RE: training for new DBA's

From: Herring, David <HerringD_at_DNB.com>
Date: Wed, 19 Feb 2014 08:09:34 -0600
Message-ID: <AD8FE6616C097545A4C9A8B0792909AC2E6BD80F4E_at_DNBEXCH01.dnbint.net>



Martin and all,

For some of our databases just a few minutes of unavailability means a world of hurt. For those we have a physical standby with RMAN backups performed on the standby (4-node RAC on both ends). The RMAN backup on the standby is a full Sun thru Fri with a full on the primary to a NAS on Sat. The NAS backup gets tested regularly by means of lower env. restores off it while the standby DBFLASH backup is tested officially once a year as part of a DR exercise. Unofficial tests happen periodically during the year due to issues with 10.2.0.2.

Dave Herring

From: oracle-l-bounce_at_freelists.org [mailto:oracle-l-bounce_at_freelists.org] On Behalf Of Martin Bach Sent: Wednesday, February 19, 2014 2:07 AM To: oracle_at_1001111.com; Dave Morgan; Oracle-L Subject: Re: training for new DBA's

Good morning everyone,

I have been following this thread and found it very interesting. So far all I did was lurking but the last post prompted a more general reply (and reminded me of Sir Alan Sugar).

So when a DBA is in a position to having to restore service, what do you do? I personally opted for Data Guard or similar replication technology in all the sites I worked. Today's database sizes make it difficult to restore within a reasonable amount of time. Unfortunately "reasonable" does not necessarily reflect what's technically possible but what the "business" is willing to tolerate.

Activating a standby removes the need to restore, but the problem must be suitable. Logical corruption will require a point in time recovery and you'd have to have another standby with delayed redo application to counter that, and a bit of luck. Having image copies of your database in the FRA seems like a nice thing to have but effectively doubles your space requirements. Those image copies can be rolled forward very quickly and cut the restore time to near 0 though.

So I wonder who has the time (-> management behind your desk) to first create a backup of say a 10 TB database and then begins working on the restore operation. Please don't get me wrong, I agree that you must be able to get back to where you started, but at the same time wonder if that is possible. While you are taking the backup you obviously can't start working on restoring the service. A split mirror ("BCV copy") would be almost ideal. Split the mirror, take the backup of the broken database while working on restoring service....

What is your experience?

Martin
Martin Bach
http://www.enkitec.com
http://martincarstenbach.wordpress.com
On 19 February 2014 04:40:32 CET, Dave Morgan <oracle_at_1001111.com> wrote: On 02/17/2014 08:42 AM, Jeff C wrote:
 Do you have them backup all the datafiles before starting recovery? If you have a 500g database that will take a long while?

Basic rule for a (production) DBA is he/she must be able to return to where they started at ALL TIMES!

If a recovery cannot be done because something is "changed or lost" then the outage is of infinite length. Therefore, the amount of time a successful recovery takes is irrelevant.

All it takes is one corrupt file in the backup/recovery system and a company can be in a world of hurt.

Ohh, I can't do that, it will take tooo long to do it properly. Wrong attitude, you're gone!

YMMV
Dave
i0zX+n{+i^ Received on Wed Feb 19 2014 - 15:09:34 CET

Original text of this message