Some, but not all, of the ESXi hosts in the cluster connected to VMFS datastores backed by NVMe over TCP storage devices may have frequent path loss events occur.
"Resetting controller" messages along with dead path(s) messages may be intermittently seen in vmkernel.log:
2025-08-12T00:26:51.653Z In(182) vmkernel: cpu50:2467849)NVMEDEV:8245 Resetting controller 266 (nqn.X:Y.Z#vmhba65#.#.#.#:4420)...2025-08-12T00:26:55.184Z In(182) vmkernel: cpu50:2097840)HPP: HppNvmeUpdateNamespaces:535: Marking paths dead - controller:nqn.X:Y.Z#vmhba65##.#.#.#:44202025-08-12T00:26:55.184Z In(182) vmkernel: cpu50:2097840)HPP: HppPathGroupMovePath:688: Path "vmhba65:C0:T2:L54" state changed from "active" to "dead"2025-08-12T00:26:55.185Z In(182) vmkernel: cpu50:2097840)HPP: HppNvmeUpdateNamespaces:535: Marking paths dead - controller:nqn.X.Y:flasharray.Z#vmhba65##.#.#.#:4420vobd.log will report the path(s) as degraded and down as well:
2025-08-12T00:26:55.187Z In(14) vobd[2097958]: [psastorCorrelator] 3684877445497us: [esx.problem.storage.redundancy.degraded] Path redundancy to storage device eui.## degraded. Path vmhba65:C0:T2:L54 is down. Affected datastores: "DATASTORE_NAME".ESXi
NVMe over TCP storage configuration
A NVMe over TCP adapter binds to vmkernel interface(s) for connectivity to storage. If there is a duplicate vmkernel IP assigned between ESXi hosts and bound to their respective NVMe over TCP adapter, this will introduce instability into the NVMe over TCP storage connections/sessions.
Review IP addresses across the ESXi cluster to ensure there are no duplicate IP addresses in use in respect to vmkernel interfaces bound to the NVMe over TCP adapter.
Tools that can be used: