Batch System
Batch farm at Atlas HU computing cluster
Description
The batch system installed on the Atlas computer cluster is Torque which is based on the Open Portable Batch System, OpenPBS. At present there two separated batch system running on the msXXb and the asXXb machines.Before submitting batch jobs make sure the jobs are tested thoroughly. Try to share the workload among the different machines and also to keep network traffic (e.g. by reading from data disks of other machines) low. Please refer to the rules for the usage of the HU Atlas cluster.
Still, the commands of the former batch system dqs are valid. The relevant commands for managing batch jobs are qstat, qsub, qdel, qhold and qrls. For details have a look a their respective man pages.
Hello World Example
File runscript.sh
#!/bin/sh #Hello world run script echo "Hello world!"
Command line
qsub runscript.sh
Selecting the Node
In order to submit jobs to a certain node use the -l option, eg
qsub -l nodes=ms03b myrunscript