Skip to content
System Status: All systems are operational • Services are available and operational.
Click for detailed status

Slurm upgrade to 25.05

Slurm will be upgraded in the morning or Jan. 13 2026.

During the upgrade:

  • the Slurm partitions (queues) may be down, meaning no new jobs will be started
  • various slurm commands (sbatch, srun, sacct) may be unresponsive

13.01.2026 14:30

The upgrade to Slurm 25.05 is complete.

21.01.2026 15:00

The myjobs script has been fixed to again show the correct CPU utilization, which was wrong due to a bug in Slurm (see known issues below).

Known issues:

  • Programs linked to old Slurm versions (23.02 or lower) will not function correctly. Please contact us if you suspect your program is affected by this Invalid Protocol error.
  • For CentOS 7 compatibility containers a fix was implemented in the afternoon of 14.01.2026.

  • Job cancellations: your job will be CANCELLED if its startup takes over 120s or if you replace the shell by using the exec command. This issue manifests most frequency in cases where the .bashrc or .bash-profile contains commands that cause a long loading time. For instance initializing a conda environment on /cluster/work

  • Due to a change in the new Slurm version, some jobs that requested more than 60 hours of runtime and contain an srun call in the jobscript are terminated already after 60 hours. All newly submitted jobs are not affected by the bug, but some already running jobs might terminate still after 60 hours if they are affected by the bug.

  • There is a bug in the new Slurm version that affects jobs which have a srun call in the jobscript. For those jobs, the total CPU time is reported as 0. This affects the myjobs command and the get_inefficient_jobs service such that CPU utilization for those jobs is shown as 0.0%, which is wrong. We are working on a fix for this issue.