Re: RMAN in RAC

"Howard J. Rogers" <howardjr2000_at_yahoo.com.au> wrote in message news:<asEJ9.1098$jM5.3154_at_newsfeeds.bigpond.com>...
> The basic principle is that the node name is not stored in the database
> controlfile or the RMAN catalog (whichever you choose to use), so RMAN must
> be able to find all the files it needs from the node where you perform the
> recovery.
>
> Without a cluster file system, that means you really need to use the NFS
> mount point solution. However, it *is* possible to copy all the required
> logs onto the node where RMAN is being run, before invoking RMAN.
>

Then, it is possible ? But what must be done in order to recover and recover the database from these moved files. Must I allocate channels towards the new directories ?

> That is not considered an optimal solution, however, because it means
> (assuming you copy the logs as part of recovery operation) that you are
> copying logs onto one location merely to have RMAN replay them on the node
> you've just copied them from. That's a lot of old network traffic for no
> particular reason.
>
> I'm not entirely sure what makes NFS read mount points so 'dangerous', since
> that is certainly Oracle's own (strong) recommendation.
>

There are two points:
1/ Oracle really prefers CFS.
2/ My client does not like NFS

> I also have no idea what you mean by the 'width' of a backup.

By width, I mean the size. Sorry for my english :-) When I make statistics on my tables in order to know the exact size of my data, I see that the backupset of the corresponding datafile is bigger than these statistics results. What does that mean ?
>
> Regards
> HJR
>

Thanks a lot for your answer. I hope you will send to me a new one :-) Received on Wed Dec 11 2002 - 11:29:54 CST