FC-NVMe Host Logs In but No Namespace Appears
A PowerStore 5200T is configured for NVMe-oF over Fibre Channel. A new Linux database host has two 32 Gb FC HBAs and uses nvme-cli. The SAN administrator creates two single-initiator, single-target zones: host HBA1 to array node A FC port 0, and host HBA2 to array node B FC port 0. The array ports are online and the switch shows the host logged in. In PowerStore Manager, the host initiator is present and the NVMe namespace is mapped to the host, but the host's nvme list shows no namespaces and nvme connect fails with a transport address error. The host OS is configured for FC-NVMe, and the array is running PowerStoreOS 3.2. The customer suspects the array. The zoning uses the correct WWPNs.
1. What is the most likely configuration gap, and what would you check first on the array and host?
Read the worked answer
The likely gap is that the host was added to PowerStore as a SCSI FC host rather than an NVMe-oF host, or the NVMe host object is missing the correct host NQN. A physical FC login only proves that the HBA and switch see the array port; it does not prove that an NVMe controller and namespace were exposed to that host. If the host object type is wrong, mapping the namespace to it will not create an NVMe subsystem for the host.
Check the host object type in PowerStore Manager and confirm that the NVMe initiator includes the host NQN and the correct WWPNs. On the host, confirm that the nvme-fc transport is loaded, that the initiator is using the intended NQN, and that discovery uses the array's FC port WWPN as the transport address. Also confirm the array FC port is enabled for NVMe-oF, not only SCSI.
2. Describe the correct sequence to add the host and map the namespace for FC-NVMe.
Read the worked answer
First, verify that the PowerStore FC ports support and are configured for NVMe-oF. Create or edit a host object of type NVMe, supplying the host NQN and the two HBA WWPNs. Place the host in the appropriate host group if one is used, and confirm that the initiators show as logged in. Then map the NVMe namespace to that host or host group.
On the Linux host, run nvme discover against the array port transport address, then nvme connect for the discovered subsystem. Confirm that nvme list shows the namespace and that multipath is active across both HBAs. If discovery succeeds but connect fails, compare the NQN and subsystem NQN in the connect command with what PowerStore presents.
3. If the array shows the initiator logged in but no NVMe controller, what does that tell you?
Read the worked answer
It tells you that the physical fabric and zoning are probably working, but the NVMe session has not been established. The host HBA has logged in to the FC port, yet no NVMe controller exists for that host from the array's perspective. That points away from a bad WWPN zone and toward host type, NQN, discovery, or namespace mapping configuration.
Check the host object type and NQN first, then the namespace mapping and the array's NVMe-oF port configuration. On the host, confirm the nvme-fc module is loaded and run discovery with the correct transport address. If the host is still logged in as a SCSI initiator, remove the incorrect host object or recreate it as an NVMe host before mapping the namespace.