Re: shrinking rbs?

johnt_at_tman.dnsalias.com wrote:
>
> Doug O'Leary <dkoleary_at_ro05-24-29-232-217.ce.mediaone.net> wrote:
> > Thomas Kyte <tkyte_at_us.oracle.com> wrote:
 

> >> The problem is a query doesn't "USE" rbs. It might need the contents of
> >> rbs at some point to provide a read consistent view of the data -- but
> >> it is not using rbs. A "write" transaction will use RBS and if it was
> >> using it -- we would not shrink it.
 

> > That's the concept that I was missing - and explains the risk. Thanks
>
> But if there is no 'task' using the rbs, wouldn't a read-consistent view
> already be in the table itself? Therefore, by shrinking a RBS I can't
> see how you can cause a problem with any query.

To achieve the read consistent look, the rollback segments need to be visited - not necessarily at query start time but as it wanders through the blocks on the table. If sometime after your query starts and it finds a block that warrants leaping out to the rollback segs to reconstruct it (and then finds that roll seg extent gone because someone has shrunk it), then you get the error.

hth
connor

-- 
===========================================
Connor McDonald
http://www.oracledba.co.uk (mirrored at
http://www.oradba.freeserve.co.uk)

"Some days you're the pigeon, some days you're the statue"