Skip to main content
Have a request for an upcoming news/science story? Submit a Request

Home Filesystem Maintenance - All Clusters

Conte has been returned to normal operations as well now. This concludes the home directory maintenance on all systems.

Update: September 27, 2016 11:55pm

All systems other than Conte have been successfully returned to normal operations with the new home directory filesystem. Work continues at this point on Conte to ensure the Phi accelerators are properly reconfigured.

Carter has also been given a new scratch filesystem during this maintenance. This should alleviate some of the problems with the previous scratch filesystem on Carter. For more details, please see the Carter-specific announcement on this topic: New Carter Scratch Filesystem

Reminder:

This is a reminder of the Home Filesystem Maintenance taking place next week on Tuesday, September 27th.

Details below.

Original Message:

All of the research clusters (Carter, Conte, Hammer, Hansen, Hathi, Peregrine1, Radon, Rice, Scholar, Snyder, and WinHPC) as well as some other minor systems will be unavailable beginning at Tuesday, September 27th, 2016 at 7:00am EDT, for scheduled maintenance. All clusters other than Conte will return to full production by 11:59pm.

Conte will return to partial capacity by that time, but will not return to full production until the following day. Many Conte nodes will remain offline and gradually be returned to service over the following 12-24 hours to allow for power reconfiguration in the data center. Please see the separate article on Conte: Conte Cluster Maintenance.

During the large all-systems maintenance Tuesday, the /home filesystem used by all Research Computing systems will be replaced by a new filesystem. The new filesystem will be based on DDN's GRIDScalar technology and running on new hardware dedicated exclusively to Research Computing home directories.

All files on the existing /home filesystem will be migrated to the new system during the maintenance window and prior to any of the clusters returning to service.

In the coming weeks, any jobs which request a walltime which would take them past Tuesday, September 27th, 2016 at 7:00am EDT will not start and will remain in the queue until after the maintenance is completed.

Originally posted:
Last updated: