Read/Write and POWERFAILED on VA7400 LUN

[Eugeny Brychkov]

Jim,
syslog errors provided are for c8t1d3 (should be VA LUN11, and frontend aborts showed for LUN11 also). Looks like your SAN was configured improperly. VA should be in fabric login, host behavior table (or default host beavior) should be set to hpux. VA firmware should be HP18 and battery 4.3 (call HP for update).
Host software configuration is also important: check if server has recent FC driver installed http://software.hp.com , Diags, GR and HWE patch bundles installed from same CD release.
If you would zip 'armdsp -a' output and brocade's supportshow output and attach to your next reply I may be will be able to tell more
Eugeny

[James A. Donovan]

The VA firmware is HP16
The battery firmware is 4.3
VA is attached to the fabric in point-to-point.
VA host behavior table is set to hpux.

The server in question:
# swlist -l bundle|grep GOLD
GOLDAPPS11i B.11.11.0206.4 Gold Applications Patches for HP-UX 11i, June 2002
GOLDBASE11i B.11.11.0206.4 Gold Base Patches for HP-UX 11i, June 2002
....plus some additional patches for Oracle and Oracle Apps 11i

I don't think that there is something wrong with the SAN configuration. This is only one of 6 servers attached through this switch to this VA-7400, and it is the only one that has exhibited this behavior. This server (pollux) has been attached to the SAN for several months and just began issuing the diagnostics messages yesterday.

[Eugeny Brychkov]

Jim,
if you're saying that everything worked fine before and now it complaints that having power fails then really something changed. Think about it - what? VA looks like healthy. I see in supportshow temp reported 36C and it logs errors. Are you sure it does not exceed 40C? It should not, otherwise brocade will be in 'marginal' state. Recommended (reported by brocade) is 30-32, thus room temperature 22-25C and proper airflow from front with cool air feeded to the rear with hot air withdrawn
Eugeny

[James A. Donovan]

I've found one thing wrong with the configuration of this server. The logical volumes did not have bad block relocation disabled. I'm going to go with the theory that the heavy updates that were going on that day were encountering bad blocks on disk that got relocated by the VA array and the OS did not know. Bad block relocation has now been turned off, and I haven't seen these errors reappear since that day.

[Fabrice Meynard]

Hi Jim,

if the problem reoccurs, have a look to PV time out, must be to 60 with VA arrays (default is 30).

Fabrice