Re: How to clear Rollback segments

From: Howard J. Rogers <dba_at_hjrdba.com>
Date: Wed, 5 Dec 2001 06:20:40 +1100
Message-ID: <3c0d21f2$0$27196$afc38c87_at_news.optusnet.com.au>


I entirely agree, and the very strong recommendation on my website is that OPTIMAL is the last resort of scoundrels and wastrels, and has no place in a properly managed database (well, not *quite* put in those words, but the sentiment's the same).

I was merely pointing out that, *as posted*, the advice to simply 'shrink' won't work -it needs a size component, unless optimal has indeed been set. That was never an endorsement of optimal!

Regards
HJR

--
Resources for Oracle: http://www.hjrdba.com
===============================


"Mladen Gogala" <mgogala_at_earthlink.net> wrote in message
news:pan.2001.12.04.08.19.00.215602.898_at_earthlink.net...

> On Tue, 04 Dec 2001 03:03:43 -0500, Howard J. Rogers wrote:
>
> > That command will only work if OPTIMAL has been set (and if OPTIMAL
*has*
> > been set, it's a bit unlikely he'd need to manually shrink them in the
> > first place), otherwise you have to specify a size to shrink *to*.
> And having "OPTIMAL" set means that your rollback segments will be shrunk
> from time to time. That means that there are transactions which will
> extend rollback segments beyond optimal and that SMON will shrink them
> back. Now that means that some lucky transactions will have to deal with
> the extents allocation in addition to the normal rollback space allocation
> and on the periodical basis. This is the introduction of a lottery system
> which will make some transactions 10 times slower then the other. If you
> are lucky, you will have some "snapshot too old" beauties to worry about.
> What I am trying to say is: do not set the OPTIMA parameter. Let you
> rollbsck segments be fruitful and multiply because they're doing so for a
> reason.
> --
> Mladen Gogala
Received on Tue Dec 04 2001 - 20:20:40 CET

Original text of this message