[OpenAFS] Debian setup problems (afs rootvol and afs-newcell)

[Marcus Watts]

Matt Weatherford <mbw@u.washington.edu> writes:
> Message-ID: <3F96EF21.4040806@u.washington.edu>
> From: Matt Weatherford <mbw@u.washington.edu>
> User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.5) Gecko/20030925
> Cc: openafs-info@openafs.org
> References: <3F945C2F.9080102@u.washington.edu> <tsl1xt7h5sv.fsf@mit.edu>
> 
> 
> Folks,
> 
> Im having lots of trouble getting the setup under debian to work
> ... Im getting bound up in afs-newcell (protection database exists 
> errors) or in afs-rootvol (fs:'/afs': Connection timed out, Failed: 256)
...

You might want to use "script" to save output from running several
of these and report this as a bug to the debian folks.  They
can't fix things if they don't know that they're broken.
My guess is that Sam Hartman will be very interested in your problem.

As a completely separate path, you could also investigate setting
things up without the handy scripts.  It's a bit more work, but
it does expose you to the various pieces much more closely.
Here's my notes for
	/afs/umich.edu/user/m/d/mdw/wp/uniq.2y
this includes debian specific details so might be easy to follow.  On
the other hand, the bits aren't necessarily well explained and might be
out of date or otherwise quirky, so it may be hard to follow.  One of
the pieces here that may be hardest to follow is running "pt_util" --
that's a handy trick borrowed from the afs-newcell script.  pt_util has
a really tricky syntax and confusing messages.

Yet another thing to consider -- the afs-rootvol script probably
requires that it be run on an afs client that has your cell available,
and may not be smart enough to ensure that this is so on its own.
You might be trying to run this on a db server where you (quite
correctly) don't expect to run a cache manager.  An easy fix therefore
would be to bring up the cache manager before running that piece.
Of course, if you don't have a root volume this may be hard.
Looks like there's a "prerequisites" notice that says all this.
One way to work around
the need to have a running cache manager is to do a vos restore.  Of
course that presents a chicken and the egg problem in that you
would need a suitable "vos dump" file from which to restore.

A slightly more clever "afs-rootvol" script could either contain
a suitable generic "root vol" vos dump, or perhaps a perl script or
C program that generates a suitably customized vos dump.
This would completely eliminate the need to run a cache manager
to bring up your cell.

				-Marcus