Oracle FAQ | Your Portal to the Oracle Knowledge Grid |
![]() |
![]() |
Home -> Community -> Usenet -> c.d.o.server -> Re: Weird Backup, can it be done without resetlogs?
"Francisco J. Garcia Humphries" <fjgarcia_at_e-humphries.net> wrote in message news:<burtni$6q0$1_at_news.ya.com>...
> My company is implementing a backup solution for a 2 instance (A and B) RAC
> based on the following:
>
> - At a certain time we start mirroring the database (datafiles,
> controlfiles, redologs and archives) to a disk that gets shared with server
> C.
>
> - After about two hours, when the shared disk is an exact live copy of the
> database, we issue an "alter tablespace begin backup" on the database.
> Detach the mirror disk (so it doesn't get changed any more)and the "alter
> tablespace end backup" on the production database.
>
> - At this point i would expect the database on the mirrored disk to be
> pretty similar to a database that has just crashed ¿right?.
>
> - The backup strategy is based on a new instance on server C that opens the
> database files on the mirror disk and then get's backued up by third party
> software. The problem we are encountering is that whenever we try to open
> this copy of the database with a new instance
> we need to do a RESETLOGS in order to be able to open the database. I don't
> get why i should have to open resetlogs and i need to be able to open the
> database without RESETLOGS
>
> Does anyone see a problem in the way this process copies the database? I'm
> thinking there might be a problem for Controlfiles and Logfiles. Is there
> any problem with the database being opened by a single instance when it was
> a RAC at first?
>
> Any help will be greatly appreciated.
If you are on version 9+ then look at the suspend command. You would want a global checkpoint to occur immediately prior to suspend and you would need the online redo logs and control files to be part of the mirror.
I do not see the point of the begin/end backup. It would cause the dirty buffers for the affected files to be flushed to disk, but a log switch on each instance would do the same thing for the entire db.
HTH -- Mark D Powell -- Received on Fri Jan 23 2004 - 18:31:12 CST
![]() |
![]() |