[GE users] relationship between qsub and qstat and queue allocations

Margaret Doll Margaret_Doll at brown.edu
Thu Nov 13 14:02:58 GMT 2008


On Nov 13, 2008, at 3:52 AM, reuti wrote:

> Hi Margaret,
>
> Am 12.11.2008 um 23:36 schrieb Margaret Doll:
>
>> qsub is not working the way that I thought it should.  Each qsub may
>> start several instances of a job, but it will
>>
>> create only one instance of a running job showing in qmon
>> only one instance of a queued job using "qstat -f" and
>> seems to count  as only one job on one  of the  compute nodes.
>>
>> For instance, I am using
>>
>> qsub -q mem16.q shll
>>
>> where shll includes:
>>
>> #!/bin/bash
>> #$ -o $HOME/works-1/Out
>> #$ -j y
>> /opt/openmpi/bin/mpiexec -v -n 17 -machinefile $Home/works-1/
>> machinefile $Home/works-1/mad
>
> although Open MPI has a tight integration into SGE built in, you will
> need to define and request a PE (parallel environment), instead of
> supplying your own list of machines.
>
> http://www.open-mpi.org/faq/?category=running#run-n1ge-or-sge
>
> -- Reuti
>

I found the settings.sh in /opt/gridengine/default/common/settings.sh

I believe, however, before I run

qsh -pe  orte 4
mpirun -np 4  a.out

I have to set up a parallel environment named orte.  How do I set up  
numerous
parallel environments that correspond to the queues that I set up in  
qmon?


>
>>
>> machinefile includes:
>>
>> compute-0-10
>> compute-0-11
>> compute-0-10
>> compute-0-11
>> compute-0-10
>> compute-0-11
>> compute-0-10
>> compute-0-11
>> compute-0-10
>> compute-0-11
>> compute-0-10
>> compute-0-11
>> compute-0-10
>> compute-0-11
>> compute-0-10
>> compute-0-11
>> compute-0-10
>> compute-0-11
>>
>> I have each of the  compute nodes set to run  only eight queued jobs
>> at  a time.
>>
>> A queued job will show up on compute-0-10 when I run "qstat -f"
>>
>> compute-0-10 will be running  9 instances of the program;  
>> compute-0-11
>> will be running 8.
>>
>> What am I doing incorrectly?
>>
>> ------------------------------------------------------
>> http://gridengine.sunsource.net/ds/viewMessage.do?
>> dsForumId=38&dsMessageId=88651
>>
>> To unsubscribe from this discussion, e-mail: [users-
>> unsubscribe at gridengine.sunsource.net].
>
> ------------------------------------------------------
> http://gridengine.sunsource.net/ds/viewMessage.do?dsForumId=38&dsMessageId=88664
>
> To unsubscribe from this discussion, e-mail: [users-unsubscribe at gridengine.sunsource.net 
> ].

------------------------------------------------------
http://gridengine.sunsource.net/ds/viewMessage.do?dsForumId=38&dsMessageId=88680

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



More information about the gridengine-users mailing list