[OpenAFS] Unclean shutdown: AFS client ate my filesystem (ext3)

[Helge Bahmann]

Hello list!

Is there some wicked interaction with the afs cache manager and ext3, or
is this "normal" for unclean shutdown?

Today I had to reboot the afs fileserver (Debian Woody, afs-client and
-server 1.2.3), though for reasons external to afs.

During shutdown "something" kept /afs busy, I don't know what; as a result
the afs client could not be stopped, kernel module could not be unloaded
and afsd could not be terminated, also keeping /var busy (where the afs
cache is located).

At system startup /var was garbled so badly that journal recovery could
not bring the filesystem back into a consistent state, only manually
fscking the thing brought it back to life.

What I have learned is to put the cache into its own partition but I am
dismayed that openafs-client appears to be unable to handle this situation
sanely -- is this true or am I having just bad luck?

Regards
-- 
Helge Bahmann <bahmann@math.tu-freiberg.de>             /| \__
The past: Smart users in front of dumb terminals       /_|____\
                                                     _/\ |   __)
$ ./configure                                        \\ \|__/__|
checking whether build environment is sane... yes     \\/___/ |
checking for AIX... no (we already did this)            |