Back to all scenarios
Scenario #329
Storage
Kubernetes v1.20, iSCSI volumes
Pod Gets Stuck Due to Ghost Mount Point
Pod failed to start because the mount point was partially deleted, leaving the system confused.
Find this helpful?
What Happened
After node crash, the iSCSI mount folder remained but device wasn’t attached. New pod couldn’t proceed due to leftover mount artifacts.
Diagnosis Steps
- 1CSI logs: mount path exists but not a mount point.
- 2mount | grep iscsi — returned nothing.
- 3ls /mnt/... — folder existed with empty contents.
Root Cause
Stale mount folder confused CSI plugin logic.
Fix/Workaround
• Manually deleted stale mount folders.
• Restarted kubelet on affected node.Lessons Learned
Mount lifecycle must be cleanly managed.
How to Avoid
- 1Use pre-start hooks to validate mount point integrity.
- 2Include cleanup logic in custom CSI deployments.