Job Queueing on the Bayreuth Clusters

Job handling on the Bayreuth Clusters is done with the Torque/PPS queueing system. To run a job on a cluster (besides the executable program doing the actual calculatiuons) a job file is needed, which is submitted to the queue. A typical job file consits of two parts:

Description of the resources needed

The resources needed like the number of computenodes needed or the maximum of walltime to execute this job can either be sepcified on the commandline of the qsub or specified in lines starting with #PBS within the job file

Example File

An example job file executing a program named doit.exe on one cpu-core on one computenode for a maxiumum of 5 Minutes could look like
#!/bin/tcsh
#PBS -l nodes=1:ppn=1,walltime=00:05:00
#PBS -j oe
#PBS -m abe
#PBS -M joe.user@uni-bayreuth.de
cd $PBS_O_WORKDIR
date
./doit.exe
date

Job Handling

  • To submit a job use qsub [JobScript]
  • To see all jobs in the queue use qstat
  • To remove a job (either waiting or executing) use qdel [JobID]

    Note

    In the event of a compute node crashing while executing a job the user will be unable to remove the job from the dead node. Please contact the cluster administrators.
    Last modified December 2016 by Dr. Bernhard L. Winkler