Seeing Duplicate Disks in Cluster

[Bing59]

I have a W2k3 CLUSTER WITH DUAL HBA's connected to a Clariion CX600. After doing a Flare OS upgrade to the Clariion Array, we are seeing the disks belonging to the other node in the cluster as "Not Initilaized, Unallocated". I have been told this is a Microsoft issue, an have found the Hot Fix for it. But, I am wondering why don't I see this with the cluster nodes that only have a single HBA in them?
Has anyone else had this issue?

 

[comtec17]

I believe this has something to do with Microsofts MPIO.

 

[xmsre]

Wrong answer. Care to try again?

EMC does not license MPIO, they have their own proprietary multipathing software - powerpath (well there is powerpath for iSCSI, which does use Microsoft MPIO, but that's another story and the term HBA was used here so I don't think so).

Let's step back a minute and parse this out:

DUAL HBA's connected to a Clariion CX600

- Multipath

After doing a Flare OS upgrade to the Clariion Array

- Change to the environment

we are seeing the disks belonging to the other node in the cluster as "Not Initilaized, Unallocated".

- yup, 820127 describes an issue where vds.exe incorrectly
translates the disk status of "ERROR_BUSY". Hmmm, but is this your issue?

why don't I see this with the cluster nodes that only have a single HBA in them?

- The difference between the single HBA and dual HBAs is the path management software.

Do we see that not initialized, unallocated wording in any other situations?

- yup, when there's a pathing/zoning/mapping issue

http://support.microsoft.com/default.aspx?scid=kb;en-us;311081

I suspect it's a bit more than the vds issue.

I have been told this is a Microsoft issue...

- Let me guess, by EMC tech support. Now, if you ask MS, I bet they'll say it's an EMC issue. About as productive as setting two phones to call forward to each other and dialing one of them, isn't it?

You could probably save yourself a lot of frustration by cranking up the logging on the HBAs and viewing the status code that gets returned in both the single path and dual path situations. If the single path configuration returns the same status as the dual path configuration, and it's ERROR_BUSY (the HBA vendor should have documentation on how to break out the scsistat, srbstat, and sense code data. If it gets logged is scsiportlogerror format, log at offset 0x38 for 0x08). If say, both configurations don't send the same error code and the dual configuration is sending busy, then it would be vds.exe. If not, you have issues in the path.

 

[Bing59]

xmsre,
Thanks,
For the help. I am going forward with the Firmware and driver update along with upgrading W2K3 with SP1 which includes the HotFix for this issue. Trying to find out how to get logging turned on for the emulex adapters.

Your dead on about vendors pointing fingers at each other and leaving us end-user/engineers between a rock and a hard place, quite possibly with our Jobs and Credibilty on the line.
Again, Thanks for the assist.
Bing

 

[xmsre]

One thing on the Emulex storport miniport drivers: You do want to update. There were some nasty bugs in previous versions (resource leak).

That said, EMC uses a full port or monolithic driver. (one big reason they can't use MS MPIO). I think the current version is 2.30a2.

These links may help

http://www.emulex.com/ts/docoem/framemc.htm

The troubleshooting section should help

http://www.emulex.com/ts/docfc/w2kport/230a2/pdfs/manual.pdf

 

[xmsre]

Forgot to metion, it's the local driver trace mask that you want to set. Should set a mask value of 0x00000002 to get the srbstatus and scsistatus messages. To find a complete list of the codes, you'd download the platform sdk from MS and look in scsiport.h. A fairly decent list is also available in Microsoft KB 244780.