Hello there,
After 3 years now I needed to rethink our system setup and I figured
out what ridiculous condition caused the shutdown problem.
The server machine had 2 VPN network interfaces and it turned out,
that...
$ vos listaddrs -printuuid
UUID: xxxxxxxx-xxxx-xxxx-xx-xx-xxxxxxxxxxxx
domain.tld
192.168.x.x
10.8.x.x
...AFS was listening on those 2 interfaces too, but the clients weren't
connected to the VPNs, and despite that they assumed a multi-homed
volume. It caused not only a very slow shutdown because of the timeout
but also another delay of about 15 seconds to log in while waiting for
the volumes on the VPNs.
Back then I accepted that this won't be solved in the near future -
well, thanks to journalctl the problem was successfully identified
recently in a virtual environment and it turned out that the same
problem was affecting our workstations...
Actually I don't know whether it counts as a bug or not - maybe it
does, because it means that if there is a multi-homed volume and it is
not available via a network interface, then it will induce the shutdown
problem despite the acknowledgement of not reaching the volume that can
be read in the logs.
The quick solution for me:
$ vos setaddr <UUID> <HOST> <the only address in use>
Thank you for all your help, including your other guidelines for
various problems you responded to in emails, on IRC, etc. :)
Best Regards,
István
