Due to maintenance work on the electrical infrastructure of the Kältezentrale, the Fritz and the Helma cluster will be down on
Wednesday, September 30th, starting at 11:00
We expect normal operation to resume by the evening. As usual, jobs that would collide with the downtime will automatically be postponed until the downtime is over. Most frontends and fileservers will be available through the downtime. However, short interruptions are possible here, too.
UPDATE 2026-09-30 @ 18:00 for Fritz
- Most parts of Fritz are back in operation
- Two racks still have issues with their power feed
- Slurm on Fritz currently probably cannot send email notifications
UPDATE 2026-09-30 @ 18:00 for Helma-GPU
- there is not ETA yet; we have to check recent announcements from NVIDIA concerning privilage escalation bugs
UPDATE 2026-09-30 @ 18:00 for Helma-CPU
- Helma-CPU had its power restored today (see the other announcement at https://hpc.fau.de/2026/09/28/uncheduled-downtime-of-helma-cpu-from-september-28/)
UPDATE 2026-09-30 @ 19:00 for Helma-CPU
- about 40% of Helma-CPU are back
UPDATE 2026-09-30 @ 20:30 for Helma-GPU
- about half of the H100 and half of the H200 nodes of Helma-GPU are back
UPDATE 2026-09-30 @ 22:00 for Helma-GPU
- Helma-GPU is fully back
Two racks of Fritz are not back yet, and Helma CPU is not operating at full capacity yet. Everything else is back, and batch processing is working normally. We will thus now mark this as resolved.