Mark,

we have few dozens of databases only so I can't really help you. The
only thing we found helpful is calendar-based scheduling - at least this
way we were able to separate file system and DB backup plus separate
Full and Incremental backups for SQL servers.

Dmitri

-----Original Message-----
From: markjessup AT northwesternmutual DOT com
[mailto:markjessup AT northwesternmutual DOT com] 
Sent: Sunday, November 16, 2003 9:44 AM
To: veritas-bu AT mailman.eng.auburn DOT edu
Cc: briandiven AT northwesternmutual DOT com; markjessup AT northwesternmutual 
DOT com
Subject: [Veritas-bu] Backing up Thousands of Databases

Hi All,


I am wondering what method or strategy people use for backing up
application databases such as sybase, oracle, or db2. We have over 1500
databases in our shop and are in the middle of migrating to Netbackup
Datacenter from an older mainframe based system. We currently use the
Netbackup database extension for each flavor of database out there,
Sybase, Oracle, and UDB so we are doing hot online backups.  We backup
each individual database and the way we have it set up today is with a
separate Netbackup policy for each database dump and also for the
database log dump. This method has introduced over 1200 policies to
Netbackup so far and it has caused performance issues when expanding the
policy tree in the Windows Admin console. It takes us over 3 minutes to
bring up the initial policy list.
The bigger issues is that with each database being a separate policy, we
lose almost all control over backup management because all of the
database backups are not part of a single policy but are individual
policies. With this method, we also experience status 58, 202, and 205
errors alot nightly because each database backup which is a separate
policy can go back to the same physical Unix client so we could have
many, many database backup jobs trying to run and connect to the same
Unix database server at the same time and then trying to communicate
back to our media server. This is all being done within Netbackup's
internal scheduler with hot backups to disk or tape based on the size of
the database.  We have increased kernal parameters, introduced local DNS
cachin, etc with little improvement on the status 58, 202, 205 when
Netbackup does the scheduling.  We have the default of 10 mintues set
for the scheduler interval. I was wondering how other big database shops
backup application databases, what process or method and how you
schedule them and monitor the database backup jobs.
Also, is this process supported by the central backup team or by the
database admin team for these application database backup jobs. 


Any feedback, ideas, or thoughts on this topic or what I described above
regarding our current process would be greatly appreciated. Thanks!





Mark Jessup
Northwestern Mutual