Re: Sounds good in theory, but will it work?

TB wrote:
> We have Oracle 8.1.7 running on an IBM p615 single processor server.

Oracle DataGuard does not exist in 8i R3. It was introduced in 9i and is quite mature in 10g. Perhaps if you upgrade to a newer, supported release you can use provided utilities to meet your objectives without rolling your own. The practice of managing a standby database within Oracle is a time-honored one, whether its managed (dataguard) or handled by roll-your-own scripts.

> I'm an AIX admin, not an Oracle admin. We don't have an Oracle admin
> on staff.
>
> I need to ensure some level of recovery (somewhat quickly) in case of a
> total disaster; building burns down, tornado hits,... whatever. I'm
> not looking for anything fancy and costly like clusters or
> (a)synchronous replication. What I do have is a full hot backup every
> night with RMAN and an offsite DR facility.

does the rman backup include archived redo logs. wait, is the database running in archivelog mode? If not, are the rman backups performed in a mount state? have you ever restored such backup sets and recovered them? It is Backup/Restore/Recovery after all, not just "backup".

> Here are my thoughts...
> Buy a second AIX server (I have a quote for a p520 dual proc) for our
> DR site. Backup the rootvg (mksysb). On the weekends shut down the
> Oracle database and do a savevg on the data vg's. Bring Oracle backup
> and continue to do full RMAN backups durring the week.

Depending upon the size of the database and transaction volume (redo log generation) it might be more efficient to only ship the archived redo logs. If unrecoverable operations take place, they are not propagated in the archived redo log stream and other methods would need to be used (such as what you describe above).

> Build the new server at the DR sight with the mksysb. If a disaster
> happens, then I could take the savevg(s) to the DR facility and restore
> those vgs. This should give me a fully functional Oracle server as of
> the weekend backup. Then I could use the RMAN backups to restore to
> the weekly full backup. With this process we could potentionally lose
> a full days work, but we are comfortable with this risk.

Congrats on determining this design parameter (acceptable loss of data, "recovery point objective")

> Has anyone done anything like this? Think it will work? What are your
> thoughts?

After the day's worth of data is lost in this database but exists in other databases you might think otherwise. If it really is stand-alone data or can be easily re-entered/reloaded good for you. You really don't know until you tell the application owner "we have lost 5 hours of transactions due to a corrupt archived redo log that we cannot recover through".

What is the recovery time objective, or how long after the disaster is declared is the system expected to be up and running? A standby database might reduce the mean time to recovery.

You appear to be well along your way to producing a solution. Seeing that you're on 8i R3, you will have to do alot more work yourself rather than leveraging functionality provided by the vendor that you may already be licensing.

-bdbafh Received on Thu Nov 09 2006 - 00:23:39 CST