Re: Testing an rman backup

  • From: "Niall Litchfield" <niall.litchfield@xxxxxxxxx>
  • To: john.hallas@xxxxxxxx
  • Date: Mon, 11 Jun 2007 10:59:58 +0100

On 6/11/07, Hallas, John (EXP N-ARM) <john.hallas@xxxxxxxx> wrote:

I have never understood why you have to connect to the source database
and what happens when you do.


Rman reads the controlfile from the source(production) database. In the case
of a 'set until time' duplication I'd have thought that in principle Oracle
could read a backup controlfile, but for duplicate more generally you will
want the current controlfile information.

One issue I have had to battle with is change control. Even though you
are building a copy database on a test/development node you still have
to make a connection to the production system , which requires a CC.


I don't know if your organisation operates ITIL change management practices,
but if they do then this sort of activity would exactly fall under the guise
of a standard (i.e routine) change.  I'm quite agressive with change control
myself - meaning if in doubt record it, but it does seem to me that making a
copy of a production system absolutely is something that should be recorded
and approved (maybe that's the ex auditor trying to get out).

I raised a metalink call with Oracle on this matter several years ago
(probably v8) but I IIRC the answer was waffle and there was no good
reason.

And why oh why is the source database called 'target'


well the why is because that is what RMAN always calls the primary database,
but I agree it's an awful name, as is auxiliary. I quite like the idea of
calling the primary database the source and any auxiliary the target - even
for backups - since in my head we backup sources and copy to targets and
never backup targets. Still as Dave Ensor famously pointed out if you can
have a product called multi-threaded server that is neither multi threaded
nor a server what hope is there for naming.



--
Niall Litchfield
Oracle DBA
http://www.orawin.info

Other related posts: