Upcoming Maintenance for GACRC Computing Cluster

The University of Georgia’s Georgia Advanced Computing Resource Center (GACRC) has announced scheduled maintenance for the Sapelo2 computing cluster, set for March 26-27, 2024. This maintenance is essential for upgrading the cluster’s storage capacity and ensuring its continued performance and reliability.

Maintenance Details:

  1. March 23, 2024: UPS maintenance at the Boyd Data Center. During this period, all GACRC resources, including Sapelo2, xfer nodes, the teaching cluster, and Open OnDemand, will be unavailable, and cluster jobs will be suspended.
  2. March 26-27, 2024: Sapelo2 will undergo a significant upgrade involving the migration of the /scratch and /work file systems from the current 6-year-old DDN Lustre storage device to a new Lustre system. The new system boasts 1.8 PB of usable space on flash drives and 7.9 PB of usable space on spinning disks, offering enhanced storage capacity and improved support for increased I/O loads. This upgrade is necessary to accommodate the growing number of compute nodes in the Sapelo2 cluster.
    • Preparation: An initial copy of all data on the current Lustre system has already been made to the new system. During the maintenance, a final synchronization of files will be performed to ensure all data in /scratch and /work remains available post-migration.
    • Downtime: Maintenance will begin at 6 a.m. on March 26 and is expected to conclude by 6 p.m. on March 27, although it may extend if the file synchronization takes longer than anticipated. Sapelo2, Open OnDemand, and xfer nodes will be unavailable during this period. The teaching cluster will remain unaffected.
    • Job Scheduling: Jobs with requested walltime extending past 10 p.m. on March 25 will not start and will remain pending until after the maintenance. Any jobs still running at 6 a.m. on March 26 will be terminated.

For questions or assistance, users can contact GACRC via the Web form.

This maintenance is crucial for enhancing the storage capacity and performance of the Sapelo2 cluster, ensuring it meets the future needs of the UGA research community. Users are advised to plan accordingly to accommodate these scheduled downtimes.