Hazel Cluster Planned Outage Information
- /home and /usr/local file systems will transition from NFS file systems to native GPFS file systems on login and compute nodes
- /home quota will increase from 1GB to 15GB per USER
- iNode quotas will be enforced - exceptions will be allowed with demonstrated need
- /home - 10K per USER
- /usr/local - 250K per GROUP
- /share - 1M per GROUP
- LSF will be updated to current version 10.1.0.15
- LDAP transition for login and compute nodes will be completed
- Changes made via the Research Services web application will be effective within approximately 10 minutes - rather than being implemented once a day at 6:30pm
- All compute nodes and login nodes other than login01 have been running LDAP for some time already with no significant issues
- Friday 19 December - 8:00am LSF queues off - no new jobs scheduled
- Friday 19 December - 5:00pm
- All running jobs suspended
- All logins disabled
- Final sync of data from /home and /usr/local initiated from NFS storage to GPFS storage
- Saturday 20 December
- Remove NFS /home and /usr/local from cluster
- Mount new GPFS /home and /usr/local on all cluster nodes
- Update LSF to 10.1.0.15
- Implement new USER and GROUP iNode quotas
- Implement new USER quotas on /home
- Complete LDAP transition work for login and compute nodes
- Saturday 20 December - 6:00pm
- Restart LSF
- Unsuspend jobs that had been running at 5pm Friday - most will likely fail
- Restart queues scheduling new jobs
- Enable logins
Additional GPFS storage capacity is being added to the Hazel cluster. A portion of this new space will be used for expanded /home and /usr/local file systems.
In order to make the transition to new /home and /usr/local file systems a complete outage is required. HPC staff plan to use the required outage as opportunity to perform some other maintenance work more efficiently and to avoid potential other disruptions.
Significant Work Planned for 19-20 December 2025