Nothing solid here but a couple of interesting thoughts:
Dr. Tom Blinn from Campaq suggested it was a bug and / or we had an
unsupported configuration. He suggested, as a workaround, removing the
SWXCR, upgrading the system, and reinstalling the SWXCR. Alas, the machine
boots from the SWXCR. It's possible the configuration is slightly screwy.
The machine was originally a 2100 4/200 with a single CPU and has since
been upgraded with a new backplane and two 5/250 CPUs and memory. Could be
some sort of one off hardware problem. Much as hardware vendors would like
us to believe, I'd find it hard to believe every combination of hardware
could get tested. Dr. Blinn also pointed out that the installation CD-ROM
uses a special installation kernel and not genvmunix so it might be
specific to the install kernel. Though, since booting the 4.0A CD-ROM did
cause the panic we're going to use that as a diagnostic on our other
2100A's before the upgrade.
wmills_at_wellsfargo.com suggested using the 5.2 firmware instead. The
rationale being that 5.3 may have enhancements not needed (or dangerous).
I kind of like this one. Our mania for the latest and greatest might be
getting us into trouble here. I might start using the minimum required
firmware for upgrades rather than the latest.
What we did do because of time pressures, was a new install of 4.0D and
back-filled things from backups. Copying the old root partition to an
unused partition and having it mounted during the back-filling was very
useful in recovering config files and such. Luckily almost all our data
was out on an SW800 so we didn't have to touch it at all. If at all
possible I recommend segregating your data and system disks. It was
tedious but it seems to have worked OK. Upgrades are much easier than
installs given a choice.
Original question:
>We were trying to upgrade a 2100A from 3.2G to 4.0D. At the halfway point
it needs to be upgraded to 4.0A. We had upgraded the firmware using the
5.3 disk (eventually we were going to 4.0E). When the upgrade finished to
4.0A the machine would panic when trying to initialize the SWXCR-EB board
streaming an dump on the console. The SWXCR at that point was running
firmware rev. 2.15. Assuming we needed to upgrade the firmware to 2.16 we
eventually did that but still had the panic. It would panic booting either
the 4.0A kernel on disk or the CD-ROM ending with a Severe Machine Check
(660) error.
>
>Time being short we restored 3.2G and the machine booted fine. It would
also boot the 4.0D CD-ROM fine. When we called Compaq the software guys
hadn't heard of the problem and said that it was probably a hardware
problem. The hardware guys said it was strange that it would boot with
4.0D and 3.2G but not 4.0A and that if the board were broken it should boot
anything.
>
>Other picky details: It's got 3 KZPSA boards in it and it's a 5/250 with
1.5GB of memory. The SWXCR is the only SWXCR and it's a 3-channel board
with 4 disks ea on two of the channels.
>
>Any clues out there? We ended up doing a fresh install on this machine
but we have three other 2100A's running on a production system we'd much
rather upgrade than do a fresh install and have to reconstruct.
>
----------------
Grant Young, MIT Information Systems
Administrative Servers Service Team
W91-219B -- 253-7529
Received on Tue Apr 13 1999 - 14:06:13 NZST