Re: Wrong results using decode when db upgraded to 9205

From: Daniel Fink <Daniel.Fink_at_Sun.COM>
Date: Wed, 05 May 2004 12:07:49 -0600
Message-id: <40992D75.4F7CBAC4@sun.com>

In my testing, I've found the RULE hint still works. If I remove the RULE hint from the previous statement, it uses hash joins. With the hint, it uses nested loop joins. Also, if I rerun the query slightly altered (to get a new hard parse) with 10053, it shows that the CBO is not used, no calculations are made, no join orders are attempted. This tells me that the RBO engine is being used.

The old rules about RBO and CBO still apply.

Regards,
Daniel

Mladen Gogala wrote:
> Well, it is alive, but not well. If you don't change defaults, 10g does
> dynamic sampling all by itself and you will always get some statistics.
> which means that you will always get CBO, unless you reset dynamic sampling
> and disable statistics collection with dbms_scheduler.disable('GATHER_STATS_JOB')
> DBMS_SCHEDULER is DBMS_JOB on steroids. Also, you cannot set optimizer_mode to
> rule in init.ora (OK, OK, it's spfile.ora), there is no /*+ RULE */ hint and
> oracle did everything to make the use of RBO as inconvenient as possible.
> Also, there is no documentation, so I don't know whether RBO will prune
> partitions, do hash joins or perform an index skip-search (OK, I know that RBO
> will not do thiat, but not from 10g documentation). Fortunately, the event
> 10053 still works as it is supposed to, so we can take a peek under the hood.
>
> --
> Mladen Gogala
> Oracle DBA
>

Please see the official ORACLE-L FAQ: http://www.orafaq.com

To unsubscribe send email to: oracle-l-request_at_freelists.org put 'unsubscribe' in the subject line.

--
Archives are at http://www.freelists.org/archives/oracle-l/
FAQ is at http://www.freelists.org/help/fom-serve/cache/1.html
-----------------------------------------------------------------

Received on Wed May 05 2004 - 13:06:31 CDT