MPIRUN 与 PBS 脚本

 

Submiting a Batch Job

Batch jobs are subbimted using the command:

[depietri@albert PBS]$ qsub -q [queue]  [batchscript.sh]

or, if a different than defalt number of nodes is request, by the commnad:

[depietri@albert PBS]$ qsub -q [queue]  -l nodes=[nnodes:queue]  [batchscript.sh]

Here:
[queue] is one of the five configurated queues n1, n8, n16, n64 or n88;
[batchscript.sh] is the name of the script file to be submited;
[nnodes:queue] is the number of processor that should be allocated for the batch job.

The number of proccessor to be allocated may be declared on the [batchscript.sh] script file or the queue default number of nodes is allocated. On albert100 the default number of allocated processor is: 1 for the queues n1, 8 for the queues n8, 16 for the queues n16, 64 for the queues n64 and 88 for the queues n88

Main commands:

qstat -q (shows theconfigured queue)
qsub  jobscript (submit the job to the default queue)
qsub  -q  <qname> jobscript (submit the job to the queue <qname>)
qsub -a (shows the submitted jobs)
qdel jobid (cancels the specified jobid from the default queue)

Example of a single node Job Script

    #PBS -S /bin/sh
#PBS -m ae
#PBS -M depietri@albert.pr.infn.it
## --------------------  PREAMBLE -----------------------------------
export RUN_HOST=`hostname`
n=`wc -l < $PBS_NODEFILE`
cd $PBS_O_WORKDIR
## --------------------  END PREAMBLE --------------------------------
echo "================================================================================"
echo "."
echo ".                                           Albert100 single processor BATCH JOB"
echo ".                                           ------------------------------------"
echo "."
echo ". Job Running on queue: $PBS_QUEUE"
echo "."
echo ". Running on HOST:    $RUN_HOST"
echo ". Allocated processor: $n"
echo ". Working Dir:  `pwd`"
echo "."
echo "================================================================================"
echo .

 

Example of a MPICH/Cactus Job Script

    #PBS -S /bin/sh
#PBS -m ae
#PBS -M depietri@albert.pr.infn.it
## --------------------  PREAMBLE -----------------------------------
export MFILE="$PBS_O_WORKDIR/mfile.$PBS_QUEUE.$PBS_JOBID"
export MPIRUN="/opt/mpich-1.2.0-smp/bin/mpirun"
export RUN_HOST=`hostname`
n=`wc -l < $PBS_NODEFILE`
cd $PBS_O_WORKDIR
/opt/bin/machinefile2smp < $PBS_NODEFILE > $MFILE
## --------------------  END PREAMBLE --------------------------------
echo "================================================================================"
echo "."
echo ".                                                Albert100 mpich 1.2.0 BATCH JOB"
echo ".                                                -------------------------------"
echo "."
echo ". Job Running on queue: $PBS_QUEUE"
echo "."
echo ". Root host:    $RUN_HOST"
echo ". Running on:   $n processors"
echo ". Working Dir:  `pwd`"
echo ". Using mpirun: $MPIRUN"
echo ". -machinefile: $MFILE"
echo "--------------------------------------------------------------------------------"
cat $MFILE
echo "================================================================================"
echo .
$MPIRUN -np $n -machinefile $MFILE ./BenchADMsmp  BenchADM.par
\rm $MFILE

Example of a LAM/Cactus Job Script

    #PBS -S /bin/sh
#PBS -m ae
#PBS -M depietri@albert.pr.infn.it
## --------------------  PREAMBLE --------------------------------------------------
##
## Here: We set as working directory the one where the qsub command as been lunched.
##       We generate the machine file for the requested nodes
##       The appropriate mpirun command is set
##
source /opt/bin/LAM-6.5.6.sh
export MFILE="$PBS_O_WORKDIR/mfile.$PBS_QUEUE.$PBS_JOBID"
export MPIRUN="/opt/lam-6.5.6-usysv.1/bin/mpirun"
export RUN_HOST=`hostname`
n=`wc -l < $PBS_NODEFILE`
cd $PBS_O_WORKDIR
/opt/bin/machinefile2lam < $PBS_NODEFILE > $MFILE
## --------------------  END PREAMBLE ------------------------------------------------
echo "================================================================================"
echo "."
echo ".                                                  Albert100 LAM 6.5.6 BATCH JOB"
echo ".                                                  -----------------------------"
echo "."
echo ". Job Running on queue: $PBS_QUEUE"
echo "."
echo ". Root host:    $RUN_HOST"
echo ". Running on:   $n processors"
echo ". Working Dir:  `pwd`"
echo ". Using mpirun: $MPIRUN"
echo ". -machinefile: $MFILE"
echo "--------------------------------------------------------------------------------"
cat $MFILE
echo ================================================================================
echo .
lamboot -v $MFILE
$MPIRUN -np $n  ./BenchADM-lam  BenchADM.par
lamhalt
# -----------------------------------------------------
# We can now delete the temporary generate Machine File
# -----------------------------------------------------
\rm $MFILE