Re: Do I really need more than 1 rollback segment?

Sorry guys, I think I've disgraced myself.... :-)

What I meant by "as small as you can make them" was pertaining to my taking an outer limit on rollback segment sizes and creating the segments to fit that. I arrived at these limits during testing in the, umm, test databases and then monitoring having re-created RBS in LMT. (Nuno, you may remember my initial statspack report where my inherited databases had 7 segs of 350m apiece in 2 extents).

Anyway, now, in my two production databases, one has 4m extents and the other has 2m extents (minextents 20). Optimal size is 50m. I didn't mean create them as small as the database would allow though reading it again that sure sounds like my implication.... ;-)

They rarely extend (like maybe *one segment* will extend by one extent) on a Saturday night, but that's it. They *never* extend during daily processing. Ever.

I don't see a problem with using optimal given that my segments are sized appropriately to safeguard against them extending (and thus requiring shrinking) in the first place. Sure, if they were extending and shrinking throughout the day, that would be mad. Hopefully, this sounds a bit more sane!

Interested in responses as always.

Andrew

"Howard J. Rogers" <dba_at_hjrdba.com> wrote in message news:<3c110a2a$0$29051$afc38c87_at_news.optusnet.com.au>...
> I agree with your disagree! Setting rollback segments 'as small as you can
> make them' is just plain daft, and setting optimal at all is equally mad.
> Optimal is there for those databases that don't have a full-time DBA, and as
> such it can be a useful 'light-weight management' tool. But in a
> properly-managed production database, it's an utterly lousy idea.
>
> Regards
> HJR
> --
> Resources for Oracle: http://www.hjrdba.com
> ===============================
Received on Tue Dec 11 2001 - 06:16:00 CST