NVMe-OF attached storage is not visible to host after configuring NVME adapter Policy for "FC-NVMe Initiator"

Products

• VMware vSphere ESXi

Issue/Introduction

This article addresses NVMe over Fibre Channel (NVMe/FC) connectivity issues where an ESXi host successfully establishes a fabric login but fails to detect or mount storage namespaces (datastores). Although the nfnic driver is active and the host communicates with the SAN fabric, the vSphere Client fails to display devices, paths, or controllers, leaving the host "blind" to the storage.

Environment

VMware vSphere ESXi
Cisco UCS (Virtual Interface Cards/VIC)
NVMe over Fibre Channel (NVMe/FC) storage arrays (e.g., Pure Storage, NetApp, Dell)

Cause

When the SAN fabric and hardware are functioning (confirmed via successful FLOGI and queue negotiation), the issue typically resides at the storage array's masking/presentation layer. In NVMe/FC environments, storage arrays often require the mapping of both the WWPN and the Host NQN (NVMe Qualified Name) to the NVMe Subsystem/Host Group to authorize namespace access. If the array is not configured to present storage to that specific host's NQN, the host will not "see" the storage.

 

Resolution

Follow these steps to diagnose and correct storage presentation issues:

1. Retrieve the Host NQN: In NVMe/FC, modern storage arrays require the Host NQN for security and namespace masking.

2. Run on the ESXi host:

esxcli nvme info get

3. Locate the Host NQN field. Provide this string to your storage administrator to add to the NVMe Subsystem definition on the array.

4. Verify Storage Array Masking: The storage administrator must confirm the following on the array:

• The host is explicitly added to the correct NVMe Subsystem or Host Group.
  
  
• Both the WWPNs and the Host NQN are associated with that specific host object.
  
  
• Storage volumes are mapped to that host object as NVMe Namespaces.

5. Force a Deep Storage Rescan: Once the array configuration is updated, force ESXi to rediscover the namespaces:

• Run a comprehensive rescan:
  
  esxcli storage core adapter rescan --all
  
  
• Verify namespace visibility: esxcli nvme namespace list
  
  
• Check for Unresolved VMFS Volumes: If the namespace list shows devices but they do not mount as Datastores, ESXi may be detecting them as snapshot LUNs or signature conflicts.
  
  
• Check for volumes requiring manual intervention:
  
  esxcli storage vmfs snapshot list
  
  
• If listed, refer to
  
  Troubleshooting LUNs detected as snapshot LUNs in vSphere
  https://knowledge.broadcom.com/external/article/323142/troubleshooting-luns-detected-as-snapsho.html

Additional Information

If the steps above do not resolve the visibility issue:

Vendor Engagement
Have user contact Cisco TAC and their Storage Array Vendor support.
Provide the symptom report: "The HBA is recognized at the driver level, but discovery fails for NVMe subsystems" to narrow the focus to fabric provisioning or storage masking.

Joint Vendor Call
If independent troubleshooting does not resolve the issue, recommend coordinating a joint vendor call for full collaboration among all parties.