[OpenAFS] Handling stale copy of RW volume on server back up

John Tang Boyland boyland@uwm.edu
Tue, 07 Nov 2017 11:03:41 -0600

Dear OpenAFS,
   This is probably a FAQ, but I couldn't find it.
This references the somewhat common situation when a fileserver
goes down.

1. fileserver fails
   -- we try to resuscitate it, but fail, and people need access to data,
2. We use the daily backup to create restore volume on a new server
3. delentry to remove the entry to the "down" fileserver
   (we don't want to either wait for the server to come back up, because
   it's not up, and also don't want the data to go away in case
   we are belatedly able to get the fileserver back up.)
4. vos rename the restore volume to the original volume name
5. continue to work on the down fileserver.
6. A week later, the down fileserver comes back up!

Now the previously down file server has a stale copy
of the RW volume, but it may include changes from between the backup
time and the time the fileserver crashed.  So I don't want to lose it,
but nweither do I want it to be used or backed up in place of the
current copy.

I have NOT syncvldb yet -- I'm afraid of what might happen.

How do I rename the volume on the fileserver to get out of the way of
the replacement volume?  I could ZAP it, but I'd like to preserve the
data.  vos rename works on the VLDB, not on the server.

The only thing I could find is that one could vos dump each of the stale
volumes and then restore as a new volume and then zap the stale volume.

Is that the best?
Please reply to me as well as openafs-info (I get a delayed digest).

Best regards,