RE: Backup on standby database

I'm rusty. But does is cumulative incremental still part of RMAN? I *thought* that accomplished keeping it to one L0 and one L1 to achieve a recovery, and it seems that may be a useful consideration in this conversation. I don't *think* they ever implemented merge L0/L1 to produce a new L0 having only the last version of every block, which is what I actually suggested as part of MOSES, but I thought the "no more than 2 required for recovery" approach with incremental cumulative gave me little room to continue bit...., er, complaining.

I had opened Rightsizing, Inc. with Jerry Ireland and was doing billable consulting by the time it was ready to kick the tires on RMAN. The three key ideas of RMAN as I understood it was to make it hard to get it wrong (1), only have to read blocks Oracle had used in the datafiles (2), and (3) read the blocks the same way they are read in a select to make sure the blocks in the backup are valid. Most of my clients were still using "hot" backup with split mirrors to make them quick while RMAN was still too green to rely on when market demand switched me mostly to either repairing horrible media infrastructure choices that made the disks slow enough to be important or repairing queries so that the media farm was the limiting factor. (And then things moved to speed of memory into chips...)

Anyway, back to my question: Are incremental cumulative backups relevant to this discussion?

mwf

-----Original Message-----
From: oracle-l-bounce_at_freelists.org [mailto:oracle-l-bounce_at_freelists.org] On Behalf Of Mladen Gogala Sent: Friday, May 14, 2021 2:31 PM
To: Andy Klock; oracle-l_at_freelists.org Subject: Re: Backup on standby database

Hi Andy!

Yes, level 0 still requires scanning of every block and it will take longer to do L0 than it would take to do L1. However, that is offset by the fact that you have only a single backup to restore. And you must restore a full backup. The performance hit doesn't really matter since the backup is taking place on the standby, as per OP. The best of both worlds is what Pap has said: to have a backup suite capable of merging L1 and L0 backup into a new L0 backup. That is known as "synthetic full backup". He essentially keeps doing incremental backup forever and merging them into a new full backup on the daily basis. I would really like to know which backup software he's using.

Commvault can do that, but it requires storage snapshot. I am not sure that Commvault can snapshot ZLDRA, Oracle would have to reveal their internal mechanisms on the appliance for that, so it's probably not Commvault.

Regards

On 5/14/21 9:26 AM, Andy Klock wrote:
> Admittedly, it's taken me a bit to follow your line of thought. As an
> example, if a level 0 is taken on Sunday and the six incremental
> backups are taken on consecutive days, you are asserting that it would
> be better to instead take level 0s every day.
>
> I can see how deduplication can speed up writes and possilby conserve
> space for the backups, but there is still is a performance hit with
> this strategy, right?
>
> Level 0s still require scanning every block which takes time, so with
> BCT, RMAN already knows what has changed and the incrementals are
> going to complete much faster than dedplucated level 0s.
>
> Do you have some test cases that prove this strategy?
>
> Thanks,
>
> Andy K

--
Mladen Gogala
Database Consultant
Tel: (347) 321-1217
https://dbwhisperer.wordpress.com

--
http://www.freelists.org/webpage/oracle-l





--
http://www.freelists.org/webpage/oracle-l