-
Bug
-
Resolution: Unresolved
-
Minor
-
None
-
4.14, 4.16
-
Quality / Stability / Reliability
-
False
-
-
None
-
None
-
None
-
None
-
None
-
None
-
None
-
None
-
None
-
None
-
None
-
None
-
None
-
None
Summary
Checkpointing a pod in OpenShift fails when it has TCP sockets in the TCP-CLOSED state, causing the pod to terminate unexpectedly.
Steps to Reproduce
- Deploy a pod with an app that opens and closes a TCP connection.
- Wait until the socket enters the TCP-CLOSED state.
- Run criu dump or trigger OpenShift checkpoint.
- Observe the pod is terminated.
Expected Behavior
Pod should remain running. TCP-CLOSED sockets should be handled gracefully during checkpoint.
Actual Behavior
Pod is terminated during checkpoint due to unhandled TCP-CLOSED socket state.
Impact
Blocks live migration, debugging, and resilience features relying on checkpoint/restore.