Re: Things to consider during upgrade/migration

Yes Mark, so my point was, we already have ~9 sqls which are using baseline in current 11.2. version and of course
have optimizer_use_sql_plan_baselines set as TRUE(which is default). But Now that we will be setting optimizer_capture_sql_plan_baselines to TRUE a few days before 19C upgrade on production to get hold of all the possible sql and their plans(say we captured ~1000 sqls baseline), but we don't want all of those to be enabled at that same moment. rather we want only to enable the baseline for those ~9 old sqls + additional few sql (say e.g. 3 sqls ) which misbehaved/changed plan post upgrade to 19c. And we want to drop/get rid of those other ~997 sql baselines from our 19C version database (which we captured initially just for our safety). So considering its inplace upgrade and we don't have an exact test environment like RAT, is there a way to achieve this in the existing production database? I am still unable to get it clearly.

On Wed, Nov 17, 2021 at 10:29 PM Mark W. Farnham <mwf_at_rsiz.com> wrote:

> I thought the point was he has enabled the use of all the baselines he
> currently has en masse by a default parameter, so he wants to be able to
> get back to just those being stored.
>
>
>
> Did I misread something?
>
>
>
> *From:* Andy Sayer [mailto:andysayer_at_gmail.com]
> *Sent:* Monday, November 15, 2021 5:29 PM
> *To:* mwf_at_rsiz.com
> *Cc:* Clay Jackson (cjackson); Oracle L; loknath.73_at_gmail.com
> *Subject:* Re: Things to consider during upgrade/migration
>
>
>
> “ As i had posted earlier, I am still wondering how to isolately enable
> the sql plan baselines to only specific sqls manually rather than forcing
> it to be enabled for all the captured baselines? ”
>
> Capture ALL baselines (either with the parameter or just loop through your
> application sql_ids and the dbms_spm function). Keep using baselines with
> your existing parameter. You can turn off the capture just before you
> upgrade if you really wanted (I leave it permanently true).
>
>
>
> No need to do anything special with your existing baselines, just keep
> them as they are (nothing will happen to them unless you manually disable
> them).
>
>
>
> Thanks,
>
> Andy
>
>
>
> On Mon, 15 Nov 2021 at 21:45, Mark W. Farnham <mwf_at_rsiz.com> wrote:
>
> Or just export the current contents and use your favorite method to
> restore it as the contents later…
>
>
>
> *From:* oracle-l-bounce_at_freelists.org [mailto:
> oracle-l-bounce_at_freelists.org] *On Behalf Of *Lok P
>
>
> *Sent:* Monday, November 15, 2021 2:53 PM
> *To:* Mark W. Farnham
> *Cc:* Clay Jackson (cjackson); Oracle L
> *Subject:* Re: Things to consider during upgrade/migration
>
>
>
> As i had posted earlier, I am still wondering how to isolately enable the
> sql plan baselines to only specific sqls manually rather than forcing it to
> be enabled for all the captured baselines?
>
> Note- We have currently optimizer_use_sql_plan_baseline set as default
> true which is helping few of the existing sqls already. So we can not turn
> it false. And keeping it True is again going to make all the
> captured baselines to be enabled, so how should we deal with this?
>
>
>
> On Mon, Nov 15, 2021 at 7:22 PM Mark W. Farnham <mwf_at_rsiz.com> wrote:
>
> AND, (not but at all),
>
>
>
> Don’t be hesitant to discard a few false starts. Make sure your timeline
> includes time for a few fresh starts.
>
>
>
> IF you have a way to simulate the activity including transactions,
> standard reports, and most common ad hoc reports for a five month period
> of December, January, February, March, and April for a January 1st year
> end, that will cover a year end and a quarter end, which tend to be the
> most important. IF special things happen at year and quarter ends and you
> don’t do this, you are in extreme violation of #3 below.
>
>
>
> IF you have time period comparison reports for, say, last year compared
> to this year and the time to shove all data conflicts with the desire for
> multiple trial conversions and dress rehearsals make sure you get enough
> months to make the comparison reports realistic.
>
>
>
> IF you have a lot of moribund data (data that is no longer allowed to be
> changed) consider physically re-ordering, partitioning, and making such
> data isolated in read only spaces as part of the migration. As needed for
> retrograde plan or iso-plan performance outliers, extra specific indexes
> (and/or best cluster factor and attribute ordering to match queries of
> existing indexes) may be extremely useful to improve performance. The extra
> time to convert is often time well spent in both size and speed.
>
>
>
> IF you can simulate as suggested, that may tempt you to go light on
> suggestion #2 below, ignoring the things that seem fine and just snapping
> the baselines that have problems in your “test migration” when everything
> on the destination is fresh and new. Don’t skip suggestion #2 unless you
> plan to have the old machine available for years. Transactions can change
> the texture of the data over time such that plan changes erupt differently
> in the new environment as time passes. IF the old machine is gone, all
> those baselines are your time machine to keep life BORING. BORING is good
> in IT operations. The excitement should be innovative use of well running
> technology, not firefighting to keep the technology from burning up.
>
>
>
> Good luck. JL also has a blog thread of things that can go bump in the
> night when you “didn’t change anything” that is a useful read to help
> THINKING CLEARLY about things that erupt during your several dress
> rehearsals.
>
>
>
> And remember: Two complete dress rehearsals with NO CHANGES before you
> pull the trigger on the real migration. (Farnham’s Law of Sane Migrations,
> often violated at great pain).
>
>
>
> *From:* oracle-l-bounce_at_freelists.org [mailto:
> oracle-l-bounce_at_freelists.org] *On Behalf Of *Clay Jackson
> ("Clay.Jackson")
> *Sent:* Saturday, November 13, 2021 4:36 PM
> *To:* loknath.73_at_gmail.com; Oracle L
> *Subject:* RE: Things to consider during upgrade/migration
>
>
>
> What Mladen and Andy said – I would pay very serious attention to their
> wisdom; born of many years 😊. ESPECIALLY the parts about
>
> 1. How different 19 is from 11 and the Exadata is from the HP
> 2. Collecting baselines for ALL SQL – the time you spend up from will
> save more than that troubleshooting on the “back end”
> 3. TEST EVERYTHING
>
>
>
> In my position, I talk to customers/prospects going through similar
> upgrades/migrations and those who follow(ed) Steps 2 and 3 (and understood
> the differences) had SIGNIFICANTLY better success than those who tried to
> “shortcut” the process.
>
>
>
> Clay Jackson
>
>
>
>
>
> *From:* oracle-l-bounce_at_freelists.org <oracle-l-bounce_at_freelists.org> *On
> Behalf Of *Lok P
> *Sent:* Saturday, November 13, 2021 10:47 AM
> *To:* Oracle L <oracle-l_at_freelists.org>
> *Subject:* Things to consider during upgrade/migration
>
>
>
> *CAUTION:* This email originated from outside of the organization. Do not
> follow guidance, click links, or open attachments unless you recognize the
> sender and know the content is safe.
>
>
>
> Hello Listers, With respect to having a safe upgrade(say from 11.2 to 19C)
> or migration(From HP to Exadata) experience with minimal performance
> issues. Is there any guideline we should follow like setting up exadata
> system stats in case the target database is going to be exadata, Or
> verifying dictionary stats/table stats etc in a certain way. Want to know
> experts' views, if there are any such guidelines?
>
> Regards
>
> Lok
>
>

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