Another option, if you want to do this hot, and if your DB is in archive-log
mode, is to roll through the DB tablespace by tablespace.  We do this on our
hot DB's but it may take a while on your 800G DB (it's difficult to
multistream because you get too many tablespaces in hot-backup mode at once,
unless your DB is mostly read-only).

What we do, therefore on our small instances, because of RMAN complexity and
license expense, is what we call a "rolling hot backup".  
  a. Query the database for all tablespace names 
  b. For each tablespace, in sequence
     1. query oracle for datafiles related to the tablespace (place in a
file)
     2. place tablespace in hot-backup mode
     3. use bpbackup to send the list of datafiles to netbackup
     4. remove tablespace from hot-backup mode
     5. loop for next tablespace.
  c. Alter control file to trace and send that file via bpbackup, too.
  d. Roll the redo logs.
  e. Send relavent archived redo log files, this is the set that starts from
just before the first tablespace is put into HB mode up to and including the
last one that results from rolling the redo logs.

This makes a whole bunch of images, one per tablespace, that represents a
single backup of the entire database.  To track these as a set, we tie them
together with a keyword based on backup date & instance name.

Restoring is a pain compared to a cold backup because there's a lot of redo
logs to apply to get to a time-consistent image but it does work.

Otherwise, the other poster is correct, your pre-schedule & post schedule
scripts (bpstart_notify & bpend_notify) make great tools to start & stop a
database.  Then you can drive the entire backup from the Netbackup
scheduler.  I've even heard about folks who used the bpstart_notify job to
build a policy-specific include list of filenames so that they can limit the
files backed up on an active basis (rather than using a filelist in the
policy).  This may be good if your files are frequently changing locations
making a static policy filelist difficult.  Note, though, I've never tried
this myself.

HTH - M

-----Original Message-----
From: Deiter Scott [mailto:scottd AT HanoverDirect DOT com]
Sent: Thursday, September 04, 2003 12:59 PM
To: veritas-bu AT mailman.eng.auburn DOT edu
Subject: [Veritas-bu] User backup necessary ?


We need to backup up our dataware house client.
  Sun Ultra / Solaris 9
  Netbackup 4.5
  0.8Tbyte of oracle to be cold backup
The master server is the same platform/os.  In order to stop oracle we need
a script on the Client.  So scheduling the backup does not make sense due to
timing errors.  Trying to create a script on the client to use the bpbackup
command.
We will need to use the -f filelist option.  So how do we set the excludes?
I did create an exclude_list.TestBarney on our test machine.  it does not
appear that the exclude file was used during the test backup.  looking at
the docs for bpbackup I can not find any mention of excludes.

So  Does this really work?

OR

Is there a better way other than rman ?


>From   Scott Deiter
        Hanover Direct, Inc.
        Tech services
        (717) 633-3298
        scottd AT hanoverdirect DOT com

_______________________________________________
Veritas-bu maillist  -  Veritas-bu AT mailman.eng.auburn DOT edu
http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu