Re: Archived redo logs

"Jurjen Oskam" <joskam_at_quadpro.stupendous.org> wrote in message news:slrnbf1h24.m3s.joskam_at_calvin.quadpro.stupendous.org...
> In article <u471fvkt992g5d4ufh0ueee8qeab78heog_at_4ax.com>, Sybrand Bakker
wrote:
>
> >>The reason I ask is because in our current setup all files in that
directory
> >>are backed up and deleted, *except* for the most current file. The
reasoning
> >>for this is that the most current file might still be written to. I'm
not
> >>sure this reasoning is correct.
> >
> > this is exactly why you should issue an alter system switch logfile
> > and alter system archivelog all prior to backup the archived redolog
> > files, so you don't need to use this procedure.
>
> But (if I understand correctly) this still doesn't guarantee anything.
> Suppose I did this, but there are quite a lot of archived logs. My
> backup-script starts to process them, but in the meantime an online
> redo log fills up and starts to be written to a new archivelog file.
> My backup script comes along, finds this new, not-completely-written-yet,
> archived log file, and backs it up. Woops.
>
>
> I like the suggestion of querying V$ARCHIVED_LOG (sp?); that way, you're
> guaranteed that the files you backup are complete and finished.
>
> --
> Jurjen Oskam
>
> PGP Key available at http://www.stupendous.org/

The script in the DBA handbook deals with this in the following fashion - the list of logs present is stuffed in an environment variable. The rest of the script processes the env var. Anything that is created after the backup of archive logs start is not in it, will not be back upped and not deleted
- secondly the script stops automatic archival. Assuming the activity on the database is low, this shouldn't pose any problem

Of course RMAN has smarter ways of dealing with this, but basically it also makes a list of logs to be back upped.

Hth

-- 
Sybrand Bakker
Senior Oracle DBA

to reply remove '-verwijderdit' from my e-mail address