Start using Iceberg
Interactive Access to Iceberg
- When you log directly onto iceberg you will be located on one of iceberg's two head-nodes.
- Most of the applications are available only on the WORKER-NODES.
- Therefore the next logical step will be to get onto one of the worker nodes. This is achieved by using the command-
qsh "or qrsh if your terminal client does not support XWindows"
- You will know that you are on a worker-node by observing that the command prompt has changed from-
username@iceberg to username@node... or username@amd-node...
- Head-nodes provide the access point to the worker-nodes and should only be used for light computational tasks
such as file editing.
Batch Processing on Iceberg
qsub command: Having logged onto iceberg, long running jobs should be submitted to be run in batch processing mode by using the qsub command.
If your software is processor-type dependent you can use -l arch=intel* to access the new intel nodes or -l arch=amd* to access the previous AMD processors.
NOTES:
- Long running application on the head-nodes may be killed by the system administrator
- Both batch and interactive usage of iceberg is controlled by the "Sun Grid Engine" job scheduling system.
- Refer to section on submitting jobs to iceberg for further details of the qsub command.