vsphere hypervisor on IBM HS22 SAS connected IBM DS3512

[ArkovCZ]

Hi,

I have a problem with upgrade from 4.1 and also with clean installation vsphere 5 (vsphere hypervisors) on two IBM bladecenter HS22.

They were running about half year on version 4.1 without problems, and now customer want to upgrade to version 5.

The problem is, after some time or on some load storage DS3512 and datastores on them are disconnected from hypervisor. VMs looks like they are runing and response to ping, but nobody can access them, hence HA don't move them on second HS22.

After hard reset of HS22 everything goes well to another "hang". HS22 is responsing normally. In log that hanged HS22 are logged problems with LSI mpt driver and losted connection to DS3512. There is LSI1064e adapter.

I don't know what can I do with that. Both bladecenter servers behave same, after some time they disconnect.

IBM HS22 is on HCL with vspere version 5, HS22's and chassis are on newest firmware.

[cityinfosys]

On ibm's download site they only support esxi 4.1 update 1 currently. You might have to go back to esxi 4.1 update 1 to fix the problem. You could also buy the esxi usb key for the HS22 it fits under the fiber card. And boot to that.

[micbuc5583]

I am seeing a similar issue with our HS22's after an upgrade from 4.1 to 5. Out of the 8 we have 2 of them will disconnect from VCenter about once a week. After the host disconnects from VCenter the VM's will continue to run as normal unless I attempt to log into the ESXI 5 console. Loging into the ESXI console will hard lock the host and drop the VM's. Host logs display the following error.

linscsi: SCSILinuxAbortCommands: 1798: Failed, Driver MPT SAS Host, for vmhba0

So far we have tried the following.

1. Full hardware update of all HS22 firmware

2. Found and updated LSI drivers on the ESXI host

3. Blew away the two failing host and re-installed a fresh copy ESXI 5

I have verified that the HS22 and all hardware is listed on the  compatibility list. IBM also list ESXI 5 as supported on the HS22.

So far since wiping and reloading ESXI 5 we have been up and running about week without failure. I just wanted to drop a line here and see if anyone else is seeing this on HS22's and has found a solution I have not listed.

[pcamelio]

Hi,

We move one week ago 4 HS22 blade, and one meet the problem you report, under stress (backup VM) or login to ESXi console, the blade freeze and drop the VM :smileyangry:  (HA is not working). Reset of the blade "solves" the problem

Not sure what is going on ?:smileyconfused:

I reinstalled the blade two days ago (fresh install - format everything, reconfigure etc ...) and the problem still occurs. Therefore, the FC connection are still working !!!! - the VM on FC are still working, while VM on the SAS broken.

It seems to be a problem with the SAS (internal card and CIOv card) (local disk and DS3200 LUN are no more available !!!!!!!!!!!!!)

We plan to make a full upgrade of the firmware this week et see what happen

but any advice would be helpful !

Regards

[AWoroch]

Stumbled across this thread, as I'm looking for others in the same boat as ourselves.

Our situation is similar:

IBM HS22 blades

IBM/LSI SAS 1064E controller

ESXi 5

mpt2sas  hangs when checked via the ALT+F12 screen at the console (vmkernel.log).

Systems just "drop out" of vCenter, can't be managed, can't be handled.  They DO seem to still respond to VM's, but in our case all the VM's are on FC shared storage, and NOTHING but the boot OS is on the mirrored internal drives.

Check this thread: http://communities.vmware.com/thread/393736 and see if it is similar.

If so, try this IBM link: http://www-947.ibm.com/support/entry/portal/docdisplay?lndocid=MIGR-5089360&brandind=5000020&myns=x1...

It will suggest you perform a workaround, listed at VMware's site: http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=103026...

If that should happen to work for you, please let me know.  I'm trying to find other people that not only have had the issue, but who have found a long term workaround.

Avram