hanging shutdowns

Hello all,

I have a small dilemma I need to resolve.  We have several large databases
that do not shutdown quickly due to active processes (usually AQ stuff but
not always) and this has caused problems with cold backups in the past.
(I'd prefer to stop the colds entirely, but have not yet won that debate.)

Another dba wrote a script which first kills all unix processes connecting
to the database, connects to the database to kill all sessions except the
sys connections and then issues a shutdown immediate.  We now have another
instance on the server so I need to modify the scripts for the new instance
but I'm not comfortable with the current approach.

Given that I need to have something that works by next Sunday morning, what
is the best way to structure these scripts?  Continue to kill everything?
Issue a shutdown, find the remaining active transactions and kill only
them?  Issue a shutdown abort, restart, shutdown cleanly and then run the
backups?   I've seen discussions on this subject in the past, I'm just
wondering if there's any new solutions or concerns on the subject.

Databases are Oracle 9.2.0.6 on HP-UX approximately a TB in size.  (some
slightly bigger, some slightly smaller)

recommendations appreciated ...

Robyn

Other related posts: