Oracle-L: RE: Changed execution plans..

From: Robert Freeman <robertgfreeman_at_yahoo.com>
Date: Sun, 19 Jan 2003 20:28:37 -0800
Message-ID: <F001.00534247.20030119202837@fatcity.com>

Thanks Cary... I've actually considered most of those already. This isn't my database, I'm coming in to help someone else. In this case, the database schema was accidentally dropped, and recovered from an export. I'm wondering if the import might have created blocks that are packed more densely and that this might be the cause of the problem. I'm told the parameters are the same, that the physical structure is the same, they are not using stored outlines and the SQL has not been changed. No patches have been applied, so it's apples for apples with the exception of the statistics and, possibly, the data density. They have some old stored statistics that they generated pre-schema drop that they are supposed to send me, so I'm going to look at that tomorrow and run a 10053 trace on one of the changed queries and see what I can find. I was just wondering if I could be missing something obvious..... Seems like that is just the way, it's the obvious things that get missed... :-)

Thanks so much for your comments!

-----Original Message-----

Millsap
Sent: Sunday, January 19, 2003 8:54 PM
To: Multiple recipients of list ORACLE-L

Robert,

Seven reasons I can think of include changes to:

Oracle instance parameter values (changes when you edit the parameters, whether in the stored "init.ora" way, or via ALTER SYSTEM or ALTER SESSION commands)
Database table and index statistics (changes, e.g., when you run dbms_stats.gather_database_stats)
System CPU and I/O statistics (changes, e.g., when you run dbms_stats.gather_system_stats)
Database schema configuration (changes when you create/drop indexes, etc.)
Stored outlines (changes when you create or reassign outlines)
SQL text (changes when you manipulate the application SQL)
Oracle query cost model (changes when you upgrade or patch your Oracle kernel)

Cary Millsap
Hotsos Enterprises, Ltd.
http://www.hotsos.com

Upcoming events:

- 2003 Hotsos Symposium, Feb 9-12 Dallas
- RMOUG Training Days 2003, Mar 5-6 Denver
- Hotsos Clinic 101, Mar 26-28 London

-----Original Message-----

Freeman
Sent: Sunday, January 19, 2003 1:24 AM
To: Multiple recipients of list ORACLE-L

Hey Ya'all... (still got my southern roots even up here in Chicago!)

Anyone want to throw in some possible reasons why an execution plan might
change for a given table query....This is on Oracle9iR2 on SUN. I've looked
at the obvious causes:

Object has changed - Appears not to have changed.
Database parameters have changed - Appears that no parameters have changed.
Statistics (data volumes, distribution, cardinality, etc) have changed - Still looking into this, but the volumes have not changed dramatically even if they have changed.
Other physical database changes.

None of these seem to apply. I've got a database that a few weeks ago were
doing indexed
lookups using a partitioned index on a partitioned table. Now, it seems that
these queries
are doing full table scans on this partitioned table. I'm still gathering up
the details for the items above (e.g how much have the objects changed) and
I'll probably run a 10053 trace on one of the bad queries to see what the
optimizer is doing on Monday, but I'd like to just poll for some additional
ideas. I *AM* getting partition elimination (thank goodness) but I've got
two FTS on one partition of this table that are just killing it. They want
to quantify the reason why this access has changed so I'm trying to think of
what kinds of stuff I can look at to try to do this.

I will add that this table was just rebuilt recently (through imp/exp)...
can the change in row to block density make the difference.... hmmmm....

Any ideas??

--

Please see the official ORACLE-L FAQ: http://www.orafaq.net
--

Author: Robert Freeman
INET: robertgfreeman_at_yahoo.com

Fat City Network Services    -- 858-538-5051 http://www.fatcity.com
San Diego, California        -- Mailing list and web hosting services


---------------------------------------------------------------------

To REMOVE yourself from this mailing list, send an E-Mail message

to: ListGuru_at_fatcity.com (note EXACT spelling of 'ListGuru') and in the message BODY, include a line containing: UNSUB ORACLE-L (or the name of mailing list you want to be removed from). You may also send the HELP command for other information (like subscribing).

--

Please see the official ORACLE-L FAQ: http://www.orafaq.net
--

Author: Cary Millsap
INET: cary.millsap_at_hotsos.com

Fat City Network Services    -- 858-538-5051 http://www.fatcity.com
San Diego, California        -- Mailing list and web hosting services


---------------------------------------------------------------------

To REMOVE yourself from this mailing list, send an E-Mail message

--

Please see the official ORACLE-L FAQ: http://www.orafaq.net
--

Author: Robert Freeman
INET: robertgfreeman_at_yahoo.com

Fat City Network Services    -- 858-538-5051 http://www.fatcity.com
San Diego, California        -- Mailing list and web hosting services


---------------------------------------------------------------------

To REMOVE yourself from this mailing list, send an E-Mail message