We will retire several of SeaWulf's older compute resources and their associated queues as part of scheduled maintenance beginning Monday, November 9th. The following resources will be retired:
 

  • 28-core queues (Intel Haswell 28-core nodes)

  • gpu queues (NVIDIA Tesla K80 GPUs)

  • p100 queues (NVIDIA Tesla P100 GPUs)

  • Login nodes login1 and login2

Users should continue using milan1, milan2, and xeonmax to access SeaWulf and submit jobs to the queues available through these newer login nodes.

This change affects compute and login resources only. User home, project, and scratch data will not be affected.

If you are unsure which queue is the best replacement for your workload, or if you encounter problems moving an existing workflow, please contact HPC Support and we will be happy to help.

To facilitate retiring these older compute resources and to consolidate the remaining SeaWulf hardware in the campus datacenter, we will perform cluster maintenance starting at 9 am on Monday, November 9th and finishing by the end of business on Tuesday, November 10th. 

During this maintenance period, all SeaWulf queues, login nodes, and the GPFS file system will be unavailable. Any jobs still running when maintenance begins will be terminated. Users should plan their job submissions accordingly and ensure that any long-running computations are completed or checkpointed before the maintenance window begins.

We thank you for your patience while these necessary updates are made.

Completed
Your jobs Jobs still running when the work starts are cancelled. Jobs waiting in the queue stay there and start once the cluster is back. To keep a job, give it a time limit that ends before the window: Choosing a time limit.

Add to my calendarMonday, November 9 to Tuesday, November 10

← All announcements