arla 0.21 Oops on Linux 2.2.1

[Chuck Lever]

On 3 Feb 1999, Magnus Ahltorp wrote:
> > so other people on this list are running arla 0.21 on Linux 2.2.1
> > successfully?  it won't run at all for me; it just oops's when i try to
> > "cd" into our afs cell.
>
> I just tested arla 0.21 on Linux 2.2.1. I can't reproduce the oops.

ok, this may be a cascade of smaller failures.  i'll break this down into
smaller bits to see if we can get an idea what's going on...

first, with arladebug set to "all", and running "arlad --no-fork" so i can
catch error messages, i try mounting /afs, and this is what arlad says:

Thu Feb  4 17:24:44 1999: arlad: worker 0: processing
Thu Feb  4 17:24:44 1999: arlad: Rec message: opcode = 2 (getroot), size =
20
Thu Feb  4 17:24:44 1999: arlad: VL_GetEntryByNameN(root.afs): Unknown
error 4294966841
Thu Feb  4 17:24:44 1999: arlad: Failed to contact any db servers in cell
0(citi.umich.edu)
Thu Feb  4 17:24:44 1999: arlad: Cannot find the root volume
Thu Feb  4 17:24:44 1999: arlad: multi-sending wakeup: seq = 8, error =
110
Thu Feb  4 17:24:44 1999: arlad: worker 0: done
Thu Feb  4 17:24:44 1999: arlad: worker 0 waiting

the mount command is successful -- no error is reported, and "df" reports
that /afs is one of the mounted file systems.

i believe that citi.umich.edu doesn't maintain a replica of root.afs.
with earlier releases, i would also get a warning, but things worked after
that.  is there a way to decode 4294966841, or is this a sign of error
code value mismatches between our AFS servers and arla?

the "Failed to contact" message started appearing after upgrading to 0.21.

according to tcpdump, a udp packet goes to the afs3-vlserver port on
babble.citi, and babble responds with a packet.  so arla is pinging the
correct vlserver host, and the host is responding.

        - Chuck Lever
--
corporate:      <chuckl at netscape.com>
personal:       <chucklever at netscape.net> or <cel at monkey.org>