Re: DBMS_STATS gives wrong stats for some columns/indexes

From: Mladen Gogala <mladen_at_bogus.email.com>
Date: Wed, 20 May 2009 21:37:31 +0000 (UTC)
Message-ID: <gv1t6r$lg0$2_at_solani.org>



On Sun, 17 May 2009 17:49:59 -0700, ca111026 wrote:

> If we do that then result will be much close: 5% sample size - 80,000x
> 20 = 1,600,000 distinct values 10% sample size -149,000 x 10 = 1,490,000
> distinct values 20% sample size - 286,000 x 5 = 1,430,000 distinct
> values 50% sample size - 704,000 x 2 = 1,400,00 distinct values 100%
> sample size - 1,357,000 distinct values
>
> This was observed in Oracle 9.2.0.7 and 9.2.0.8 (Sun, AIX)

If you are suspecting DBMS_STATS bug, which is not impossible or unheard of, you can run VALIDATE_STATS package, as described in the ML note 337096.1. This is a published diagnostic package, you can use it freely.

-- 
http://mgogala.freehostia.com
Received on Wed May 20 2009 - 16:37:31 CDT

Original text of this message