Re: [EXTERNAL] Question on sql plan management

Hello,

Your 12c SPM execution plans will be used only if they are reproducible in 19c. Otherwise, Oracle will use the 19c plan it produces at hard parse time.

So, in 19c, you will be in one of the two following situations:

1. CBO comes up with a new 19c plan present in the SPM baseline
   1. You will use the 19c CBO (or the 12c SPM plan)
2. CBO comes up with a new 19c plan not present in the SPM baseline
3. Your 12c SPM baseline plans are reproducible: you will use the SPM plan
4. Your 12c SPM baseline plans are not reproducible and you will use the 19c plan

> Thank You Jonathan.
> Actually we were not having an environment which is close to production in
> terms of data volume and data pattern and infrastructure. And currently we
> have turned on the "capture baseline" flag to TRUE in production itself
> which is on 12.1 version and we already have hundreds of thousands of sql
> baselines getting captured in production on a daily basis.And i noticed the
> "use baseline" flag is already set as default TRUE. And we are planning to
> set the capture baseline flash now to FALSE as we have all types of
> workloads covered till now. In this situation, how should we be able to
> make use of 19C features or plans on top of the existing baselines post
> migration?
>
> On Sat, Sep 3, 2022 at 8:49 PM Jonathan Lewis <jlewisoracle_at_gmail.com>
> wrote:
>
>>
>> One fairly standard approach is to test on a full-size backup copy of
>> production with no SQL baselines to find the statements that show much
>> worse performance in 19c, then go back to the 11g versions and capture
>> plans for only those statements before the live upgrade.
>>
>>
>> Regards
>> Jonathan Lewis
>>
>>
>> On Sat, 3 Sept 2022 at 14:21, yudhi s <learnerdatabase99_at_gmail.com>
>> wrote:
>>
>>> Thank you All.
>>> I understand for some of the specific scenarios like fixing wrong
>>> results related to bugs from some transformation etc, it may happen that a
>>> certain plan is not reproducible exactly on 19c which was there in 11.2
>>> /12.1 versions. In these cases in the19c version, it will come up with a
>>> new plan which will eventually not match with any of the accepted old
>>> baselines and thus the optimizer will forcibly choose a different path
>>> altogether and create that as a new accepted baseline.
>>>
>>> However, doesn't it sound odd that in cases where, of all those hundreds
>>> of thousands of sqls using one of the captured baselines from 12.1 by auto
>>> capture process which are now in an accepted state from the 12.1 version,
>>> has to be manually evolved and accepted to use the 19C feature or new
>>> plans? Considering enabling auto evolve jobs as riskier to go for. So I
>>> wanted to understand what is the standard or lesser riskier suggested way
>>> of doing this when moving to the 19c version?
>>>
>>>>

-- 

Houri Mohamed

Oracle DBA-Developer-Performance & Tuning

Visit My         - Blog <http://www.hourim.wordpress.com/>

Let's Connect -
<http://fr.linkedin.com/pub/mohamed-houri/11/329/857/>*Linkedin
Profile <http://fr.linkedin.com/pub/mohamed-houri/11/329/857/>*

My Twitter <https://twitter.com/MohamedHouri>      - MohamedHouri
<https://twitter.com/MohamedHouri>

--
http://www.freelists.org/webpage/oracle-l