We experienced the same issue.  The issue is the lack of communication
between the various stages in the process.  When the backup completes
(successful or failed) it calls bpend.  Once bpend is called there is no
more communication with NB.  So the problem arises on failures because the
scheduler can retry before the bpend is complete.  This causes bpend and
bpstart to trip over each other.  I had wanted to add checking in
bpend/bpstart that would be aware of retries and failures so that they could
avoid issues.  However, I could not get any feedback on the failure
conditions that won't retry.  So, the dba added a test to bpstart to see if
bpend was running.  This solved the compounding problems of bpstart/bpend
running simultaneously but there are times when the database shuts down and
starts up 3 times in a night for failed backup attempts.

-----Original Message-----
From: TWGIBBAR AT up DOT com [mailto:TWGIBBAR AT up DOT com]
Sent: Thursday, May 03, 2001 2:14 PM
To: veritas-bu AT mailman.eng.auburn DOT edu
Subject: [Veritas-bu] bpend_notify Question


We have a situation where were using home grown built oracle scripts for
hot database backups.  Everything works just great with the bpstart and
bpend scripts unless the job fails with a media or time-out error.  The
bpend doesn't run to clean things up before the job retries and reruns the
bpstart script.  This causes errors to pop up in the alert log because the
tables are already marked for backup and also causes issues when it
completes because everything isn't always properly cleaned up.  I guess my
question is, how do I ensure the bpend script runs if the jobs fails after
the bpstart has already ran?  I hope this mumble jumble makes sense to
someone.

Thanks,
Terry

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