[torqueusers] checkpoints and walltime

Charles Henry chenry at ittc.ku.edu
Mon Dec 3 16:05:41 MST 2012



----- Original Message -----
> From: "Brian Contractor Andrus" <bdandrus at nps.edu>
> To: "Torque Users Mailing List (torqueusers at supercluster.org)" <torqueusers at supercluster.org>
> Sent: Monday, December 3, 2012 1:22:40 PM
> Subject: [torqueusers] checkpoints and walltime
> 
> All,
> 
> I am seeing something odd.
> If someone request enough walltime to land in my short queue rather
> than my tiny queue (default), /etc/profile is not sourced.
> The setting that seems to make a difference is checkpoint_defaults.
> If that is enabled, various env variables are not set (HOSTNAME in
> particular)
> If it is disabled, everything seems to be set as expected.
> 
> Is there something about checkpointing that affects environment
> settings that I am missing?

Ah-ha!  Now, that makes sense.  
I observed the same problem on our system (just now--I had not previously tested that checkpoint defaults were related to my non-login shell issue).  I had previously seen that cluster jobs would not run in a "login shell", except when running them interactively.  You can read my correspondence with (the very helpful) torque-dev list here:
http://www.supercluster.org/pipermail/torquedev/2012-October/004266.html

That helps a lot to narrow it down perhaps.  I will re-file my bug report with the updated symptom:
http://www.clusterresources.com/bugzilla/show_bug.cgi?id=224

Chuck


> 
> 
> Brian Andrus
> ITACS/Research Computing
> Naval Postgraduate School
> Monterey, California
> voice: 831-656-6238
> 
> 
> _______________________________________________
> torqueusers mailing list
> torqueusers at supercluster.org
> http://www.supercluster.org/mailman/listinfo/torqueusers
> 


More information about the torqueusers mailing list