[torqueusers] job runs only in 1 node

biolabadmin at ibt.lt biolabadmin at ibt.lt
Wed Mar 15 04:15:39 MST 2006


hi again,
i have some more questions:)
here is my runing script for torque + maui+ mpich
#PBS -N bandymas
#PBS -l nodes=5:ppn=2
echo Starting Job:
date
echo
echo Running script on...
hostname
echo
cd /home/skriptai/mpi/
mpiexec -n 10 /home/skriptai/mpi/Profil_MPI_KLP.ex ch.txt

when jobs runing, sheduler says what jobs runing only on 5 nodes, each
2processors, but problems is what jobs are runing on 10 nodes, each 1
proc.
so mpiexec not really looks to PBS permissions. how to say to mpich use
only  5 PBS nodes with 2 processors? thx

p.s. sry for my poor english

>>> How to make torque add jobs on all asked nodes? thx
>
>> I think you are confusing between a parallel job and multiple serial
>> jobs executing in several nodes simultaneously. If you want to execute
>> the same serial job on several nodes simultaneously, but through PBS,
>> then mpiexec is a good way to go (as you have already figured out). If
>> you do not necessarily need PBS / Torque, then look at c3
>> (http://www.csm.ornl.gov/torc/C3/) from ORNL.
>
> To slightly rephrase:
> Torque is only managing and assigning resources (nodes and/or cpus and/or
> memory).
> You are responsible to get you application(s) running on the nodes you
> get.
>
> Usually this is done via mpiexec or similar scripts running on the first
> node of your assigned one spawning children on the other nodes.
>
>
> Cheers,
> Ronny
> _______________________________________________
> torqueusers mailing list
> torqueusers at supercluster.org
> http://www.supercluster.org/mailman/listinfo/torqueusers
>




More information about the torqueusers mailing list