HBA Failover Questions???

[VMdawg]

Ok so 85% of our VM environment is booting from SAN...So basically your looking at 2 Qlogic HBA's per ESX host running to a Brocade director 1 HBA port per switch managed through an SVC. Now I realize that SVC is not supported/compatible yet (see my other post for info on that:

http://www.vmware.com/community/thread.jspa?threadID=81799&start=0&tstart=0). My question has to do with HBA failover or multi-pathing...What we have found through testing is that the only way one of our VM servers doesn't completely lock up when we unplugg one of our HBA's (i.e. testing HBA failover, switch port failover) is to make sure both HBA's in a host are set to "enable bios" and enable "boot from LUN"...once we did that and unplugged one of the HBA's, there was a small pause but then the VM became responsive again without booting etc...

This is an important test when you boot from SAN over fiber, if I loose an HBA in a host over the weekend, I don't want to have to come in and manually change the zoning on the SVC or replace the HBA physically...anybody else test this, have best practices?

[williambishop]

It's really dependant on your setup, or anyone else's for that matter. If it's not failing over, then you have something wrong. It SHOULD failover, if properly implemented, with the briefest of pauses at most.

[mwheeler1982]

I thought that multipathing/failover was not currently supported with the boot volume?

[williambishop]

I know that was a problem in 2.5, I don't think it applies in 3, or if so, there are sure a lot of people without the problem. The biggest problem with boot from san is that you have to give it a lun that only it can see. I could be misremembering our early research(we ended up using local storage, just because it's simpler and it's only used to boot with).

[christianZ]

We have tested here multipathing with SLES10 and have many problems with multipathing of "/" - coudn't get it running.

The Novell support confirms this is not functioning for now (RHEL4 can't do it too) - so I guess it won't be possible to do it with ESX.

But maybe I'm wrong ?

[VMdawg]

I know that was a problem in 2.5, I don't think it

applies in 3, or if so, there are sure a lot of

people without the problem. The biggest problem with

boot from san is that you have to give it a lun that

only it can see. I could be misremembering our early

research(we ended up using local storage, just

because it's simpler and it's only used to boot with).

Not a problem??? I must be looking on a different forum...VMware says not many people are even using SAN/SVC environments. I doubt they would be working on a new ESX build to support SVC if it wasn't a HUGE problem.

[VMdawg]

We have tested here multipathing with SLES10 and have

many problems with multipathing of "/" - coudn't get

it running.

The Novell support confirms this is not functioning

for now (RHEL4 can't do it too) - so I guess it won't

be possible to do it with ESX.

But maybe I'm wrong ?

According to VM/IBM VMware will be releasing an ESX build this year that addresses SVC compatibility issues: http://www.vmware.com/community/thread.jspa?threadID=81799&start=0&tstart=0

*on that thread a VM Engineer says we are supposed to hear something in a couple weeks...

[VMdawg]

I thought that multipathing/failover was not

currently supported with the boot volume?

Like I said we were able to get it to work, but you're right...it's not supported yet.