Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Usenet -> c.d.o.server -> Re: Archivelog Mode Recoverability: How Close to Point of Failure?

Re: Archivelog Mode Recoverability: How Close to Point of Failure?

From: Howard J. Rogers <hjr_at_dizwell.com>
Date: Fri, 13 Feb 2004 18:29:31 +1100
Message-ID: <402c7cdd$0$19705$afc38c87@news.optusnet.com.au>

"Skinny Guy" <skinnychineseguy_at_hotmail.com> wrote in message news:f9b5a61d.0402122306.4b08bbc2_at_posting.google.com...
> Ok, I understand. So in the case of media failure, I would restore all
> files from tape to disk (except redo log files)

Absolutely not. In the case of media failure, you restore only those files which have actually failed. You don't restore the control files (unless they've gone walkabout), and you don't restore ALL datafiles, unless you've lost all of them. Just the file(s) which is actually lost.

>, hopefully have a

> multiplexed redo log files on a separate drive that didnt' crash,
> hopefully have archive logs on a separate drive that didn't crash, and
> Oracle will be able to automatically use both online redos and the
> archive log files to fully recover?

Well, not "automatically", in the sense that having restored the problem data file, you won't be able to open the database. You therefore have to say 'startup mount' followed by 'recover automatic datafile X', where X is the number of the problem file. Only at that point will Oracle automatically apply all the redo, from both archives and online logs. But yes, it will work out where it can get the required redo from automatically, and apply it regardless of its source. After recovery is complete, you just say 'alter database open'. There are variations on this theme, but that's essentially it: restore damaged file, startup mount, recover damaged file, alter open.

> My problem is we only have a single raid device which has several
> logical drives, so I can't really multiplex onto a different drive.
> Would having the RAID be good enough, so I won't have to multiplex the
> redo and archive files?

Absolutely not. It's better than nothing, but RAID only protects you against hardware failure. It does nothing to protect you against software corruption or user error ("let me just practice my rm* skills..."). You need both, if you care about not losing data that is.

> My IT server guy says RAID is good enough so it'll be hard for me to
> push to use another drive.

Your IT server guy doesn't understand Oracle databases then. RAID is a hardware failure tolerance mechanism. It doesn't protect you under all circumstances (for example, pity the poor DBA who called all his online logs "something.log" who then had an "IT Server guy" run a clean-up script which deleted *.log -they are disposable text files after all... er, you mean that wasn't a text file?... er, you mean it wasn't disposable... etc etc etc. It's happened.).

As far as persuasion goes, cost it up. How valuable is your data? How much would it cost to re-enter X hours or minutes of data? How costly would it be if you couldn't actually re-enter it reliably at all? How much is your time worth? How much does it cost not to have the database available for a couple of hours? Because without the online logs, you're going to have to do an incomplete recovery, which is lengthy, requires down-time, and can easily be botched (and hence require repetition... yet more down time). Tot it all up, and compare with the price of another disk or another RAID array. If the database is a production one of any merit whatsoever, then I know where I'd place my bets on where the balance of costs v. advantages comes out.

> But doesnt RAID make a lot of the oracle
> multiplexing and other designs unimportant?

Absolutely not (sorry, getting repetitive I realise). RAID is a hardware issue. Multiplexing is a software and user issue. You *need* both.

Regards
HJR

-- 
--------------------------------------------
Oracle Insights: www.dizwell.com
--------------------------------------------
Received on Fri Feb 13 2004 - 01:29:31 CST

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US