NVMe-oF Datastore Inaccessible Due to Path Failover Failure

Products

• VMware vCenter Server

Issue/Introduction

An NVMe-over-TCP VMFS datastore becomes inaccessible, causing an outage. All VMs on the datastore hang, which require manual intervention. 

Environment

ESXi 8.x

 

Cause

The root cause is an issue with Asymmetric Namespace Access (ANA) on the storage array. Unlike SCSI, where the host can actively manage path states, the host in an NVMe environment is a passive consumer of path state information. When a path fails, the host depends on the storage array to report the new ANA Group state via the ANA Log Page. In this case, the storage array failed to properly change the path's state from a healthy state to a non-optimal state, which prevented the ESXi host from properly failing over to a standby path.

Resolution

VMware ESXi is functioning as expected. The resolution must come from the storage vendor.