Hi, everybody,
I'm seeing some weird behaviour from an Emulex FC HBA (sold as HP 81E, but is in fact Emulex AJ762B). I have three of these cards installed in three all the same Supermicro boards and all they show the same behaviour, so at least faulty hardware can be ruled out.
First I had vanilla ESXi 5.5u1 on all three boxes. Consistently:
1. When the motherboard boots, the Emulex BIOS is accessible, meaning the card itself works and is seen by the system BIOS.
2. While the ESXi 5.5 still boots (no VMs loaded, no even ping yet to the server) the board connects to the SAN. The SAN reports channel is up at 8 Gbps.
3. As soon as the ESXi comes up and is reachable by the network, the SAN reports the channel has disconnected.
4. vSphere does not show the HBA at all.
5. Even more weird, lspci does not show the HBA at all - as if it is not plugged in the bus. However, in p.1 that the card's BIOS was accessible before ESXi began booting.
6. Kernel log is totally silent about anything related to the HBA.
Originally my ESXi showed that the available driver is lpfc_10.0.100.1-1vmw.550.0.0.1331820.
Knowing the HBA is listed as supported in the HCL, I decided to see if a driver update may help.
- HP site pointed me to a page in VMWare site with this driver: lpfc-10.0.725.203-1OEM.550.0.0.1331820 (released December 2013)
- Emulex site pointed me to a page in VMWare site with this driver: lpfc-10.0.727.44-1OEM.550.0.0.1331820 (released January 2014)
Tried both, none of them helped in any way.
Has anybody seen anything like this before? No matter how weird, it seems that the ESXi kernel somehow kills the HBA in such a way that it does not even show up on the PCI bus... or maybe somehow blocks the access to its address space (or whatever)? The boxes run the latest available BIOS from the manufacturer; going through them did not reveal any obvious setting that may affect the way the HBA is seen on the PCI bus.
Any ideas are highly appreciated.
WWell,
Assen
