[4810] in testers
Re: Updating to Athena 9 on Ultra 5 fails
daemon@ATHENA.MIT.EDU (Mitchell E Berger)
Fri Jun 15 15:36:12 2001
Message-Id: <200106151936.PAA04966@byte-me.mit.edu>
To: "Thomas H. Grayson" <thg@MIT.EDU>
cc: testers@MIT.EDU, dsheehan@MIT.EDU, miki@MIT.EDU
In-Reply-To: Your message of "Fri, 15 Jun 2001 15:15:38 EDT."
<200106151915.PAA29410@coast.mit.edu>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Date: Fri, 15 Jun 2001 15:36:08 -0400
From: Mitchell E Berger <mitchb@MIT.EDU>
This seems identical to the failure mode joust is experiencing in the test
cluster. The problems just described are not the only ones. Two nights ago,
I managed to force joust past this point of the installation, and the
experience I had suggests that by doing something very similar, you could save
the machine and have it continue running 8.4, but since you don't really have
any state on it, a reinstall is probably easiest.
To describe the rest of the failure, here's what I did: net boot off of
tae-kwon-leap, specify a custom install with custom partitions and type "!sh"
at the format program's prompt to get a shell, fsck -F ufs -y
/dev/rdsk/c0t0d0s3 (every single inode was reported as damaged), mount it (it
appeared empty after having cleared all the inodes), create the vice directory
on it with an etc and cache directory, set up a tftp server on another machine
to serve CellServDB, CellServDB.public, SuidCells, SuidCells.public, ThisCell,
cacheinfo, and cacheinfo.old (I'm not sure if the last two are needed),
unmount, fsck /dev/rdsk/c0t0d0s0 and /dev/rdsk/c0t0d0s1 as well for good
measure, and reboot.
Then the disk slice mounted properly and after an immensely long cache scan,
afsd came up, and it attempted to continue running the install from the
miniroot. However, it complained at a certain point about nawk failing. For
the most part, this only seemed to hurt a couple of packages and the rest of
the install seemed to go ok, but Beland and I were sort of working while it
was installing, so we could have missed something. In the end, it ended up in
an infinite boot loop where it complained that it couldn't find a device
number foor rootnex. Sure enough, it wasn't in name_to_major. I hand-typed
several different name_to_major files into it, including one from Larry's
Ultra5 running Athena 9.0, and none of them made it work, so I finally gave up.
For the record, I do remember during the 8.4 install that joust has a Segate
Medalist hard-drive, which means the AFS issue is the one Greg just mentioned,
but this doesn't necessarily explain why name_to_major is coming out wrong and
nawk is failing.
Mitch
