On Thu, 4 Nov 2004 16:14:15 -0500, Hand, Michael T <handm_at_polaroid.com> wrote:
>
> I understand the change in extent size as segments get larger in LMT's with
> autoallocate, but read on a 3rd-party application web site that the smaller
> extent sizes (8M, 1M, 64K) are used if there isn't enough freespace for 64M
> extent. My experiments so far (using allocate extent and table insertion)
> have shown that if there is not 64M of freespace, the process errors out.
> Has anybody heard of this "feature", and under what conditions it manifests
> itself? Or is this unnamed 3rd party smoking something? The platform is
> Tru64 using 9.2.0.5.
Richard Foote and I had a discussion about this a while back. Richard
has a script that does use smaller extents when the larger size is not
available, I have a script (sounds similar to yours) that doesn't. We
could not conclude what the differences were. I'd suggest that it
might be worth constructing a test that dumped the extent allocation
bitmap after each allocate/deallocate of an extent to see what was
going on in the extent bitmap under this sort of activity, but to be
honest given that it is shown that this behaviour can fail with system
allocation, and we know that it can't with uniform allocation i don't
see any reason not to stick to uniform LMTs. I also suspect that my
test might be the start of a rather convoluted investigaion - and as
Mark states - the autoallocate behaviour is subject to change.
--
Niall Litchfield
Oracle DBA
http://www.niall.litchfield.dial.pipex.com
--
http://www.freelists.org/webpage/oracle-l
Received on Fri Nov 05 2004 - 06:26:49 CST
