HPE Storage Users Group

We upgraded from 5.1 U1 vsphere to 5.5 U1 and noticed our QLogic HBAs would disappear or not always log in. The common denominator is inform 3.1.2 MU3. After pulling teeth with HP support for 2 weeks we get answer that VMware 5.5 needs inform 3.1.3. Seems in VMware 5.5 VMware went to using the native HBA drivers without putting any translation wrapper on them. Not sure how this broke things and am waiting on HP to deliver final root cause, but they did say they are getting this issue with other customers as well.

Do you know of any workaround? New driver/firmware for the qlogic-hbas? Anything in the pipe from VMware?

_________________
When all else fails, read the instructions.

We were told the 3par needs to be 3.1.3 before upgrading vmware. I will post more once I get the official write up they owe me in a few days.

As you mentioned VMware has introduced an optional driver model in ESX server 5.5,
called "Native Device Driver architecture" that gets rid of the vmklinux shim layer between the linux based driver and the vmkernel.

This should in theory give us more stable and better performant drivers.

Qlogic is one of the first to take advantage of this with the qlnativefc driver,
wich also is a consolidated driver supporting 24/25/2600 series and 80/81/82/8300 series in one single driver.

I am suprised over the HP response as HP has certified 3PAR for 3.1 3.1.2 and 3.1.3 with the exeption of metro storage cluster, that is only supported under 3.1.3.

We did extensive pre-production testing, but decided to go for 3.1.3 because of certification. Holding back other environments, but 5.5 is coming, and most of our installations are using peer persistance...

What driver version are you using? It might be advicable to use the latest oob drivers,
as the inbox drivers are lagging.
The new driver have also started with the 7.x releases of the HBA FW,
and I suspect we might stumble on to some potential early issues there.
We never even tried the inbox driver on 5.5, got the latest from Qlogic.

You can get the 1.1.20.0-1 driver here:

https://my.vmware.com/group/vmware/deta ... FC-11200-1

We've also been vigilant about upgrading Brocade FOS prior to this,
but that can backfire as well. I suspect we might have run into a compability issue with older FW from the 5.1 inbox drivers...

We are gathering the driver info, but we have not updated anything since 5.1U1 and ther are using the oob drivers with 5.5U1, I am going to check but 5.5U1 is using whatever QLogic drivers came with 5.5U1 vmware distribution. I was aware of the new native driver model and like the concept. I also like QLogic, but we have seen our fair share of QLogic HBA firmware/driver issues, while our Emulex HBAs have been solid and far easier to manage.

For what it's worth, I'm running 2 7200's with 4 VMware 5.5 nodes on Emulex and everything has been rock solid.

I think Emulex is still using the old wrapper driver. 3par has admitted they made improvements in 3.1.3 to how it handles device log in requests, which is why we do nto see the issue in 3.1.3

I'm sure you checked with SPOCK and saw that ESXi 5.5 U1 is definitely listed as "supported" with Inform 3.12. And it also includes a variety of Qlogic cards, firmware, and driver versions.

I'm curious as to what HP/3Par support said in response to your bringing the "supported" status to their attention ?? Did they really acknowledge that they actually tested your particular mix of gear ??

thanks,
don m.

Does not matter what is on the list, 3par is saying it is best to be on 3.1.3. Considering how bad vmware QA is lately I don't trust anything they say either.

We deployed our first 5.5u1 on a Cisco UCS platform. We are on 3.1.2 with two attached T400, and have seen no issues after extensive pre production testing.

Maybe we are lucky, or the fact that Cisco doesn't use Qlogic have insulated us from seeing this error.

In any case, with the Heartbleed issue, those arrays are scheduled for upgrade to 3.1.3 next month.