Path failover on HP blade with Emulex HBAs

[omasse]

I'm having SAN reliability issues running ESX 3.5.0.1 on all my HP bl480c blade servers, each have two Emulex LPe1105 HBAs. The HBAs are hardwired to an HP Virtual Connect module, which is plugged in my SAN on a Silkworm. The luns are mapped from an EVA8000 which has active/active controllers. All firmware and software revisions are recent.

To make a long story short, I tested failover scenarios by unplugging the HBAs, one at a time, using the Virtual Connect. The link does go down but path failover fails to work as expected: it switches once, but the failed HBA becomes unusable, even after it's been reconnected for a while. Doing rescans won't help... although there is a link up, the luns cannot be rediscovered. Only resetting the card using HBAnyware, or rebooting the ESX server, makes it functional again. If I do similar tests on the exact same hardware, but with Win2003, it works well. So I'm thinking this might be an issue with the lpfc driver.

Does anybody have a similar setup and have tested this? I would be glad to hear from you.

I did open a call on this, but it's been lagging so that's why I'm poking around to see if I'm alone in this boat.

Thanks

[weinstein5]

Welcome ot the forum - Have you changed the failover policy for each LUN from the default MRU (Most Rectly Used) to Fixed? Also, when they are reconnected can you manually switch paths?

Check out for more information -

[omasse]

Path policy is set to Fixed. And esxcfg-mpath doesn't consider the paths as OK when I plug back the HBA so I cannot switch back. Running esxcfg-rescan once the HBA is plugged back in does not make them come back either, as a matter of fact it removes them, as if they were no longer visible at all.

Thanks

[dwight · Accepted solution]

Use emulex firmware version 2.70A5




RHCE, VCP

Blog: http://computing.dwighthubbard.info

[omasse]

I said to myself "You've got to be kidding".... but I gave it a try with firmware 2.70a5 and it now works. I think it's the first time I actually have to downgrade a shipping firmware.

Kudos to Dwight. You made my day.

Thanks

[dwight]

That was exactly my reaction when VMware support told us that would fix the issue.

RHCE, VCP

Blog:

[Hommer]

I have a similar issue on a new blade "C" class enclosure as described here, but I'm not sure how to downgrade the firmware when ESX is installed to version 2.70A5. Any pointers?