Comments on: Plan stability through Upgrade to 11G-Introduction

By: Plan stability through Upgrade to 11G- Building a test case « Coskan’s Approach to Oracle

Mon, 14 Feb 2011 12:05:51 +0000

[...] <-Previous Post in Series Next Post in Series–> [...]

By: Plan Stability Through Upgrade-Why is my plan changed?-bugfixes-1 « Coskan’s Approach to Oracle

Mon, 14 Feb 2011 12:04:22 +0000

[...] on 11.2.0.2 performance stor…coskan on Session Snapper on dutyorawiss on Plan stability through Upgrade…ittichai chammavanij… on 11.2.0.2 performance stor…Coskan Gundogar on 11.2.0.2 [...]

By: orawiss

orawiss — Fri, 28 Jan 2011 22:46:09 +0000

Really Nice project, I will try to be updated with your articles.

Thank you,
Wissem
http://www.oracle-class.com

By: coskan

coskan — Thu, 27 Jan 2011 09:54:55 +0000

14 over 1800 unique sqlid and 10 of them fixed 4 of them not fixed. Fixed ones are higher cost ones thats why they are fixed. Even It can bee seen like we did not used baselines heavily, it worked for us. We found 11.2.0.1 fairly good for plan generation apart from the cost plan changed caused by extra nested loop addition without changing the cost we had probably 359 plan change and very small amount of plan change for bad.plus cardinality feedback related instability on plans

By: Dom Brooks

Dom Brooks — Thu, 27 Jan 2011 09:34:05 +0000

How much of your SQL is protected by baselines?

By: Marko Sutic

Marko Sutic — Thu, 27 Jan 2011 09:03:48 +0000

Great topics Coskan and thanks for your willingness to share that kind of information.

We have several 9i->11g, 10g->11g migrations awaiting and your future posts will be very valuable to me.

Enjoy,
Marko

By: Damir Vadas

Damir Vadas — Wed, 26 Jan 2011 22:49:28 +0000

Nice.
If you do not have client who bought Intel Itanium CPU and end the story on the first step. Bye bye 11g!
8(

By: coskan

coskan — Wed, 26 Jan 2011 09:32:37 +0000

I wish I had them all ready but I can tell that I would do

a-) I would capture the plans just to be on safe side and if they cause problem on 11G then I would fix their plans
b-) this needs more testing. advised by Maria Colgan from Oracle but can have wider effects. To narrow it down maybe I would do logon trigger for services users using the sql or capture their plans during this parameter on and again fix those plans at least I feel more secure. optimizer_ignore_hints is good but developers are very keen on their hints and their knowledge and if something blows then you will be the one on fire. hard choice

By: Dom Brooks

Dom Brooks — Wed, 26 Jan 2011 09:22:19 +0000

Hi Coskan,

Sounds like a great series.

Trouble is I want to read it now

I’m currently waiting to shepherd an application through a migration from 9i to 11gR2.

One of the things I’m pushing heavily (and encountering FUD resistance from management) is SQL Plan baselines to a) capture some of the 9i plans (optimizer_features_enable) and b) to ignore those single use_nl hint in those 8 table join SQL statements using optimizer_ignore_hints.

I’ve got a meeting about all this in ten minutes so if you could get all the articles done by then, I’d appreciate it

Cheers,
Dominic