[torqueusers] Re: PBS with abaqus

Brock Palen brockp at umich.edu
Wed Jan 14 08:30:54 MST 2009


If in your sites/abaqus_v6.env file has:

mp_file_system = (DETECT,DETECT)

It abaqus should figure out if scratch and not scratch are local or not.

Also they ship with mp_rsh_command set to rsh by default (dumb)

just put:

mp_rsh_command='ssh -n -l %U %H %C'

in your abaqus_v6.env  file.

Brock Palen
www.umich.edu/~brockp
Center for Advanced Computing
brockp at umich.edu
(734)936-1985



On Jan 13, 2009, at 3:05 PM, Sebastien Corot wrote:

> Nicolas,
>
> When using abaqus with PBS, I suggest you to use the python scripts
> available on the simulia support page (see support answer ID 473: "How
> can I run Abaqus/Standard or Abaqus/Explicit in a batch queue?"). This
> answer is theoricaly available for costumer without support.
>
> Once these files installed as explained, you will be able to manage  
> many
> pbs options within the abaqus command line; the PBS script is
> automatically generated.
> Furthermore, stuff like multiple input files study, user elements,
> restarted study, etc... are fully supported and correctly submited  
> to the
> PBS subsystem by simply typing the usual abaqus command line !!
>
> I installed it on our cluster and things works fine with it.
>
> Concerning the answer suggested by Brock, the python scripts I talk
> about, are able to detect shared or local filesystem.
>
> Sebastien
>
> On Mon, 12 Jan 2009 16:04:36 +0100, Nicolas Moulin wrote:
>
>> Hi
>> I'm setting up a torque-abaqus cluster. I'm getting some errors  
>> when I
>> submit an abaqus parallel job. When I start the calcul without PBS in
>> standard or parallel, for exemple:
>>
>> /abaqus job=COINv71 input=COINv71.inp scratch=$SCRATCH interactive
>> abaqus job=COINv71 input=COINv71.inp scratch=$SCRATCH interactive  
>> cpus=4
>>
>> /I have no errors.
>> When I use the next PBS script to start the abaqus calcul, I have no
>> errors: /#PBS -N jobabaqus
>> #PBS -o jobabaqus.out
>> #PBS -e jobabaqus.err
>> #PBS -l nodes=1:ppn=1
>> #PBS -l pmem=3000mb
>> #PBS -l walltime=01:00:00
>> #PBS -S /bin/sh
>> #PBS -V
>> module load abaqus
>> SCRATCH=/data/$USER
>> if [ ! -d $SCRATCH ] ; then
>> echo "No local scratch filesystem"
>> exit 1
>> fi
>> cd $PBS_O_WORKDIR
>> abaqus job=COINv71 input=COINv71.inp scratch=$SCRATCH interactive  
>> rm -rf
>> $SCRATCH/*/
>>
>> But, if I try to start a parallel job with a PBS script with the next
>> script:
>> /#PBS -N jobabaqus
>> #PBS -o jobabaqus.out
>> #PBS -e jobabaqus.err
>> #PBS -l nodes=4:ppn=1
>> #PBS -l pmem=3000mb
>> #PBS -l walltime=01:00:00
>> #PBS -S /bin/sh
>> #PBS -V
>> module load abaqus
>> SCRATCH=/data/$USER
>> if [ ! -d $SCRATCH ] ; then
>> echo "No local scratch filesystem"
>> exit 1
>> fi
>> cd $PBS_O_WORKDIR
>> abaqus job=COINv71 input=COINv71.inp scratch=$SCRATCH interactive
>> //cpus=4/ /rm -rf $SCRATCH/*/
>>
>>  I'm getting some errors like that :
>> /Abaqus License Manager checked out the following licenses:
>> Abaqus/Standard checked out 8 tokens. <27 out of 35 licenses remain
>> available>. mpid: Cannot set work directory: No such file or  
>> directory
>> mpid: MPI_WORKDIR=/data/nmoulin/nmoulin_COINv71_7518/
>
>
> _______________________________________________
> torqueusers mailing list
> torqueusers at supercluster.org
> http://www.supercluster.org/mailman/listinfo/torqueusers
>
>



More information about the torqueusers mailing list