-
Notifications
You must be signed in to change notification settings - Fork 0
Open
Labels
slurmIssues related to Slurm workload manager itselfIssues related to Slurm workload manager itself
Description
For an unknown reason, Slurm does not properly clean cgroupv2 hierarchy and it ends up being full and fail with no space left in cgroup virtual filesystem.
A workaround is to clean empty cgroups on compute nodes, with:
On Debian:
# clush -bw cn[1-4] 'find /sys/fs/cgroup/system.slice/cn?_slurmstepd.scope -depth -type d -exec rmdir {} +'
On Rocky:
$ clush -bw cn[1-4] 'find /sys/fs/cgroup/system.slice/slurmstepd.scope -depth -type d -exec rmdir {} +
Metadata
Metadata
Assignees
Labels
slurmIssues related to Slurm workload manager itselfIssues related to Slurm workload manager itself