Re: database could not be started

On Sat, 9 Aug 2003 17:18:12 +0200, Anton Buijs <remove_aammbuijs_at_xs4all.nl>

>
> It's Burton's remark to backup the redo files and I did not react on it
> because frankly I am in favor to backup the online redos when:
> a. the database runs in noarchivelogmode
> b. and you don't use rman to make backups (rman never backup redos
> anyway)
> For 2 reasons:
> a. if you did a shutdown abort you need them to recover the database

If you lose all your redo logs, and you're not in archivelog mode, you'd restore your previous closed backup (which is consistent and doesn't contain redo logs), and you'd say 'recover database until cancel...cancel...alter database open resetlogs'. You've now got a fresh set of redo logs and the database can be opened. So there was no need to back the logs up.

> b. when you have them you can normally open the database and you have a
> real
> complete set; without the redologs you must do an open resetlogs.

See, even you know it can be done without having a backup set of logs on hand.

> Your reaction will probably be: "make sure you do a clean shutdown" and
> "what the heck, open resetlogs isn't so bad" but my reaction would be:
> not a
> problem to backup the redo's too and I am protected against these 2
> possible
> mistakes.

But vulnerable to a much worse one, which is restoring stale logs on top of usefully fresh ones (and I've seen it done), when you've gotten into the habit of restoring *.* from a backup directory on a noarchivelog test system, and survived, and then do much the same on the archivelog production system, and burn miserably as a result.

Maybe I'm simplistic, but I like wherever possible to have one set of guidelines which cover all eventualities... wherever possible. And the advice 'don't backup redo logs ever' works under all circumstances, whatever the characteristics of your database happens to be, and is safe.

But you obviously know all the issues, so that's fine.

~QM

>
> For a database in archivelogmode I agree: never backup the redologs
> because
> you probably do a roll forward and have the archives available for that
> purpose. And you can't be tempted to restore them, loosing the real ones
> and
> therefore loosing transactions.
>
> Altough a "best practice" would be to always make a savety copy of a
> control
> file and redolog files and if possible the entire database or involved
> files
> before beginning a restore/recovery.
>
>
>

-- 
Using M2, Opera's revolutionary e-mail client: http://www.opera.com/m2/