Re: vxfs_checkpoint and database recovery

On Wed, Mar 19, 2008 at 11:50 AM, David Turner <dnt9000_at_yahoo.com> wrote:

>
> We're performing
>
> "alter system archivelog current;"
>
> before and after the backup, but not putting the tablespaces in backup
> mode. I assume this is the reason why we're having the problem, but can
> anyone explain to me the why this is required? With the snapshot/checkpoint
> as I understand it we are getting a consistent version of the datafiles,
> redo(which I remove before restore).
>

You can switch and apply logs all day, and you will never be able to open the database.
Without backup mode, the SCN in the datafile is getting updated fairly frequently, and is newer than your most recently applied archive log.

>
> Also, is there a different option that I should be using with Netbackup
> for checkpoints to make this work? In this phase we do not want to
> incorporate RMAN or any Oracle scripts if possible. The SA's want as
> complete a separation as possible.
>

I don't understand the 'separation' part. Something has to be done at the database level if you want to make a backup.

Why not RMAN? No backup mode required.

I don't know how vxfs checkpoints work, but any technology that creates a snapshot that is
updated with copy-on-write is going to incur IO from the source database on a file scan,
which is what a backup does.

Use RMAN incremental Level 0, Leve 1 with Block Change Tracking. (10g+ for BCT) Is your backup to Disk?

If so, you can experiment with "BACKUP DURATION .. MINIMIZE" load to decrease
IO incurred by RMAN. There's also ALLOCATE CHANNEL RATE=nK

-- 
Jared Still
Certifiable Oracle DBA and Part Time Perl Evangelist

--
http://www.freelists.org/webpage/oracle-l