>From time to time I've read in this forum about the scripts that have
been written to handle the NBU catalog backups. The scripts are
written to cover two major (IMHO) flaws in the "automatic" mechanism
for performing the backups: 1) only two tapes, and 2) no convenient
method to get just one backup a day onto a specific tape during a
quiet period.

They all follow the premise that you don't want the catalog updated
(a regular backup running) while the catalog backup is happening.
Subsequently you can't run a backup while the catalog backup is
running. (This usually accomplished by bprdreq -terminate.)

I also run VxVM and VxFS on my master server. I suspect a good number
of you do too.

I was wondering if it wouldn't be a good idea to insert a mount
command that uses the VxFS snapof option to create a read-only mount
of the database.

So the generalized procedure would be:
1) terminate the request daemon
2) check for active jobs and wait for them
3) mount -F vxfs -o snapof=<real db> /bkupdb
4) initbprd
5) backup the catalog
6) umount /bkupdb

Just so we're clear, I'm talking about VxFS snapshot mounts, not VxVM
snapshot volumes. You create a new volume sized a percentage of the
original volume's size and use the snapof= mount option. The new
volume then holds a copy of the data that is changed on the original
file system since the mount was made.

Since the NBU catalog doesn't change that much over the few hours
needed to push it to tape, one wouldn't need much more than 1%-5% of
the catalog size for the snapshot volume.

Anyone thought about this before?  Any reasons why this would be a
very bad idea?

Thanks, Jim

=====
Jim VandeVegt
vandevegt AT yahoo DOT com vandevegt AT avaya DOT com
There is no luck except where there is discipline.

__________________________________________________
Do you Yahoo!?
Yahoo! Mail Plus - Powerful. Affordable. Sign up now.
http://mailplus.yahoo.com