To minimize disruption to users, ICDS utilizes a combination of full outages and rolling maintenance to implement critical system updates and repairs: 

• Full Outages: During a full outage, the cluster is generally unavailable to users. The entire cluster is taken offline for large scale updates and fixes, including data center maintenance.  

• Rolling Maintenance: During rolling maintenance, the cluster remains available to users. Small portions of the cluster are taken offline, serviced, and then returned to production, until the entire cluster has been updated. To minimize the reduction in available resources, rolling maintenance may last up to two weeks. 

 Outage details are published here as they become available prior to an outage. 

Tentative Outage Schedule 

All outages begin at 5:00 a.m. on the date noted. Full outages are expected to last less than 24 hours. The outage schedule is subject to change. 

• Tuesday, March 17, 2026: ROLLING MAINTENANCE

• Tuesday, May 12, 2026: ROLLING MAINTENANCE

• Tuesday August 18, 2026: FULL OUTAGE

• Tuesday, October 20, 2026: ROLLING MAINTENANCE

• Thursday January 7, 2027: FULL OUTAGE

 Note: At the outset of a rolling maintenance window, users may notice a brief interruption in interactive desktop and submit node access. During rolling maintenance the Roar cluster remains functional with a small reduction in available resources. 

Planned system outages are conducted in accordance with the Roar supercomputer Service Level Agreement Terms and Conditions.