I know alot of issues have been brought up on how to fix the PATHETIC way
that Veritas backs up the catalog database.  I'm trying a new method and was
hoping I could get people's opinion on it.  I'm over a single tape for my
database and needed to implement something that worked

Using a Solaris 8 box with two luns from a t-3 attached.  (200GB approx each
lun). 
Using Netbackup 3.4.1 w/ dlt tape drives in multiple  ATL L1000 attached to
multiple media servers. 

Lun 1 is  where /usr/openv is mounted. Lun 2 is a shadow of this.
I use netbackup's regular schedule program to back up all of /usr/openv.

the bpstart_notify runs a subprogram which unmounts my Lun 2, syncs the mirror 
and then breaks it. it then remounts lun 2 as a broken mirror.

I then back up lun2  on a seperate pool (with its own labels).  
since I'm over a single tape the backup spans tapes.  After the backup is 
finished bpend_notify emails the backupid.  I then take the backupid, plug it
through bpimagelist -L -backupid <backupid>.  Print it out, and remove all
tapes associated with that backupid.  The tapes are then sent offsite with a
hardcopy of the bpimagelist in EACH case.  the bpimagelist is also put into a
file for ease of use in case I only lose the master server, and not the whole
data center.

I've used the tar in netbackup to test the restore, and since I'm not 
multiplexed
I've been able to do multiple tape restores w/ this method.
The advantage I see is that I can restore the whole /usr/openv environment in
one fell swoop.  The only thing I'm missing is the /etc scripts.

Does anyone see anthing that I may have missed, or could fix to make this
process better.  Does anyone think it won't work?  If so please comment.  I
just want to be able to restore the data :)