ACS V8.7-F7 Upgrade caused Tru64 Cluster Crash

[SAKET_5]

Hi All,

I upgraded one of our HSG80 based SA from ACS V.8.6F10 - V 8.7F7. The procedure used to upgrade the controllers was rolling - so one controller at a time. HP Procedure from the HSG80 Service Manual was followed and the upgrade was completed successfully with no drama.

Later on we realised that during the upgrade, the only Tru64 UNIX Cluster (2 members) experienced I/O errors and various Advfs domain panics on multiple Advfs domains. This finally led to a system crash on both members.

I thought this was quite strange and defeats the idea of performing rolling upgrades of ACS.

Anyone, any clues as to what went wrong?

Cheers!

[Erwin van Londen]

Saket,

The most made mistake is when people upgrade the controllers too fast. When you upgraded one controller wait at leat 5 minutes before doing the other. When you insert the new firmware card it takes about 3.22 minutes for the new software to do a data structure, memory and integrity check. The controllers CPU is running 100%. If you are too fast with shutting down the other one you end up with a controller that's not able to serve IO requests because of the processes that are allready running.

In the past two weeks I've upgrade 4 EMA12000 to version 8.8-2F and 8.8-2P without any pain. These arays are serving 3 4-node Tru64 clusters and one 8 node OpenVMS cluster.

Patience is the magic word here.

Kind regards,
Erwin van Londen