Dell PowerStore direct-attach FC: Emulex LPe37000/LPe38000 initiator not discovered on ESXi 8.0.2 / 8.0.3
Environment
- Dell PowerEdge R770, Emulex LPe37000/LPe38000 FC HBA (2 ports)
- VMware ESXi, tested on both 8.0.2 (build 23305546) and 8.0.3 A01 (build 24022510)
- Dell PowerStore, PowerStoreOS 4.0.0.1
- Direct-attach FC (host cabled straight into PowerStore, no FC switch)
Symptom
Everything looks fine at first glance:
- HBA port LED is green, link is up
- PowerStore front-end port shows Online
- ESXi Storage Adapters show both vmhba ports as Online with valid WWPNs
But:
- Targets: 0 on both adapters – no remote/target port discovered
- Devices tab is empty – no LUN, no enclosure visible
- In PowerStore Manager, Add Host → Select Host’s Initiators shows: „No initiators have been added or discovered.“
So: physical link is up, but the initiator never registers on the storage side, and this was identical on both ESXi 8.0.2 and 8.0.3 A01.
What we ruled out
- FC zoning – not applicable, this is direct-attach, no switch in the path
- Cabling / SFP / link speed – link is up and stable
- Missing/incompatible driver –
lpfcis loaded, adapters are recognized - Direct-attach FC-NVMe interference – disabling NVMe on the driver (
lpfc_enable_fc4_type=1) made no difference
Root cause
Confirmed VMware bug, PR 3632380:
„An ESXi host with an Emulex LPe37102/LPe38102-D FC HBA that is directly connected to a storage appliance cannot discover the LUNs.“
When these HBAs are direct-attached to a storage array, the PLOGI response and the subsequent PRLI request can arrive at the target out of order. The storage array interprets this as a protocol violation and refuses the login, so the host never registers as an initiator — even though the link itself is healthy.
Fixed in lpfc driver 14.4.0.42-35, shipped with ESXi 8.0 Update 3i (build 25205845, released 24 Feb 2026).
Fix
Upgrade vCenter Server first, then the ESXi hosts, to 8.0 Update 3i or later. vCenter must be on the same or a higher version than the ESXi hosts it manages, so the vCenter upgrade has to happen before touching the hosts.
In our case, the vendor’s customized ESXi image incorporating the Update 3i base only became available some time after the fix itself was released upstream — worth keeping an eye on your OEM’s image catalog, since the base VMware fix doesn’t help until it lands in the image you’re actually able to deploy.
The image that finally contains the fix: VMware-VMvisor-Installer-8.0.0.update03-25205845.x86_64-Dell_Customized-A10.iso (build 25205845). Plan is to upgrade vCenter to 8.0 Update 3i first, then deploy this image to the affected host and verify the initiator is discovered automatically in PowerStore Manager.
Takeaways
- A green link LED / „Online“ status only confirms the physical layer (FLOGI). It says nothing about whether the PLOGI/PRLI handshake with the target actually succeeds.
- Direct-attach FC is a less common topology than switched fabric, and bugs like this one show up specifically there.
- Before chasing zoning, cabling, or storage port configuration, check the current
lpfc/qlnativefcdriver release notes for your exact HBA model and ESXi version — a specific hardware + patch-level combination can produce exactly this kind of „phantom“ behavior. - Always verify the actual running ESXi build on the host itself — the installer image name doesn’t always match what ends up deployed.