NVMe RDMA path stays connecting after rebooting E-Series controller
Applies to
- E-Series EF80
- SANtricity OS 12.10R or earlier
- RoCE (RDMA over Converged Ethernet)
- NVMe over RoCE
Issue
- Some NVMe/RDMA paths stay in connecting after a controller reboot while other paths on the same subsystem stay live
nvme list-subsyscommand shows some NVMe/RDMA paths asconnecting
Example:
+- nvme6 rdma traddr=<TARGET_ADDRESS>,trsvcid=4420,host_traddr=<HOST_ADDRESS> live
+- nvme7 rdma traddr=<TARGET_ADDRESS>,trsvcid=4420,host_traddr=<HOST_ADDRESS> live
+- nvme8 rdma traddr=<TARGET_ADDRESS>,trsvcid=4420,host_traddr=<HOST_ADDRESS> connecting
+- nvme9 rdma traddr=<TARGET_ADDRESS>,trsvcid=4420,host_traddr=<HOST_ADDRESS> live
- Linux host log repeats messages showing I/O queue creation is rejected:
Example:
nvme nvme8: creating 4 I/O queues.
nvme nvme8: Connect rejected: status 10 (stale conn).
nvme nvme8: rdma connection establishment failed (-104)
