Re: rollback segment question...

gdas_at_my-deja.com wrote:

> Can anyone explain to me what could cause a rollback segment to
> automatically go offline? (v. 8.1.5)
>
> I was looking through an application log and saw the oracle error:
> unable to extend rollback segment rb28 by 32.
>
> We were doing a stress test and we artificially, throttled up the
> processing load of the application. At that point I started noticing
> these errors written to our log file.
>
> So I think the root source that caused the problem is simply that the
> space allocated to the rollback segments couldn't handle the number of
> current transactions we were hitting the database with.
>
> But my bigger concern is that once we stopped the test and I went to go
> check on the rollback segments...5 of them were no longer online,
> included rb28. The system is set to acquire these at start up, and
> given the error message above, it does indicate that at one time rb28
> was online...but somehow, something caused it to go offline.
>
> Can anyone explain what might have caused this? If rollback segments
> run out of space, do they automatically go offline?
>
> Thanks
> Gavin
>
> Sent via Deja.com
> http://www.deja.com/

Are these "public" rollback segments. If not they probably should be unless you are set up in some type of parallel server configuraiton.

Secondly, check the Alert log to see if there are any error messages there. Also, check the bdump, udump and cdump directories for trace files that might have pertinant entries. Received on Wed Dec 20 2000 - 07:49:19 CST