Re: Weird behaviour on a RAC

From: -={Giorgio}=- <grgs27_at_gmail.com>
Date: Tue, 17 Jan 2006 10:47:57 +0100
Message-ID: <dqiegd$316j$1@newsreader1.mclink.it>

Sybrand Bakker wrote:
> On 16 Jan 2006 08:04:38 -0800, grgs27_at_gmail.com wrote:
>
>

>>The only things I can say are:
>>- It seems that wverything started after a bounce of the instances
>>(but no parameters were changed whatsoever)

>
>
> buffer cache has been cleared

Yes AFAIK this might impact the performance of a query not its cost... and this, however, does not explain why the *very same* query explained 4 times in sql*plus gives you cost 7 and 3 seconds after cost 15k and the 4th time cost 7 again.

>

>>- Statistics has been collected correctly

>
> Please define 'correctly'

with package and not analyze if you mean that... parameters i don't have at hand now but i can post. But no histograms and sampling is around 30%.

g

>

>>- It seems that with problem #1 changing on table involved from IOT to
>>Heap the problem disappears.
>>- "Normal" activity in global cache

>
> --
> Sybrand Bakker, Senior Oracle DBA
Received on Tue Jan 17 2006 - 03:47:57 CST