Re: rollback segment too small

If you monitor V$ROLLSTAT.WRAPS you will see this count increasing over time. If your rollback segments have 20 extents, then your before images are preserved for 20 wraps in each rollback segment. If you have 6 rollback segments, your before images are preserved for 120 total wraps. If you record the minimum time between over-writes (120 wraps in this case), you will know the maximum query time that is "safe".

If you increase the size (or the number) of the rollback segments, you will see that the minimum time between over-writes increases, thus lengthening the maximum "safe" query time.

We are into queueing theory here -- doubling the size (or number) of the rollback segments may not double the minimum time between over-writes! So it boils down to econonomics -- you may not be able to afford enough rollback disk to guarantee a *long* query during peak updating

Glenn Baron wrote:

> Essentially , you are always likely to encounter this when running *long*
> queries online against a volatile database.
Received on Sat Apr 24 1999 - 10:50:29 CDT