[GE users] prioritized queues - suspend jobs.

Reuti reuti at staff.uni-marburg.de
Thu Oct 5 21:27:31 BST 2006


Am 05.10.2006 um 21:51 schrieb Jesper Krogh:

>> Am 05.10.2006 um 14:27 schrieb Jesper Krogh:
>>>> Am 05.10.2006 um 06:06 schrieb Jesper Krogh:
>>>> Are you requesting any resources in your qsub?
>>>>
>>> I'm not sure what you mean by that.. All jobs are compute-intensive
>>> stuff, so they dont need any special resources.
>>
>> So you didn't request anything with -l?
>
> No. just qsub (-cwd -V) job.sh
>
>> Can you please post your queue definitions, and the output of  
>> "qstat -
>> j <jobid>" of one of the waiting jobs? - Reuti
>
> This is really strange.. I could provoke the wrong behaviour just a  
> few
> hours ago, but not a the moment. Here's the queue definitions  
> anyway in
> case you can see somehing wrong in them.
>
> http://krogh.cc/~jesper/longjobs.txt
> http://krogh.cc/~jesper/shortjobs.txt
> http://krogh.cc/~jesper/allqueue.txt

If all of your nodes have 2 CPUs, it's just easier to specify "slost  
2" and remove all the special definitions for each machine. The other  
thing is: don't attach the same PE btmpich to more than one queue (at  
least on the same list of hosts). As already stated, you might get a  
mixture of slots of different queues. To be on the safe side, just  
define thee PEs btmpich1.. btmpich3 and attach only one to each  
queue. In the qsub command you will have to use "-pe btmpich*".

-- Reuti

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
For additional commands, e-mail: users-help at gridengine.sunsource.net




More information about the gridengine-users mailing list