[OpenAFS] migrating /usr/afs/db/kaserver.*

[Russ Allbery]

Robin Yamaguchi <rhy@physics.ucsb.edu> writes:

> I'm going through the process of a bare-metal restore on our AFS server
> for testing purposes.  On a 2nd machine, running the same OS (redhat 9)
> and same version on OpenAFS (1.2.9).  I got the AFS server running in a
> new cell w/ no problem.

> To restore my ka and pt entries, I copied over /usr/afs/db/kaserver.*
> and /usr/afs/db/ptserver.* files from the 1st machine to the 2nd,
> properly shutting down those services before touching those files on
> both machines.

I know this doesn't exactly help with the exercise that you're going
through, but far and away the best way of replicating the database files
to another server is to just install all the server software on a second
machine with an empty database file, copy over the KeyFile (securely!),
add it to CellServDB on the existing database machine, and kick the
database server processes on both machines.

They'll talk to each other, figure out that the second machine doesn't
have a database, replicate the database via Ubik, and you'll be all set.

The best way, IMO, to do AFS database backups is to have enough machines
around that you never lose them all at once....  (Of course, if you get
corruption that then replicates, this strategy doesn't help.)

-- 
Russ Allbery (rra@stanford.edu)             <http://www.eyrie.org/~eagle/>