Back to all scenarios
Scenario #369
Storage
Kubernetes v1.24, external CSI
CSI Driver Memory Leak on Volume Detach Loop
CSI plugin leaked memory due to improper garbage collection on detach failure loop.
Find this helpful?
What Happened
Detach failed repeatedly due to stale metadata, causing plugin to grow in memory use.
Diagnosis Steps
- 1Plugin memory exceeded 1GB.
- 2Logs showed repeated detach failed with no backoff.
Root Cause
Driver retry loop without cleanup or GC.
Fix/Workaround
• Restarted CSI plugin.
• Patched driver to implement exponential backoff.Lessons Learned
CSI error paths need memory safety.
How to Avoid
- 1Stress-test CSI paths for failure.
- 2Add Prometheus memory alerts for plugins.