Multi Mastering - backups?

[Ratzie]

So we are new to multi-mastering.
10.2.0.2
We are not using streams, but advanced replication.

We are just trying to wrap our heads around how to do backups.
If we are using multi mastering, are backups required, if so what?
From what I am thinking, if we have a catastrophic failure, we confirm which is master core. Severe multimastering, completely install new database, and treat as new core and restart multimastering, and all data will then sync across.

So what really needs to me backed up?
This is really confusing...

If we do a hot backup at the same time, on both cores, and we have a problem, and do a restore on the problem server, with this create any issues with Multimaster transactions?

[Steven E. Protter]

Shalom,

As I understand this concept multiple backup streams go to the same tape, making the backup go much faster.

With some versions of Veritas this makes the restore single stream with many passes on the tape. This makes DR painfully slow.

Test your technology on the restore side at your DR site.

SEP

[Duncan Edmonstone]

SEP,

Not sure what you are talking about there - this is a question about Oracle Streams replication.

LHradowy,

I'm afraid I don't know the streams technology at all - but I can't imagine it would be a replacement for a good local backup solution...

Have you looked at the part of the streams manual that talks about recovery?

http://download.oracle.com/docs/cd/B19306_01/server.102/b14228/man_gen_rep.htm#sthref627

HTH

Duncan

[Volker Borowski]

Well,

you surely need a backup!
Think of untechnical errors like a user who deletes wrong data.
This will be a true and solid DB-transaction, which might be surely replicated anywhere else and if that already happend before you noticed... busted.

In terms of databases for the same reason:
- Mirroring is no technique that saves you from taking a backup
- Clusters are not a technique that saves you from taking a backup
- Replication is no technique that saves you from taking a backup
- export is no backup at all

Only a backup gives you the opportunity of doing a point-in-time recovery, meaning turn the inside db-clock backwards.

The same applies for a crash situation. Do not think of a db-backup as a filesystem backup, which you restore and get everything back to yesterday 04:00 am.

In case of hardware error, you will always want to recover every transaction and not yesterdays last backup. So you'll have to use yesterdays backup, but require db-logs up to the point of crash to get everything that happend.

By the way, coming back to my user-error scenario: You have a strategy to recover from this, have you? Otherwise I'd recommend to think about this urgently.

Volker

[Yogeeraj_1 · Accepted solution]

hi,

Allow me to add one further step to what Volker has posted above.

The backups need to be tested. Even better actual recovery must be done on the DR or recovery site as frequently as possible. This way you ensure that both your system is running and the backup is still valid.

If you use RMAN backup, there is one command that allows you to validate your actual backup without the need to actually doing the real restore on a server:
restore validate database;

You should also look into Incremental Merge RMAN backup option for faster recovery...


hope this helps too!

kind regards
yogeeraj

[Ratzie]

Thanks, I knew I was going to get this type of feedback...