Re: Looking for a good database shutdown script for Unix
Date: 19 May 1993 17:33:43 GMT
Message-ID: <1tdr1n$p24_at_usenet.rpi.edu>
In article <1tb1da$2r1_at_spock.dis.cccd.edu>, markb_at_spock.dis.cccd.edu (Mark Bixby) writes:
|> At the present time, we run ~oracle/bin/dbshut before shutting down the
|> underlying Unix system. This script does a sqldba "shutdown" command for each
|> database listed in oratab.
|>
|> But recently we've encountered some sort of hang problem where the database gets
|> seriously confused, and a standard sqldba "shutdown" command hangs forever. We
|> then have to kill off sqldba, and manually kill the associated oracle processes.
|> Yuck. :-(
|>
|> I'm wondering if anybody out there has implemented an Oracle shutdown script
|> that is smarter about shutting down hung databases. I.e. one that tries
|> "shutdown normal", and if it is still hung after x seconds, kills that attempt,
|> and then escalates to "shutdown immediate" or even "shutdown abort".
|>
|> Please post or e-mail; I'll summarize any e-mail replies. Thanks.
|> --
|> Mark Bixby Internet: markb_at_cccd.edu
|> Coast Community College District 1370 Adams Avenue
|> District Information Services Costa Mesa, CA, USA 92626
|> Technical Support (714) 432-5064
|> "You can tune a file system, but you can't tune a fish." - tunefs(1M)
We ran into the same problem on our Unix (IBM AIX) system. We do cold backups
every night as part of our regular file system incremental backups so we shut
down our databases at the beginning of the backups in order to be certain we
have consistent backups. Dbshut as Oracle distributes it, performs a
"shutdown normal". The problem we ran into was that if a user forgot to exit
the Oracle application when they left for the day, dbshut would wait forever
for the (long since gone) user to exit the application before shutting down
the database. Our solution has been to change dbshut to issue a
"shutdown immediate". We have found that this solution has never failed to
cleanly shutdown the database except for one time when there really was a
problem with the database.
-- +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ E. Andrew Mondore Internet: mondore_at_rpi.edu Systems Programmer Bitnet: userfmc6_at_rpitsmts.bitnet Rensselaer Polytechnic InstituteReceived on Wed May 19 1993 - 19:33:43 CEST