SUMMARY: Backing up Oracle-databases, NSR 4.4/DU 4.0D

From: Thomas Strandenęs <thomas.strandenaes_at_adm.uit.no>
Date: Mon, 31 Aug 1998 10:35:12 +0200

Honors go to:

- Steffen Gattert [sgattert_at_compuserve.com]
- bryan.fredrick_at_pca.state.mn.us
- Tom Webster [webster_at_ssdpdc.lgb.cal.boeing.com]
- Kris Chandrasekhar [Kris.Chandrasekhar_at_digital.com]
- NW User [nwuser_at_mailcity.com] (full-time Networker user?)
- Peter Ondruska [petino_at_hotmail.com]
- Nguyen, Ky [knguyen_at_ab.bluecross.ca]

for helping me out on this one.

Original question:

|I'm trying to set up a custom backup script for backing up our
|Oracle-databases. We're using DEC NSR 4.4 (The Last DEC NSR?), and
|since the 'savepnpc'-feature has not been implemented in this
|version, using the 'backup command' field in the client setup
|seems to be the only choice for taking down the databases before
|the backup starts. However, the backup command is invoked
|once *per filesystem*, and I don't want to shutdown/startup the
|database once per filesystem. This is impractical for a number of
|reasons. I want to do shut down the database before the backup, backup
*all*
|filesystems, and then resume operations. I've tried to make *one*
filesystem
|entry per client, create a 'backup command' script, and then call
|'save/savefs' in the script for each filesystem actually installed on the
|client, and then start up the database systems. This works to a certain
|extent: the filesystems seem to back up properly, but the file system
|listed for backup in the client setup always fails (Unsuccessful save set)
|in the savegroup after one retry. My backup scripts return errorcode 0, and
|no individual command called in the backup script fails, so this should not
|be the cause of the saveset failure. Why does it fail? I've also tried
|bypassing the 'savegroup' functionality altogether, and calling 'savefs'
|from a script and cron, but I want to keep the 'savegroup completion'
|features, that prints
|the bootstrap information etc.

Summary:

I got some hints on how I could return text from the custom script
to make NSR believe 'save' was a success, and indeed a few people relied
on this approach. However, Steffen Gattert & Kris Chandrasekhar suggested
I invoke a script with CRON on the hosts, which takes down Oracle
before invoking 'savegroup'. This way I preserve the savegroup functionality
without risking a unsupported setup (custom 'save's etc).

I've now written a Perl script that

1. Takes down Oracle on all the machines (using rsh on remote hosts)
2. Forks off a savegroup process
3. Waits for savegroup to finish, with a predetermined timeout
4. Takes up Oracle on all the machines, eventually after stopping
a hanging (media request etc.) 'savegroup'
5. Mails me if something goes wrong with Oracle or NSR

This works great. I've not included the Perl-script here - contact me for a
copy.

Great list! Thanks a lot for your help & time.

With friendly regards,

--
Thomas Strandenaes
Computing centre
University of Tromsoe
NORWAY
Received on Mon Aug 31 1998 - 08:32:16 NZST

This archive was generated by hypermail 2.4.0 : Wed Nov 08 2023 - 11:53:38 NZDT