Compute Cluster Policies
1. Access to the cluster.
- 1.1. The privilege of access to the CCC Compute Cluster is by request only. Requests may be sent to firstname.lastname@example.org. Requests should include the reason for the request, including the department and the type of research and the type of computation to be run on the cluster. This does not need to be described in great detail, but if it appears that user needs will be better met on a multi-CPU shared-memory system the person making the request may be directed appropriately.
- 1.2. Cluster access privileges may be withdrawn for any reason, such as repeated failure to comply with these policies, interference with other users' programs, frivolous load on the systems, termination of association with WPI, or other situations. The ruling of the HPC manager is final.
- 1.3. Cluster use must adhere to the CCC Acceptable Use Policy, including security and password guidelines.
2. Use of the cluster.
- 2.1. Programs run on the cluster must be submitted through the gridengine queue processing system. This will ensure that compute jobs do not interfere with one another.
- 2.2. If an interactive session is required, you must use the qrsh command. This submits a request through gridengine and ensures that the interactive session is opened on an idle compute node. If you need to log into a node on which your own compute processes are running, you must use ssh, but do not use ssh for any other reason.
- 2.3. Processes not started by gridengine interfere with efficient scheduling of the cluster and will be terminated if discovered. Repeated activity of this sort may result in loss of privilege of access to the cluster. Using qrsh, which schedules a node, but then not running a useful job on the node will block the node availability for scheduling. Repeated activity of this sort may result in loss of privilege of access to the cluster.
3. Storage on the cluster.
- 3.1. The disk space on ccc-cluster is for use with the cluster only. Do not use it to store non-cluster-related files.
- 3.2. The disk array on ccc-cluster is NOT backed up. Any important files should be copied elsewhere.
- 3.3. During system upgrades, effort will be made to preserve user files, but it may not be possible. Again, any irreplaceable files should be backed up elsewhere.
4. Software requests.
- 4.1. Requests for software packages may be mailed to email@example.com.
- 4.2. Requests for assistance compiling software may also be addressed to firstname.lastname@example.org.
5. Downtime and maintenance.
- 5.1. Downtime will be announced as much in advance as possible, up to a week beforehand. Effort will be made to accomodate users with long-term jobs running on the cluster.
6. Mailing lists.
- 6.1. The email@example.com mailing list has been created for discussion of the CCC Compute Cluster. It can also be used for requests for assistance or other cluster-related questions. It includes all current cluster users. The cluster manager will also use the ccc-hpc mailing list to send out announcements about the cluster, including announcements about cluster downtime and maintenance and new software packages on the cluster.
Last modified: Jul 23, 2009, 15:26 EDT