replacing faulty HSG80 - correct procedure

[Lawrence Bowen]

Hi all,

I need to replace an HSG80 controller that has hung with a "replace controller" error code on the lights. I can follow the replacement procedure ok, but I was wondering if it is also recommended to bring all the servers down before actually following the replacement steps. If this is NOT necessary, what steps should be taken (if any) to ensure that data is not lost?

thanks - lawrence bowen

[Mike Naime]

Lawrence.

This depends on the OS type of your servers. If you have Tru64 or VMS Alphaservers, go right ahead! They will survive the connection loss without killing your systems.

Experience says that non- Alphaservers like MS Windows based, and IBM/AIX systems will not survive the replacement without problems that might require a re-boot to cure.

The safest way to replace the controller is to shutdown your servers first.

Note: when you return to Multi-bus failover mode, both controllers will reboot at that time. This is when you will most likely see any connectivity problems in the replacement procedure.

[Spike Burkhardt]

Lawrence,

I only have experience with HSG80's with Solaris so take that into account. If you have dual controllers in multibus failover then you shouldn't have any problems. If the controller failed while a write IO occurred, the server should re-issue the write IO request which will be served by the other controller. This is assuming that the HSG80 returned the failure code and that it got passed up thru the OS correctly. Additionally, if the volume is in a redundant configuration(Raid) there shouldn't be any problem with data loss. Shutting down the servers probably isn't a good idea. I've had more problems when not following the documented procedures. StorageWorks was designed with hot replacement as a primary goal. No, I am not a HP/Compaq employee either :-)

As a last line of defense, check with your Technical account manager.

spike