Re: Standby database as a reporting database ?

"Christian Svensson" <chse30_at_hotmail.com> wrote in message news:ccc2a7eb.0109280523.1515cfde_at_posting.google.com...
> Greetings !
>
> I am studying for the Backup & Recovery certification (Oracle 8i) and
> wonder if there are any means to make a standby database as a
> reporting database.

Yes, you can cancel recovery on the standby, shut it down, mount it, and open it read only.

>
> What I have read so far is that if you open the standby db, it gets
> out of sync from the production db and will not apply the archived
> redo logs anymore.
>

Correct. For it to be read only, it cannot be in managed (or manual) recovery mode, and therefore archives cannot be applied to it for the duration. There's nothing stopping you putting it back into recovery mode, however, and applying the 'backlog' of logs.

> I cant be the only person thinking this is a waste of resources.

I don't see the waste of resources. A true standby is, in every sense, a complete waste of resources -until disaster strikes on production, when it becomes an extremely valuable investment. In 8i, Oracle has even allowed for such a set of resources to have an interim use, as a read-only database. I think you're being a touch hard suggesting that both goals could or should be met by the one entity.

>There
> must be some ways to accomplish this. How is it in Oracle 9i ?
>

It's pretty much exactly the same in principle, except that there's now something called 'Data Guard' that can be used to absolutely guarantee no data loss on the standby, or even no data divergence (because LGWR can be used to directly write redo from the Log Buffer in the primary site to the Standby Redo Logs on the standby site). So the levels of standby protection have been beefed up (if you want them to be). They've also made it easier to switch back and forth between standby and primary sites without major consequences for either. They've also permitted multiple standby databases, some of which could be 'zero divergence', others having a time delay built in before redo from production is applied.

The fundamental idea of a standby not being useable in any form whilst logs are still be applied to it remains, however.

Regards
HJR
> Thanks for any ideas.
>
> /Christian Svensson
Received on Mon Oct 01 2001 - 15:13:02 CDT