Scheduled Maintenance¶
ARC admins schedule maintenance periods for each cluster in order to make large, necessary changes without interrupting users. When a maintenance period approaches, users will receive notification via email including details such as which cluster will be inaccessible, when, and for how long. Users will also be given notice when the cluster is available again.
Schedule¶
Orion is typically down for maintenance once a month on the 4th Wednesday of every month. A reminder is sent to users about the scheduled downtime at least 48 hours beforehand.
Hercules is typically down for maintenance once a month on the 4th Wednesday of every month. A reminder is sent to users about the scheduled downtime at least 48 hours beforehand.
Ptolemy maintenance is aimed to take place at least twice a year after both the fall and spring semesters are out of session. More information on specific maintenance periods will be communicated to users beforehand.
Atlas is planned for maintenance on the 2nd Tuesday of each quarter.
2026 Atlas Maintenace Schedule:
[x] 2026-01-13 Tuesday, 6am - 6pm central time
[x] 2026-04-14 Tuesday, 6am - 6pm central time
[ ] 2026-07-14 Tuesday, 6am - 6pm central time
[ ] 2026-10-13 Tuesday, 6am - 6pm central time
Morrill is planned for maintenance twice yearly.
Software updates¶
Sometimes these maintenance periods include updates to critical cluster software such as SLURM and its underlying dependencies which has the potential to break user-built software. Often times, a simple rebuild will fix the issue. If you experience problems running software after a maintenance period and a rebuild does not help, please submit a Helpdesk ticket.
Unscheduled Maintenance/Downtime¶
While we try to prevent it as much as possible, some things are out of our hands. In the event that a cluster goes down, users will be notified via email as soon as possible about the downtime, and will receive notification when the issue has been resolved and the cluster is back online.