Re: alter rollback segment

"Paul Drake" <paled_at_home.com> wrote in message news:3AF42AA1.68555258_at_home.com...
> "Howard J. Rogers" wrote: (some content removed)
> >
> > Hi Paul,
> >
> > No real gotchas that I can think of, except that you have to use the
> > 'uniform size' version, not the autoallocate (anecdotally, anyway -I've
> > never bothered wth autoallocate in the first place, so I've never
 tried).
> >
> > There is only one nasty with LM Rollback tablespace: you can't create
 the
> > first rollback segment in it unless there is one non-system dictionary
> > managed rollback segment (because creating that first one requires you
 to do
> > DML on the bitmap at the header of the tablespace -and that's DML on a
> > non-system tablespace, and hence if all you've got is the system
 rollback
> > segment, you'll get that infamous ORA error to the effect that 'you
 can't
> > use system rollback segment for updates in non-system tablespace'). So,
 you
> > need to create one small rollback segment first in boring old dictionary
> > tablespace (the SYSTEM tablespace will do fine -the fact that a rollback
> > segment is housed in SYSTEM doesn't make it the system rollback
 segment).
> > Once that is brought online, *then* you can create your first rollback
> > segment in a locally managed tablespace. And once *that's* created and
> > brought online, you can drop the temporary, dictionary-managed one,
 because
> > now you *do* have a non-system rollback segment to handle all subsequent
> > updates to the tablespace's bitmap.
> >
> > Regards
> > HJR
>
> Howard,
>
> Thanks much for the detailed info.
> I upgraded the DEV server last night from 8.1.6.3.4 to 8.1.7.1.2 (on NT)
> - so once my exports and backups are finished, I'll give it a go.
>
> In seeing a demo of LogMiner - I'd like to reduce as much as possible
> the recursive SQL to uet$ and fet$ (and others). It would be nice to
> have SYSTEM as the only non-LMT.
>

Absolutely, and that's one biggie plus in favour of LMT, of course. Good luck. Let me know how it goes (or if it doesn't!)

Regards
HJR
> thanks again.
>
> Paul
Received on Sat May 05 2001 - 12:13:22 CDT