Re: Why I don't like RMAN repositories

From: Gus Spier <>
Date: Tue, 10 Dec 2013 12:30:41 -0500
Message-ID: <>

I'm an advocate of 21 days worth of history. We work with an undisciplined, though brilliant, group of users and developers and a thoroughly unreliable storage architecture. A covert error in coding, DML, or the "persistence layer" may not (and usually doesn't) become apparent more than a week after it was implemented. The capability to recover to the point in time of the error is invaluable.


On Tue, Dec 10, 2013 at 10:09 AM, Chris Taylor <> wrote:

> Out of curiosity, with nightly FULL rman backups of a production database,
> why would you recommend a value greater than 7 for keep time?
> I had a hard time imagining a scenario where you'd have to go back 21 days
> for a production recovery - I'm wondering if I'm missing some technical
> aspect here...
> Thanks,
> Chris
> On Tue, Dec 10, 2013 at 9:03 AM, Tim Gorman <> wrote:
>> Coming late to the discussion, not sure if anyone else has made this
>> point...
>> Having an RMAN repository (a.k.a. recovery catalog, etc) is belt and
>> suspenders (or belt and braces for many). There is always a "recovery
>> catalog" in the target database's control files, so you've always got a
>> belt to prevent your trousers from falling. If you don't set
>> CONTROL_FILE_RECORD_KEEP_TIME to at least 21, then it's more cheap string
>> than a belt, but it's a belt.
>> A recovery catalog is a replicated copy of the recovery catalog with more
>> history, thus suspenders/braces in addition to the belt.
>> It is not technical merit, but rather personal/corporate choice, that
>> determines whether one wears one or both.

Received on Tue Dec 10 2013 - 18:30:41 CET

Original text of this message