Werk #20108: NFS mounts: report unreadable mounts as UNKNOWN instead of CRIT "Hanging"
| Component | Checks & agents | ||||||
| Title | NFS mounts: report unreadable mounts as UNKNOWN instead of CRIT "Hanging" | ||||||
| Date | Jul 22, 2026 | ||||||
| Level | Trivial Change | ||||||
| Class | Bug Fix | ||||||
| Compatibility | Compatible - no manual interaction needed | ||||||
| Checkmk versions & editions |
|
When the Checkmk Linux agent ran as a non-root user that lacked read permission on an NFS mount point, the NFS mount service went to CRIT with "State: Hanging" — even though the NFS server was fully reachable and the mount was working (the NFS IO stats service reported OK for the same mount at the same time).
The cause was twofold:
-
The agent probes each NFS mount with
waitmax ... stat -f <mountpoint>. It treated every non-zero exit code identically and reported"state": "hanging"— so an immediate "permission denied" (EACCES) was indistinguishable from a genuine 5-second timeout (a real hang). -
Even when a mount was correctly identified as inaccessible, the check plugin mapped the "Permission denied" state to CRIT.
This has been fixed:
-
The Linux agent now inspects the output of
statand, when it fails because of missing read permission, reports"state": "Permission denied"instead of"state": "hanging". Genuinely hanging mounts continue to be reported as hanging. -
A mount that the agent cannot read due to missing permissions is no longer reported as CRIT. Its health cannot be determined by the agent, so the NFS mount service now reports UNKNOWN with "State: Permission denied".
To benefit from the improved agent behaviour, the agent on the monitored
host has to be updated in addition to the Checkmk site. Sites running an
older agent (which still sends "state": "hanging") are unaffected and
continue to behave as before. No manual configuration change is required.