Server abends when replacing dead SCSI disk

[BrianH1]

We have a Proliant 2500 and Insight Manager reported that one of the disks had failed. We called out the hardware engineers who replaced the disk. The server is running NetWare 4.11 with latest service pack.

There are two logical drives on the server as it has had the array expanded in the past.

When the new disk is inserted the first logical drive's status changes to "Rebuilding" and the second logical drive's status changes to "Ready for rebuild". In a few seconds, these are reversed.

When the second logical drive has finished rebuilding, the server restarts. At boot, the server asks to (F1) retry array recovery or (F2) continue without recovery.

When the server comes back up, Insight shows several abends:
FreeableProcedure found an invalid deleted file
Kernel detected a process switch during interrupt time
CPQDSA: a previously timed out request completed causing potential data or memory corruption

Then ASR detected by system ROM.

ABEND.log doesn't contain an entry for these, and console.log is garbage at the time of the error.

Similar happened to this server a year ago and this was due to several (dead) replacement drives. This time 5 drives have been used and we have been assured that the last one is good. While I can't rule out another dead disk, is there any other explanation for this?

 

[marvhuffaker]

I would recommend contacting Compaq.

I also generally shut off the ASR so I can generate an error rather than a reboot with not much to go on.

You should also go into the Compaq Array Utility and see what the status in there shows. This is independent of the OS and may give you some clues.

Marvin





Marvin Huffaker, MCNE

http://www.redjuju.com

 

[terry712]

is the disks on the same controller as your backup device?

 

[BrianH1]

No, we use Tivoli Storage Manager for backups over IP, there are no other SCSI devices connected to the controller.

 

[terry712]

i always cpqdasa for for monitoring scsi attached storage but more for dlt, dats etc rather than the disk storage

i would be inclined to tempoararily unload the cpq agents then try

 

[BrianH1]

Replaced the server with a spare to allow the hardware engineers some more time to look at the box. Turns out that there's a fault with the plug in board on the array controller. Once this was replaced the array rebilt.