Have you thought about a user-based backup?  You could do away with the
filelist altogether because the filelist would be build on the client.

I've got big backups I do on demand ~500G to 1500G.  The filelist constantly
changes. 

I've got a script that starts with a raw list of files, you could generate a
list like this using the find command or similar.  I then feed it into a
script that breaks the list into four parts.  I used to break it up just
using filecount, the fist 25% to a filelist1, the second 25% to filelist2,
etc.  I later smartened the script to divide in such a way as to balance the
backup sets by size.  

Anyway, each list is fed to a a bpbackup command with the "-f filelist"
option.  "bpbackup" isn't privileged so even your DBA's could kick it off as
the DB username.  Each bpbackup sends the list of files to a policy that has
a user-defined scheduled.  This is where you keep control using MPX factors,
storage units, etc. so that you get as many tape drives involved as you
wish.

So the pattern would be:
1. export DB
2. Generate list of files
3. Break list into filesets
4. send each fileset via bpbackup.

It could be driven entirely by your DBA's, too.

HTH - Mark

-----Original Message-----
From: Alfonso Correas [mailto:alfonso.correas AT arrakis DOT es]
Sent: Friday, November 07, 2003 4:06 PM
To: veritas-bu AT mailman.eng.auburn DOT edu
Subject: [Veritas-bu] NB 4.5 very large backups and Streaming.


Hi to all:

  I have a question. I've been doing some tests but none has given me a 
suitable result for mine problem.

  I'm in charge of a backup platform who does a big number of backups, 
some of those backups are considered 'very sensible'. Here's the 
scenario:

- 1 Master server running NB 4.5 under HP-UX 11
- Some media servers (some on shared storage and some on dedicated 
drives).

I have some servers who does very large backups (>1Tb daily) via an 
export full (the admins of those machine choosed to do an export full 
rather than doing RMAN). And they export all the data on a structure 
like this:

/oracle/T11/sapdata
...
...
...
...
...
/oracle/T11/sapdata

  Since this is an unattended process (both: the export & the backup) I'm 
trying to find a suitable way of getting all the directories where SAP 
exports the data and pulling into the library (the number of directories 
can grow without previous notice) and I need to copy all.

Tried many attempts.

Case 1: Explicit declaration of all directories like....
NEW_STREAM
/oracle/T11/sapdata00
/oracle/T11/sapdata01
...
NEW_STREAM
/oracle/T11/sapdataxx

  This approach is unmanageable because I need to keep an eye over this 
server daily to check if there's any new directory to do backup. But I 
can control on how many jobs is divided the complete job and use two 
drives, take profit of the array of disks and the HBA's.

Case 2: Backup of all directories named sapdata like...
NEW_STREAM
/oracle/T11/sapdata*

  This isn't acceptable, because puts all the copy under one only big job 
and wanted to use the two dedicated drives I have on a Media server for 
this duties.

Case 3: Use expressions to divide and make the backups:
NEW_STREAM
/oracle/T11/sapdata0[0-4]
/oracle/T11/sapdata1[0-4]
NEW_STREAM
/oracle/T11/sapdata0[5-9]
/oracle/T11/sapdata1[5-9]
...
...
...
NEW_STREAM
/oracle/T11/sapdata8[5-9]
/oracle/T11/sapdata9[5-9]

This case is not acceptable, because some directories not exist and the 
software expands those expressions and try to do some jobs and the 
directories doesn't exist (it gaves an Error 71) and shoots over the 
IT/O console, giving scaring the operators by thinking that one or more 
directores are lost.

Any other suggestions for doing this backup without needing to check 
daily the number of directories and modificating the policy when is 
needed?

Best regards,
  Alfonso Correas

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