TRID-20800: Trident 26.06.1 Fails to Mount/Unmount NVMe Volumes After Node Drain in Kubernetes
Issue
After upgrading NetApp Trident to 26.06.1 and performing node drain/maintenance operations in Kubernetes clusters using NetApp AFF-A90 storage running ONTAP 9.17.1P9, NVMe volume mount and unmount operations began to fail persistently across multiple nodes. The following symptoms were observed:
- Kubelet logs on affected nodes showed repeated MountDevice and UnmountDevice failures for specific PersistentVolumeClaims (PVCs), e.g.:
MountVolume.MountDevice failed for volume "pvc-xxxx" : rpc error: code=Internal desc=failed to stage volume: nodevice found for the given namespace
UnmountDevice failed for volume "pvc-xxxx" : rpc error: code=DeadlineExceeded desc=context deadline exceeded
- Host kernel logs (dmesg) showed repeated NVMe connection errors:
nvmeX: ConnectInvalidDataParameter, subsys nqn "nqn.1992-08.com.netapp:sn.xxxxx:subsystem.x.x.x.x_xxxxx"
nvmeX: Failed reconnect attempt N
nvmeX: Reconnecting in 10 seconds...
- nvme list-subsys showed controllers stuck in “connecting” state for NetApp subsystems, e.g.:
+- nvme0 tcp traddr=10.x.x.x trsvcid=4420 connecting
+- nvme1 tcp traddr=10.x.x.x trsvcid=4420 connecting
- Affected nodes could not mount or unmount NVMe volumes, causing pods to get stuck, volume attachments to remain in a terminating state, and widespread disruption to workloads (e.g., CI/CD pipelines).
- The issue was observed globally across all nodes in the customer’s environment after node drain events.
