Cluster migration to new infrastructure, 2026

The cluster migration is complete. Log in with the same username and password, and you may continue using the same IP address.
Categories:
Tags:

Logging in after migration

The head node hostname has changed: the old IP address has been reassigned to the new cluster’s head node, so you can use either the old or the new IP shown after login. Because this is a new head node, SSH will raise a host-key warning on first login. Clear the old key (replace <hpc_ip> with your cluster’s login IP):

ssh-keygen -R <hpc_ip>

or remove the corresponding line from ~/.ssh/known_hosts.

Note: any jobs and compute nodes still running over the migration weekend were terminated. Re-submit affected jobs.

What changed

  • Refreshed OnDemand compute options. New a/b instance groups with updated cost/performance. See Compute Pricing.
  • Read-optimized data path /mnt/s3 (early testing). A read-only path mirroring existing data to reduce read latency. See Storage Pricing.
  • New desktop partition for interactive workflows (VS Code, Jupyter, RStudio, DCV) without SSH tunneling. See Getting Started.
  • New nvme partition providing fast local scratch at /scratch. See Storage Pricing.
  • Archiving option to reduce storage costs. See Storage Pricing.

Backups

The new environment does not currently include automated backups. Guidance on backup procedures will be provided separately.

Support

For any issues, contact the HPC team via Teams.