I received a bunch of helpful replies on this problem. Thanks to Barry
Kiesinger, Kevin Jones, Bryan Lavelle, Corinne Haesaerts, Wendy Fong, Kevin
Partin, Udo de Boer, Joe Carrico, Mike Shield, and Joe Fletcher.
The first suggestion from most folks was to check the SCSI IDs of the two
controllers to make sure there is no conflict (there wasn't).
The other overwhelming suggestion was to double-check termination. We
double-checked the controllers (all terminators on the cards removed) and
the y-cables, etc. and all looked fine.
Several people suggested to check the physical cables for bent pins, etc.
(no problems there--everytime we removed TCR1.6, the systems came up fine.)
There is a patch for the KZPBA cards with SMP systems and multiple
controller cards. We downloaded the patch:
http://ftp1.support.compaq.com/public/dunix/v4.0f/KZPBA_v40f_bl13.README
and applied it, but this did not rectify the problem. I gave up for the day
at that point, but the client's lead sysadmin stuck with it on the phone
with someone from Compaq (didn't catch the name). When I checked email this
morning, I had a message from Mike Shield suggesting we double-check the
termination and jumper settings in the differential BA shelf. Sure enough,
we had to make the following change to the personality module (BA35X-DA):
Termination for the differential UltraSCSI bus is also on the personality
module, and is controlled by the SCSI bus termination switches, switch pack
S4. For normal cluster operations, S4-1 and S4-2 should normally be on.
Section 3.6.2.2 discusses these switches in more detail.
The TCR hardware config docs can be found at:
http://www.unix.digital.com/faqs/publications/cluster_doc/cluster_16/TCR_HWC
FG/TITLE.HTM
When I got back to the client site this morning, come to find out that they
figured this out at about 9pm, so we were able to get back to the training
class right away today.
It apears that the rack was installed as if the systems were to be used as
standalones rather than in a cluster. At first I thought this was a bit
screwy, but the client informed me that the MC2 cards did not come in with
the original delivery of the equipment. That made more sense.
Thanks again to everyone who replied. This was my first exposure to the
differential shelf and the BA35X-DA personality module. Interesting
experience!
Cheers,
Ed.
---
Edward J. Branley, elendil_at_yatcom.com, lawfive_at_hotmail.com
Seashell Software, New Orleans, LA
+1.504.251.3357
Homepage/PGP Key at http://www.yatcom.com/ejb
Received on Thu Feb 10 2000 - 19:26:38 NZDT