VMware Cloud Community
vmuniverse
Contributor
Contributor
Jump to solution

Unplugging HBA cable on the ESXI Host causing VI Client to hang

We're using HP DL380 G5 with the embedded version from HP (ESX 3.5i U3 + latest hotfix)

This server has all the available features available (HA, DRS etc.) The server has 2 EMC QLogic 2340 HBA's with the latest firmware from EMC in it.

If I look in the VCenter at the Storage properties of a disk it shows me 4 available paths to the SP on our SAN. (all MRU settings. we have Active/Passive SAN)

Now here is the problem we have:

As soon as I unplug the fibrecable, the Vcenter should use another active path. It looks like it does the VM Machines are continue to work, but only after a rescan of the hba's it also shows that 2 paths are broken

But when I close the VI Client and start it up then the problems begins:

  • VM Machines loses connections (pinging irregular and after a while unavailable)

  • VI Client freezes and is not responding

Ultimately, after I reconnect the fibrecable (because everytime I start the VI Client it hangs) all problems are gone and everything will work again.

If I then take a look at the VM Machines some are in BSOD, some have write delayed Failed

Unplugging a fibrecable should not give any disruption, thats why we have redundant HBA's. But it does!

Anyone has suggestions?

Reply
0 Kudos
1 Solution

Accepted Solutions
celak
Enthusiast
Enthusiast
Jump to solution

The VMkernel drivers for QLogic HBAs include firmware which (effectively) hot-loaded onto the HBA as the driver loads.

See http://communities.vmware.com/message/325181#325181

So, don't bother with the firmware level of the HBA.

See latest "Fibre Channel SAN Configuration Guide"

page 99 : Setting the HBA Timeout for Failover (qlport_down_retry value)

Good luck.

View solution in original post

Reply
0 Kudos
5 Replies
vmuniverse
Contributor
Contributor
Jump to solution

Today I just received a brand new QLA-2340 and installed it into my ESX Host.

This ESX host was already equipped with a QLA-2340.

Both upgraded to firmware 1.54. HBA1 had fimrware version 1.26, and HBA2 had firmware version 1.52

Now I notice that when I unplug the fibre cable from the hba card that already was installed on the ESX it will switch to the secondary hba without any interruptions

How come that with this new card the failover works?

Reply
0 Kudos
RParker
Immortal
Immortal
Jump to solution

First of all what are you doing? Why do you need HBA connections from VC, it's obviously a physical machine. For best practice it's not a good idea to have anything running on the VC except VC, not even the database.

Secondly why are you unplugging the cable? How often are you going to need that? If you are running something on there, apparently VC is attached to something on the SAN, so that is probably why it hangs, there is a share or something that is causing a disrupttion, and thus the service hangs, but if aren't running anything else on that machine, the HBA shouldn't matter.

I hope you aren't using your VC as your VCB backup also, cause that could be mean disaster.. also with dual paths VCB doesn't work very well..

But maybe if you give some extra info as why failover for VC when it needs no storage, and what your ultimate goal is, maybe we can help you. You don't need HBA for VC, unless the storage is on the SAN.

Reply
0 Kudos
vmuniverse
Contributor
Contributor
Jump to solution

I have updates the subject, because It didn't explained well our situation

The situation is that we have 3 ESXI Host U3.

1st Host = 2 QLA-2340 which had firmware 1.26. Upgraded to firmware 1.54 (confirmed)

2nd host =2 QLA-2340 which had firmware 1.26. Upgraded to firmware 1.54 (confirmed)

3rd host = 1 QLA-2340 which had firmware 1.26. Upgraded to firmware 1.54 (confirmed)

Today we have received a new QLA-2340 hba. I placed this hba in my 3rd ESXi host as redundancy. This hba had firmware 1.52 installed. Upgraded to 1.54 as well

When I have a VM Machine running on Host 1 or 2 with an active path on hba1 it freezes the VI Client as soon as I unplug the cable from that hba. Also the VM Machine eventually stops and causing a BSOD

The second hba appears that it doesn't fail over (Host 1 and 2 have this situation)

Now with the new installed hba that purchased this day (also QLA-2340) unplugging the cable on the first HBA will failover to the newly installed hba2, without any disruption. It also shows that from the 4 available paths 2 are broken (dead)

This is the way it should work as VMWare says, but not for our 2 first ESXi host

Hope this clearify something

Reply
0 Kudos
celak
Enthusiast
Enthusiast
Jump to solution

The VMkernel drivers for QLogic HBAs include firmware which (effectively) hot-loaded onto the HBA as the driver loads.

See http://communities.vmware.com/message/325181#325181

So, don't bother with the firmware level of the HBA.

See latest "Fibre Channel SAN Configuration Guide"

page 99 : Setting the HBA Timeout for Failover (qlport_down_retry value)

Good luck.

Reply
0 Kudos
vmuniverse
Contributor
Contributor
Jump to solution

Ok I'll take a look at this.

Reply
0 Kudos