PREP, TRACE, and POST¶
Assumptions:
Job scheduling on cluster is managed by PBS (Portable Batch System).
Prerequisites:
an environment variable PREP_PATH has to point to directory containing PREP executable
your PYTHONPATH should contain the directory to a directory containing trace.py
Prerequisites for use on a cluster:
for batch mode: /etc/CLUSTERNAME is an ASCII file containing the name of the cluster
for interactive mode: an environment variable CLUSTERNAME has to be set
Start PREP on clusters and workstations¶
$ prep.py --pyhelp
Usage: /home/mors_ch/venv/mojo-develop/bin/prep.py OPTIONS
-np --numberOfProcesses Total number of processes [mandatory]
-q --queue PBS-Queue in which the job will be submitted (only
on clusters).
-l --logFile Filename in which the processes stdout and stderr
will be written.
-e --executable Executable to start. If not given the
corresponding project path will be searched.
-j --jobname Name for the job that will appear in the pbs
queue. This will also determine the names for the
output-/scriptfiles.
-n --nodetype Select a certain node type: either 'w'/'westmere',
'ib'/'ivybridge' or 'sb'/'sandybridge'. (default:
'ivybridge')
-w --waitUntilFinished Waits until the job is finished before returning.
--version Prints the version info.
-h --help Prints help messages of executable and python
script.
-? --pyhelp Prints python command line options only.
-ppn --numberOfProcessesPerNode
Number of processes for each node (e.g. on a
cluster). If not specified a default will be
calculated.
-wt --wallTime Wall time required for this job. Must be smaller
than maximum wall time of queue. Format:
[[[days:]hours:]minutes:]seconds[.milliseconds].
-di --dependJobID The job starts if the job with this ID ends.
-A --account Account for cluster costs.
-m --mailAddress Email address for status information (via local
mailing agent or SLURM.
-sw --slurmSwitches <count>[@minutes] requests a node allocation with
at most <count> network switches within <minutes>
by SLURM
Start TRACE on clusters and workstations¶
$ ./trace.py --pyhelp
bash: Zeile 1: ./trace.py: Datei oder Verzeichnis nicht gefunden
Start POST on clusters and workstations¶
$ post.py --pyhelp
Usage: /home/mors_ch/venv/mojo-develop/bin/post.py OPTIONS
-np --numberOfProcesses Total number of processes [mandatory]
-q --queue PBS-Queue in which the job will be submitted (only
on clusters).
-l --logFile Filename in which the processes stdout and stderr
will be written.
-e --executable Executable to start. If not given the
corresponding project path will be searched.
-j --jobname Name for the job that will appear in the pbs
queue. This will also determine the names for the
output-/scriptfiles.
-n --nodetype Select a certain node type: either 'w'/'westmere',
'ib'/'ivybridge' or 'sb'/'sandybridge'. (default:
'ivybridge')
-w --waitUntilFinished Waits until the job is finished before returning.
--version Prints the version info.
-h --help Prints help messages of executable and python
script.
-? --pyhelp Prints python command line options only.
-ppn --numberOfProcessesPerNode
Number of processes for each node (e.g. on a
cluster). If not specified a default will be
calculated.
-wt --wallTime Wall time required for this job. Must be smaller
than maximum wall time of queue. Format:
[[[days:]hours:]minutes:]seconds[.milliseconds].
-di --dependJobID The job starts if the job with this ID ends.
-A --account Account for cluster costs.
-m --mailAddress Email address for status information (via local
mailing agent or SLURM.
-sw --slurmSwitches <count>[@minutes] requests a node allocation with
at most <count> network switches within <minutes>
by SLURM