Scheduled Maintenance

The cluster undergoes scheduled maintenance 2-3 days twice a year, typically in June and December. During these maintenance windows, we upgrade hardware components and software including Slurm, CUDA, Ubuntu, and other system packages. We strive to keep software relatively up to date, but some upgrades require the cluster to be shut down completely.

Specific dates and details for upcoming maintenance windows are announced in the public Slack channel with at least one week of advance notice.

For real-time cluster health status between maintenance windows, see the Cluster Status page.