Upcoming Maintenance - November 16-22, 2018

GACRC is planning a maintenance, beginning at 5 p.m. Friday, November 16 and ending at 8 a.m., Thursday, November 22.

The Sapelo cluster will be decommissioned during the maintenance and will no longer be available.

The Sapelo2 cluster will be unavailable beginning at 5 p.m. on Friday, November 16 and should be available again no later than 8 a.m. on Thursday, November 22.

Please be aware that, except for the teaching cluster, all GACRC compute and storage resources, including login nodes, transfer nodes, and compute nodes will be affected by this downtime. This maintenance will require that we power down all storage and cluster systems, as we will be bringing new systems online, and updating operating systems and various clients on all compute nodes and storage servers.

As a result, all jobs still running when the maintenance begins at 5 p.m. on Friday, November 16, will be lost. Please note that we have implemented a standing reservation on the queueing system that will only dispatch jobs that request a walltime that will not extend beyond the beginning of the maintenance window.

After this maintenance window, Sapelo2 will have a new Lustre scratch file system, which has more capacity and offers better performance than the current /lustre1. All files in /lustre1 at the start of the maintenance window will be copied into the new Lustre scratch storage system, but file retention and purging policies will be implemented soon thereafter and we will communicate those policies in a separate message.

Sapelo will be fully decommissioned. All remaining nodes on Sapelo will be migrated into Sapelo2 during the maintenance window.

Please note that the teaching cluster and its transfer node will not be affected by this maintenance window.

IMPORTANT DATES TO REMEMBER:

Friday, November 16th, 2018 at 5PM - start of maintenance window

Saturday, November 17th, 2018 - the Sapelo cluster will be fully decommissioned

Thursday, November 22nd, 2018 at 8AM - end of maintenance window

As always, we understand the inconvenience experienced by the maintenance, but these operations are imperative to maintaining stable systems and expanding the cluster resources. We thank you for your understanding and patience.