Jeffrey> How are others managing this?  Big question is do you bother
Jeffrey> locking the vob's or just go with the snapshot?

You *must* lock the VOBs, othwewise your backup isn't worth anything.
But when you've got the VOBs on a netapp, it's quite fast, since what
you do is:

    foreach vob (`cleartool lsvob | grep volume | awk ... `)
            cleartool lock -tag vob:${vob}
    end

    rsh toaster snapshot volume

    foreach vob (`cleartool lsvob | grep volume | awk ... `)
            cleartool unlock -tag vob:${vob}
    end

And then you backup the snapshot.  Quick and easy.      

Jeffrey> I know of people running Oracle on filers that don't even
Jeffrey> bother with hot backup mode, and so far (years) have had no
Jeffrey> issues.

I'm not qualified on Oracle, but I'd say that they are lucky.  

Jeffrey> NBU server tells client it's time to backup.  Client runs
Jeffrey> through it's bpstart script and locks the vob's.  Backup runs
Jeffrey> and script unlock's vob's.

Jeffrey> For filer it works similar, but part of the script would be
Jeffrey> to create a snapshot after the vob's are locked, which
Jeffrey> requires rsh access to the filer; something we don't allow
Jeffrey> from application servers.  Heck, only 2 machines out of 8k
Jeffrey> have that access but I'm willing to add a master server since
Jeffrey> it's IT controlled.

I'd probably just make the MediaMaster the machine that backs up the
ClearCase box, and give IT root privs (or even just clearcase admin
privs) on the client that hosts the VOBs.  

Jeffrey> Also considering an rsh from client to master server that
Jeffrey> triggers a script on the master to create a snapshot, but
Jeffrey> that seems to add a fair amount of complexity since the
Jeffrey> master would then need to let the client know when snapshot
Jeffrey> is complete.

Very true.  I actually did do something like this for our setup, where
we were using a bunch of sockets to coordinate the process between the
backup server, the server that ran clearcase, and the master server
which has RSH access to the filer.  Worked quite well.  

This then had the backup system backup the snapshot directory, which
never changed name.  Worked well until the project got nuked.  I'd be
happy to share the scripts I used.  Most of them came from SysAdmin
magazine (the ones for coordinating sockets).

John
   John Stoffel - Senior Unix Systems Administrator - Lucent Technologies
         stoffel AT lucent DOT com - http://www.lucent.com - 978-952-7548