Oracle FAQ Your Portal to the Oracle Knowledge Grid

Home -> Community -> Usenet -> c.d.o.server -> Re: Backup & Restore

Re: Backup & Restore

From: joel garry <>
Date: Thu, 20 Sep 2007 11:38:03 -0700
Message-ID: <>

On Sep 20, 6:45 am, Brian Peasland <> wrote:
> Ravip via wrote:
> > Hi Brian,
> > Thanks for comprehensive detail. Obviously I'm not in a position to lose data
> > generated in last minute as well. Again as you said it depends on backup
> > strategy.
> > My clarification is once the incremental backup taken on yesterday night,
> > should I need to retain the
> > the backup of archive redolog backups taken prior to yesterday's backup.
> > Because the db incremental backups are available from yesterday and previous
> > days. As I mentioned earlier that the backup of DB and archive log files are
> > executed separately.
> > Thank You.
> You will not need the archived redo logs from prior to the incremental
> backup. However, if the last incremental backup was unusable for some
> reason, you would need the archvied redo logs since the previous
> incremental backup. Just to be safe, I keep the archived redo logs since
> the previous incremental level 0. And I keep at least 3 level 0 backups
> on hand. So if I generate a level 0 once a week, I'll have 3 weeks worth
> of backups on hand. To make sure that I can use any of those backups, I
> keep the archived redo logs for the last 3 weeks.
> Keeping more than one generation of your backups is an old practice. The
> most recent backup is generation 0. The backup before that is generation
> -1. You may have generation -2 and maybe generation -3, depending on
> your retention policy.
> The reason to keep multiple generations is that backups are often stored
> on tape media and tape media can fail. If there is only one backup and
> that tape fails, the safety net is lost. But by having more than one
> generation, we ensure we can get access to at least some of the data.
> The other reason to keep multiple generations around is that one may
> discover a problem (i.e. bad data) and that problem is in the backup.
> One may occasionally desire to restore before that data became a
> problem. This is hard to do with only one generation. Multiple
> generations do not guarantee success here either.
> In your reply, you stated "Obviously I'm not in a position to lose data
> generated in last minute as well.". This makes me ask...are you doing
> some sort of log shipping? Are you ensuring that your logs switch and
> get archived at least once a minute? If not, then you may be at risk of
> losing more than one minute's worth of data. Have you looked at Data
> Guard in your data protection strategy?

This and the previous post about requirements are an excellent description. I only want to add a couple of things:

There are some strange situations that can be gotten into with respect to whether older archived logs are needed. But rather than try to be paranoid enough to figure them all out, periodic test restores will highlight any issues, as well as validate written procedures and keep people practiced in skills which are rarely used, but needed in critical situations.

Redundancy of archived log backup redundancy is a good thing! :-)


-- is bogus.
It don't take a weatherman to punch a bozo in the nose.
(Actually he shoved the guy down the rocks)
Received on Thu Sep 20 2007 - 13:38:03 CDT

Original text of this message