(2026-09-07 19:00) The maintenance was completed around 18:40.
2026-09-07
As a result of the GPU driver update performed during the summer maintenance, an issue has occurred on multiple nodes where GPUs enter an error state. Therefore, emergency maintenance will be carried out to roll back the GPU driver and related settings.
1. Schedule
2026-09-07 10:35-15:00 (approx.): New jobs will be prevented from starting.
2026-09-07 15:00 (approx.) and onward: The compute nodes will be rebooted sequentially (running jobs will be interrupted). Service will be resumed on each node as soon as the work on that node is completed.
* The completion time may change depending on the progress of the work.
2. Impact
During the maintenance, new jobs will not enter the running state (including jobs with reservations).
Running jobs will not be interrupted until 15:00, but will be terminated after 15:00.
Compensation related to running jobs affected by this maintenance will be added to this article later.
Compensation regarding the impact of this maintenance on existing reservations will be communicated individually.
3. Main work items and purpose
Rollback of issues caused by the system update