SuSE - on ESX 3 - on IBM SVC

[harryc]

We have had ESX servers lock up twice now with a SCSI error, the ESX box locks up and drops off of VC, but the individual VMs stay up.

Consoles all show "SCSI error on target 08:02." scrolling contiuosly, consoles unresponsive.

- 'dmesg' output contains the following:

SCSI disk error : host 1 channel 0 id 0 lun 3 return code = 18

I/O error: dev 08:51, sector 838849664

SCSI disk error : host 1 channel 0 id 0 lun 0 return code = 18

I/O error: dev 08:21, sector 838849664

The response from VM-Ware is ( bold is mine )

...snip...


I would strongly recommend that you upgrade you ESX servers to version 3.0.2,
which now has support for IBM SVC virtualized storage, please note there are
some caveats:

* - VMware supports Storage Virtualization Devices with ESX Server 3.0.2 only.*
* - Only Windows Guest Operating Systems are supported with Storage*
* Virtualization Devices.*

- Backend storage arrays must be listed on both the ESX Server 3.x
Storage/SAN Compatibility Guide:
(http://www.vmware.com/pdf/vi3_san_guide.pdf) and the Vendor supported list.
- Do not share the same LUN of the backend storage array between SVD and any
other host.
- IBM SVC is supported with Qlogic HBAs only.
...snip...

Is anyone else running SuSE 9 or 10 under ESX 3.01. with IBM SVC ?

Any other Ideas ? ( besides VM's only run MS clients suggestion)

[spex]

If there is no solution - and I'm afraid there is none at the moment - perhaps you can still use your native storage system luns. (after some heavy rearrangments)

Reagrds Spex

[kbk00]

Harry, have you gotten any more action from IBM or VMware on this issue? I'm in the same boat as today I see an error scroll on a console of mine. This is mine: I/O error: dev 08:02, sector 1929744

which is part of the dmesg output that you listed. I have no way of interacting with the COS but the VMs are up. I'm not going to shut the system down and attempt a reboot unless I'm absolutely sure that there's no other route.

I've got everything on the list that is required:

-3.0.2

-Windows 2003 guests

-Supported SVC and storage

-Not sharing

-QLogic HBAs

The hardware is an IBM x3655 with 2 dual-core procs with 16 G of ram

I've never seen this with 3.0.1 or earlier so I'm a bit nervous. I just did the upgrade to 3.0.2 on this farm. Thoughts? Have you (or anyone else) seen this issue consistently?

Any help would be appreciated.

Ken