[OpenAFS] What happens if a client doesn't see a "break callback" caused by a volume move?

Kim Kimball kim.kimball@jpl.nasa.gov
Sun, 18 Feb 2007 11:38:17 -0700

Hello folks,


a. A volume is moved
b. The moved-from fileserver breaks all callbacks for that volume.
c. A given client does not see the break callback message.

What happens? 

IIRC the file server puts the client on a "client down" list and ignores 
the client until the client contacts the file server again.  The client 
is oblivious to changes in the volume, but may eventually contact the 
moved-from file server again and 1) be told to break all callbacks for 
volumes on the file server and 2) that the moved volume is no longer 
available on the moved-from fileserver.

I also believe that there are timeouts on cached VLDB information, 60 
minutes IIRC, and that the cached volume information should expire and 
the client should go back to the vlserver to refresh location info for a 
given volume, thereby discovering within 60 minutes that the volume was 

Could someone please either shoot my theories full of hooools (as one of 
my ESL students once said) and/or fill in the blanks?

Thanks very much.

Kim Kimball