[OpenAFS] Manually killing fs processes

[Dr A V Le Blanc]

On Fri, 25 Jul 2003 at 18:28:39 -0400, Jeffrey Hutzelman <jhutz@cmu.edu> wrote:
> Of course, if you do find yourself in this situation, you can use 'vos 
> unlock' to unlock the VLDB entry, and abort the dump-in-progress by killing 
> the volserver (send it SIGTERM; the bosserver will restart it 
> automatically.  Do _not_ send any signals to the fileserver).

Incidentally, we have long had a problem in that our bosserver processes
die.  (This is on Silicon Graphics servers; we've never had the problem
under Linux.)  This means that the bosserver process is gone, but
all the child processes are there, in particular fileserver and volserver.
If you try to restart the bosserver without killing everything off, there
are problems, because of course you either end up with two of some
process, or the new one fails because the old one is using the port.
But I have never found a way to kill off the orphaned fileserver
processes cleanly.  The result is, the new fileserver processes do
a salvage on all /vice partitions, which normally takes nearly an hour.

Is there a signal we can send to fileserver processes to tell them
to shut down cleanly, so that no long salvage will be required when
the new bosserver restarts everything?

     -- Owen
     LeBlanc@mcc.ac.uk