Re: changing rollbacks

One again:

John Higgins wrote:

> Oracle never does any space management excepr when necessary!
>
> When you dropped the segments, Oracle changed the allocated extents into
> 'free extents'. Oracle dod not coalesce all the free extents in the (now
> empty) tablespace into one single free space extent.
>
> Then, when you re-created the segments at their new, bigger extent sizes,
> Oracle proceeds as follows:
> 1) Is there a free space extent of extent of exactly the requested size? If
> so, reuse it.
> 2) Is there a free space extent larger than the requested size? If so, break
> off the requested size (if the stub would be less than 5 blocks, just use it
> all)
> 3) If there are any 'contiguous' free spaces => than the requested size,
> coalesce them and go to 2.
>
> So, Oracle bypassed your previously used extents and created new extents
> from above the old HWM.
>
> You could have, after dropping the old segments, issueed a 'ALTER TABLESPACE
> ____ COALESCE;' to coalesce all the free space into one big extent and the
> re-created your segments.
>
> Doug Cowles wrote:
>
> > I could be just tired - but somthing's bugging me. I have three
> > rollbacks,
> > intitially sized at 70M on datafiles of 1.5 GB each. One tablespace per
> >
> > datafile, and one rollback segment for each tablespace. Nothing else in
> >
> > the tablespaces but their respective rollbacks.
> >
> > I wanted to change the intial size of the three rollbacks to 300M, and
> > then
> > shrink the datafiles to 600M. I dropped the three rollbacks, (which I
> > thought
> > would clear out the datafiles completely), and then recreated them with
> > the new
> > storage parameters. When I try to shrink the datafiles to 600M, it
> > tells me
> > , ORA-03297: file contains 641 blocks of data beyond the requested
> > RESIZE value.
> >
> > Now, it's Friday, and I'm leaving this one, but instinct tells me if I
> > re-create the
> > tablespace, everything will be dandy.. but how can this happen? If I've
> > dropped any
> > segments that are in that tablespace, corresponding to a rollback
> > segment whose
> > high water mark never exceeded 300M, why would there be data or
> > allocated extents
> > floating around the 600M mark??
> >
> > Thanks in advance,
> > Dc.
Received on Tue Jun 01 1999 - 16:31:31 CDT