Re: Do I really need more than 1 rollback segment?

We actually reported that bug. I did start a thread on it here if anyone is interested in reading it again, was around the end of Aug or so.

On Jan 7th, we will be attempting the upgrade to 8.1.7 again on a 'spare' production database that is being discarded at the end of the month. I will repost with the results of that to see if the patch actually works.

We have the patch on our development machines, but we were never able to fully reproduce the problem. We were able to show the same symptoms but never to the extent as our production instances.

Scott Gamble

Andrew Mobbs <andrewm_at_chiark.greenend.org.uk> wrote in <s2E*1cicp_at_news.chiark.greenend.org.uk>:

>Jonathan Lewis <jonathan_at_jlcomp.demon.co.uk> wrote:
>>
>>Re moving from 8.0.6 to 8.1.7
>>
>>Watch out for bug 1967363 - relating to the
>>cache buffers chains latches for index root
>>blocks becoming very hot - initial symptom
>>is a dramatic increase in waits for 'latch free'
>>which can be identified as sleeps on the
>>'cache buffers chains' latch' and then
>>tracked to x$bh.hladdr via v$latch_children.addr
>>to blocks which are the second block of index
>>segments.
>
>Yikes! Thanks for pointing that one out.
>
>I hope that's the cause of the cache buffers chains latch waits that
>have been annoying me for the past year. Certainly everything fits, I've
>already traced it down to the root block of an index. I'm not seeing
>as bad an impact as that reported in the bug though.
>
Received on Thu Dec 20 2001 - 13:07:36 CST