HSG-80 Reset due to complications with Win Cluster

[CA918374]

Hello,

First of all, I appologize for the novel. There are allot of details to this one. I hope I have minimized it without leaving out too much.

We are running an EMA12000 with a combination of Tru64 UNIX, Win 2000 and most recently Win 2003. We have one production Win Cluster (2000) and were in the process of bringing a 2003 cluster online. In fact, it was online for five days before we had our problem.

Our 2003 cluster lost it's quorum drive, it was shutdown then our 2000 cluster lost it's quorum drive about 30 minutes later. Needless to say, cluster services were terminated and both of our windows cluster were dead in the water. All other servers attached continued with production activity without any problems.

In troubleshooting the cluster we lost console, win terminal and hsxterm options into our HSG-80. The controllers froze up after we attempted to delete the unit number of the lost quorum for our production 2000 cluster. We were forced to physically reset our HSG-80 controllers to gain control again.

Has anyone else experienced the loss of an HSG-80 due to a Windows cluster? I'd appreciate any additional information/experience others may have.

Regards,
Cynthia

[Uwe Zessin]

Cynthia,
what is the firmware revision and patch level on the HSG80 controllers? I have seen some problems with Tru64 Unix cluster's quorum disks on some versions.

Do you use any form of zoning in the fabrics?

Be VERY careful using multiple management paths at the same time! Is 'win terminal' another name for the SWCC (StorageWorks Command Console)? I thought that hsxterm is no longer supported, because it can cause controller hangs.

[CA918374]

The HSG80 firmware is at V87F-4.

Yes, we use zoning on our fabrics.

Yes, SWCC is what we are using. I'm a UNIX admin and get stuck on "terminal". I didn't realize hsxterm was unsupported, this would explain a few things.

[Uwe Zessin]

Ah yes. A current SWCC agent should properly lock the controller against other agents, but I am afraid 'hsxterm' will not do that as it send comands directly through a LUN.