Re: Reporting an optimizer bug

From: Lothar Flatz <l.flatz_at_bluewin.ch>
Date: Mon, 24 Jan 2022 09:33:27 +0100
Message-ID: <bd9202da-a533-b0e2-c5a7-4438b06935b3_at_bluewin.ch>



HI,

If I were still member of RWPG I could help you right away. First of all: yes, a testcase would be really helpfull. You can generate a testcase with artificial data that does mimic the original case in both volume and value distribution. Jonathan Lewis does this quite often and you can copy his methods. Basically you write a row generator using e.g. connect by rownum <= n; I admit it is not as easy. You have to have an idea which estimate goes wrong and why.
An other way would be to export the stats for the tables referred in the query.
Second:
Escalate your SR. Quite often you have fight your way up to an engineer understanding those kind of tasks.

Thanks

Lothar
Am 24.01.2022 um 08:23 schrieb Nilo Segura:
> Hello,
> We have found what we believe is a bug in the optimizer, and we are
> trying to report this problem to Oracle with no success at all. We
> have found two workarounds that generate a very good plan. The problem
> is with one query executed by the EM 13c (related to tablespace
> monitoring). These are SYS related objects, so it is not possible to
> get all the info with dbms_sqldiag (testcase), even though support
> told us precisely to do that :(
>
> So my question is, for those of you have succeeded in reporting these
> issues to Oracle.. What is the magic trick ? We are hitting a wall
> here and it is extremely frustrating, looks like they do not want at
> all to open a bug report. And yes, we have escalated the SR with no
> positive result either.
>
> Best regards.
> --
> Nilo Segura
> Oracle Support - IT/DB
> CERN - Geneva
> Switzerland

--
http://www.freelists.org/webpage/oracle-l
Received on Mon Jan 24 2022 - 09:33:27 CET

Original text of this message