Re: moving 10g execution plan to 11g using SPM

From: Carlos Sierra <carlos.sierra.usa_at_gmail.com>
Date: Tue, 4 Mar 2014 06:39:31 -0600
Message-Id: <342870BB-CCB5-4040-9406-69A9A592DB6F_at_gmail.com>



Different object ids should matter, since the hints that make the baseline do not refer to objects by id.

In any case, i suggest to focus on the root cause of the regression.

Cheers,

Carlos Sierra

blog: carlos-sierra.net
twitter: _at_csierra_usa

Life's Good!

On Mar 4, 2014, at 6:11 AM, Ls Cheng <exriscer_at_gmail.com> wrote:

> Hi Carlos
>
> I just realized that I might have different object mapping because I moved 10g database to 11g using TTS and all the object_id have changed. Can that be the reason?
>
> I will look into SQLT and get the the xtract output
>
> Thanks
>
>
>
> On Tue, Mar 4, 2014 at 1:09 PM, Carlos Sierra <carlos.sierra.usa_at_gmail.com> wrote:
> Ls,
>
> SPM is not very verbose when it comes to not reproducing a plan. Where I can help you with is identifying the reason of the regressions you are observing. What I would need is between 1 and 3 of such regressions, and have SQLT XTRACT (MOS 215187.1) executed for each of those SQL_IDs in both the 10.2.0.5 and the 11.2.0.4 environments.
>
> Cheers,
>
> Carlos Sierra
>
> blog: carlos-sierra.net
> twitter: _at_csierra_usa
>
> Life's Good!
>
> On Mar 4, 2014, at 6:00 AM, Ls Cheng <exriscer_at_gmail.com> wrote:
>
> > Hi all
> >
> > I am currently upgrading a few databases from 10.2.0.5 to 11.2.0.4 and running SQL regression tests using SQL Performance Analyzer, I have detected a few queries that is not performing well in 11g so I transported those queries execution plan and imported as sql plan baselines in 11g. The strange thing is that these base lines are not used, after running SPM tracing I see that it cannot reproduce the execution plan (SPM: failed to reproduce the plan using the following info: is shown in the trace files) but I cannot see the reasoning. The possibilities I can think of is schema differences between 10g and 11g, object mapping is different or outline errors but these dont seem apply or at least not observed in SPM trace and 10053 trace.
> >
> > Does anyone know what other potential reasons can cause such behaviour?
> >
> >
> > Thanks
> >
>
>

--
http://www.freelists.org/webpage/oracle-l
Received on Tue Mar 04 2014 - 13:39:31 CET

Original text of this message