[GE users] Strange behaviour with Matlab runtime

reuti reuti at staff.uni-marburg.de
Fri Dec 12 17:02:34 GMT 2008


Am 12.12.2008 um 17:23 schrieb Aaron Turner:

> Dear all,
> A few weeks ago I mentioned here a strange problem we were having  
> with a
> job using Matlab runtime.
> The behaviour was that the job would run on the command line in a few
> seconds but any attempt to run it under SGE (batch, or qlogin,  
> etc.) led
> to the job hanging forever, not actually using any more CPU time than
> the command line would do, until the real time limit on the queue  
> killed
> it some hours later. The job seemed to hang in one of the routines
> initialising the MatLab runtime environment. There was no indication
> from the logging in SGE on what might be happening.
> After some more investigations (trial and error) it seems that the
> specification of vmem limits in the queue instances is what seems to
> stop the job running. I haven't seen this behaviour in conjunction  
> with
> any other executables and wondered if anyone else had.

did you set any limits in the queue configuration? If you set h_vmem,  
also h_stack and h_data will be set. Some programs will get confused  
with such a high value for h_stack and will need a lower limit there  
(e.g. Gaussian03, which will need a stack of just h_vmem=32M

-- Reuti

> Regards,
>    Aaron Turner
> ------------------------------------------------------
> http://gridengine.sunsource.net/ds/viewMessage.do? 
> dsForumId=38&dsMessageId=92399
> To unsubscribe from this discussion, e-mail: [users- 
> unsubscribe at gridengine.sunsource.net].


To unsubscribe from this discussion, e-mail: [users-unsubscribe at gridengine.sunsource.net].

More information about the gridengine-users mailing list