvis    Running     4 99:17:40:42  Mon Sep 14
14:08:17
331                jeferson    Running     4 99:23:09:18  Mon Sep 14
19:36:53
...

*$ qstat -f | egrep '(egroup|job_state|Resource_List.nodes)'*
    job_state = R
    Resource_List.nodes = 1:ppn=4
    egroup = hubert
    job_state = R
    Resource_List.nodes = 1:ppn=4
    egroup = hubert
    job_state = R
    Resource_List.nodes = 1:ppn=4
    egroup = livotto
    job_state = R
    Resource_List.nodes = 1:ppn=4
    egroup = hubert
    job_state = R
    Resource_List.nodes = 1:ppn=4
    egroup = hubert
    job_state = R
    Resource_List.nodes = 1:ppn=4
    egroup = hubert
    job_state = Q
    Resource_List.nodes = 1:ppn=4
    egroup = hubert
#############

The problem is that the policy here is that no one can surpass it's
stablished limits, no matter what, when or even if there are any spare nodes
available.

What am I doing wrong here? Should I change/add something else in the
configuration file?

Thanks a lot in advance.

Sincerally yours,

Jones
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.supercluster.org/pipermail/mauiusers/attachments/20090914/2969e023/attachment.html 


More information about the mauiusers mailing list