Re: Creating standby database: Why need to copy backup pieces to standby site manually?

On May 15, 1:48 pm, Ronny <nitely..._at_ist-einmalig.de> wrote:
> Hello,
> I created a standby database (version 10.2.0.3 on Linux) by following
> the instructions of metalink Note:183570.1 (which was written for
> version 9i). Simplified spoken, you have to do the following (executed
> on primary site), which I did successfully:
>
> a)
> rman target /
> RMAN> Backup Database;
> RMAN> Backup current controlfile for standby;
>
> b)
> Copy the backup pieces to the remote server (same directory).
>
> c)
> rman target / auxiliary sys/change_on_install_at_STANDBY
> RMAN> duplicate target database for standby dorecover;
> ...
>
> Now my questions:
> - Why is it necessary to transfer the files to remote standby server
> with OS commands (step b)?
> - Why does Oracle not use SQLNet to transfer the files? (Or do I just
> not know how to do so?)
> - If not possible today: Does Oracle plan to provide this feature in a
> future release?
>
> My considerations behind:
> At the point of time, when the standby database is setup ready and
> running well, it receives the archived redolog files from the primary
> site over a SQLnet connection (--> log_archive_dest_2 =
> 'SERVICE=STANDBY LGWR...' or similar) and there is no need to copy any
> file either with OS commands nor over a NFS share. The file transfer
> is done by only using the SQLnet protocol. So I wonder why Oracle does
> not use this technology for the RMAN 'duplicate .. for standby'
> command as well...?
>
> Thank you for your posts.
> NitelyJoy

I was going to say what David did, but he beat me to it.

In addition, note that it isn't really that you have to copy them over, it's that they must be available to the standby environment. If they happen to be on the same machine, no problem (as long as you are careful!). If you have a way of telling the standby how to get to the files on the primary environment, go right ahead. Oracle doesn't really know in advance how far (topologically) apart the two are. They may be a few hundred miles apart with a crappy partial T1, they may be in the same room with shared SAN, they may be on the same machine, they may be on different continents. The db may be 20M or 20T. Initializing the standby is based on the size of the db, keeping the standby going is based on the transaction load, the two will likely be quite different. But it is ok to wonder why they didn't at least make it possible to use sql*net. My guess is it was just a hack of recovery to begin with, 3 major releases ago (and standby was one of the reasons I thought 7.2-->7.3 should have been a major release), when sql*net and networks in general were not as mature as today.

jg

--
@home.com is bogus.
It's all fun and games with Elmer Fudd, 'til someone gets their haid
blowed off.  http://www.signonsandiego.com/uniontrib/20070515/news_1n15spector.html