Comments on: 10046 Extended SQL Trace Interpretation

By: Understanding Oracle traces and plan | oraclebits

Understanding Oracle traces and plan | oraclebits — Wed, 06 Mar 2013 16:52:18 +0000

[...] (3 part series)http://hoopercharles.wordpress.com/2009/12/01/10046-extended-sql-trace-interpretation/ (2 part [...]

By: Charles Hooper

Charles Hooper — Wed, 02 Feb 2011 15:21:03 +0000

Ahmed,

Thank you for the kind words. Randolf and I spent a great deal of time working on those two chapters, so I am very happy to hear that you have found the information to be useful. I agree that a chapter on indexes written by Richard Foote would have been a great addition to the book – I believe that I suggest it (or that Richard should write an entire book) to the lead editor. I was also hoping that Tanel Poder would write a chapter for the book.

By: Ahmed AANGOUR

Ahmed AANGOUR — Wed, 02 Feb 2011 15:05:29 +0000

I have Jonathan’s book at home. I will have a look into it.
Thanks Charles for the knowledge you share with us via this blog and the OTN forum.
I’m reading the latest oakTable book and I really appreciate the 2 chapters you co-wrote with Randolf Geist.
These 2 chapters have become my survival guide for troubleshooting performance issues.
The rest of the book is also very helpful.
With one chapter written by Richard Foote about indexes, this book would have been PERFECT.

Regards,
Ahmed

By: Charles Hooper

Charles Hooper — Tue, 01 Feb 2011 18:04:52 +0000

Ahmed,

Interesting questions. Unfortunately, I do not think that I can answer the questions quite as well as a couple of other people. What you might need to keep in mind is that there is a chance that multiple undo blocks (scattered in the undo segments) will need to be applied to a block in the buffer cache so that the block will become consistent as of the start of the query (or the start of the session’s transaction). Jonathan Lewis explained the process well in his 10+ year old “Practical Oracle 8i” book. Search the book for consistent get. Amazon’s website allows you to view partial contents of the book, see pages 14-16.
http://www.amazon.com/Practical-Oracle8i-TM-Efficient-Databases/dp/0201715848/ref=cm_cr-mr-title#reader_0201715848

Jonathan also explained the process here:
http://jonathanlewis.wordpress.com/2009/06/12/consistent-gets-2/
http://jonathanlewis.wordpress.com/2009/06/09/quiz-night/

By: Ahmed AANGOUR

Ahmed AANGOUR — Tue, 01 Feb 2011 16:21:17 +0000

Thanks Charles for your answer.
I made an other test which consists in updating a table from one session and querying this table from another session and it seems that blocks access on undo segments are always done in a mono-block way (db file sequential read).
Do you confirm that? Do you have an idea why multiblock access cannot be done on undo segments?

By: Ahmed AANGOUR

Ahmed AANGOUR — Tue, 01 Feb 2011 14:38:14 +0000

[code] PARSE #6:c=3000,e=3175,p=0,cr=0,cu=0,mis=1,r=0,dep=0,og=1,plh=2102008689,tim=1296558900981804 WAIT #6: nam='db file sequential read' ela= 129 file#=3 block#=320 blocks=1 obj#=0 tim=1296558900982272 WAIT #6: nam='db file sequential read' ela= 82 file#=3 block#=10028 blocks=1 obj#=0 tim=1296558900982440 WAIT #6: nam='asynch descriptor resize' ela= 1 outstanding #aio=0 current aio limit=134 new aio limit=166 obj#=0 tim=1296558900982490 WAIT #6: nam='asynch descriptor resize' ela= 1 outstanding #aio=0 current aio limit=166 new aio limit=150 obj#=0 tim=1296558900982568 WAIT #6: nam='asynch descriptor resize' ela= 1 outstanding #aio=0 current aio limit=150 new aio limit=182 obj#=0 tim=1296558900982700 WAIT #6: nam='asynch descriptor resize' ela= 0 outstanding #aio=0 current aio limit=182 new aio limit=166 obj#=0 tim=1296558900982747 WAIT #6: nam='asynch descriptor resize' ela= 1 outstanding #aio=0 current aio limit=166 new aio limit=198 obj#=0 tim=1296558900982869 WAIT #6: nam='asynch descriptor resize' ela= 1 outstanding #aio=0 current aio limit=198 new aio limit=182 obj#=0 tim=1296558900982916 WAIT #6: nam='Disk file operations I/O' ela= 30 FileOperation=2 fileno=52 filetype=2 obj#=118919 tim=1296558900983023 WAIT #6: nam='db file sequential read' ela= 33861 file#=52 block#=12490 blocks=1 obj#=118919 tim=1296558901016907 WAIT #6: nam='Disk file operations I/O' ela= 27 FileOperation=2 fileno=55 filetype=2 obj#=118919 tim=1296558901017007 WAIT #6: nam='db file scattered read' ela= 640 file#=52 block#=12491 blocks=5 obj#=118919 tim=1296558901017765 WAIT #6: nam='db file sequential read' ela= 139 file#=3 block#=240 blocks=1 obj#=0 tim=1296558901018839 WAIT #6: nam='db file scattered read' ela= 5712 file#=52 block#=12496 blocks=8 obj#=118919 tim=1296558901025316 WAIT #6: nam='Disk file operations I/O' ela= 24 FileOperation=2 fileno=44 filetype=2 obj#=118919 tim=1296558901028150 WAIT #6: nam='db file scattered read' ela= 17726 file#=44 block#=1638601 blocks=7 obj#=118919 tim=1296558901045902 WAIT #6: nam='db file scattered read' ela= 6292 file#=44 block#=1638608 blocks=8 obj#=118919 tim=1296558901054605 WAIT #6: nam='db file scattered read' ela= 3264 file#=44 block#=1638617 blocks=7 obj#=118919 tim=1296558901060646 WAIT #6: nam='db file scattered read' ela= 7771 file#=44 block#=1638624 blocks=8 obj#=118919 tim=1296558901070624 WAIT #6: nam='db file scattered read' ela= 6359 file#=44 block#=1638633 blocks=7 obj#=118919 tim=1296558901079753 WAIT #6: nam='db file scattered read' ela= 3538 file#=44 block#=1638640 blocks=8 obj#=118919 tim=1296558901085661 WAIT #6: nam='db file scattered read' ela= 463 file#=44 block#=1638649 blocks=7 obj#=118919 tim=1296558901088796 WAIT #6: nam='Disk file operations I/O' ela= 24 FileOperation=2 fileno=45 filetype=2 obj#=118919 tim=1296558901091702 WAIT #6: nam='db file scattered read' ela= 3352 file#=45 block#=1778304 blocks=8 obj#=118919 tim=1296558901095097 WAIT #6: nam='db file scattered read' ela= 5323 file#=45 block#=1778313 blocks=7 obj#=118919 tim=1296558901103578 WAIT #6: nam='db file scattered read' ela= 3613 file#=45 block#=1778320 blocks=8 obj#=118919 tim=1296558901109981 WAIT #6: nam='db file scattered read' ela= 3583 file#=45 block#=1778329 blocks=7 obj#=118919 tim=1296558901116539 WAIT #6: nam='db file scattered read' ela= 5434 file#=45 block#=1778336 blocks=8 obj#=118919 tim=1296558901124916 WAIT #6: nam='db file scattered read' ela= 3128 file#=45 block#=1778345 blocks=7 obj#=118919 tim=1296558901131010 WAIT #6: nam='db file scattered read' ela= 1561 file#=45 block#=1778352 blocks=8 obj#=118919 tim=1296558901135106 WAIT #6: nam='Disk file operations I/O' ela= 26 FileOperation=2 fileno=46 filetype=2 obj#=118919 tim=1296558901138859 WAIT #6: nam='db file scattered read' ela= 10658 file#=46 block#=1778306 blocks=126 obj#=118919 tim=1296558901149544 WAIT #6: nam='Disk file operations I/O' ela= 40 FileOperation=2 fileno=47 filetype=2 obj#=118919 tim=1296558901200480 WAIT #6: nam='db file scattered read' ela= 8864 file#=47 block#=1778306 blocks=126 obj#=118919 tim=1296558901209384 WAIT #6: nam='Disk file operations I/O' ela= 42 FileOperation=2 fileno=48 filetype=2 obj#=118919 tim=1296558901255836 WAIT #6: nam='db file scattered read' ela= 11523 file#=48 block#=1770114 blocks=126 obj#=118919 tim=1296558901267396 WAIT #6: nam='Disk file operations I/O' ela= 41 FileOperation=2 fileno=50 filetype=2 obj#=118919 tim=1296558901319167 WAIT #6: nam='db file scattered read' ela= 15872 file#=50 block#=1778306 blocks=126 obj#=118919 tim=1296558901335080 [/code]

By: Charles Hooper

Charles Hooper — Tue, 01 Feb 2011 14:35:55 +0000

Ahmed,

A small portion of your 10046 trace file was lost (that happens sometimes when less than signs are used in comments). Enough of your comment remained for me to understand your question. Page 103 of the book “Expert Oracle Database Architecture: Oracle Database Programming 9i, 10g, and 11g Techniques and Solutions, Second Edition” explains very briefly why undo is still written for temp tables:
http://books.google.com/books?id=HPJDlGmecwcC&pg=PA103
More information is found in the same book around page 331.

By: Ahmed AANGOUR

Ahmed AANGOUR — Tue, 01 Feb 2011 13:51:16 +0000

Hi Charles,

I created a temporary table (named HISNEG_TT) which contains 3 indexes and ran the following statement:

insert into hisneg_tt select * from hisneg where rownum<=200000;

Here is an excerpt of the 10046 trace for this statement:

PARSING IN CURSOR #6 len=64 dep=0 uid=89 oct=2 lid=89 tim=1296558900981805 hv=3351120423 ad='1d78cb390' sqlid='0uygn4z3vw2j7'
 insert into hisneg_tt select * from hisneg where rownum<=200000

 select file_name,TABLESPACE_NAME from dba_data_files where file_id=3;
FILE_NAME
----------------------------------------------------------------------------------------------------
TABLESPACE_NAME
------------------------------
/UBIXT3/data/undotbs01.dbf
UNDOTBS1

I’m just inserting some rows into a global temporary table. I do not see the link between the temporary table and the rollback segment?
Could you please help me to understand this ?

Regards,
Ahmed

By: Charles Hooper

Charles Hooper — Thu, 27 Jan 2011 17:46:08 +0000

Notice that the file# is a very low number. File# 1 would be the datafile associated with the SYSTEM tablespace, so that probably provides an indication about whether or not this non-object is in a user-managed area of the database or a system managed area of the database (not the best choice of words, but close enough). File# 2 is *probably* the datafile for the undo tablespace. Riyaj Shamsudeen provides a trace file section in a comment on the following page, which also shows “file#=2″ and “obj#=0″, and he explains that obj#=0 indicates reading from undo:
http://orainternals.wordpress.com/about/

Doug Burns also hinted that same answer in a comment in this article:
http://oracledoug.com/serendipity/index.php?/archives/1321-11g-and-direct-path-reads.html

By: walid

walid — Thu, 27 Jan 2011 17:15:55 +0000

Hi,
I have generated 10046 trace.
But when I checked the wait event for a query, I find abj#=0.
*** 2011-01-25 00:39:47.711
WAIT #30: nam=’db file sequential read’ ela= 9292 file#=2 block#=329387 blocks=1 obj#=0 tim=23252790781437
WAIT #30: nam=’db file sequential read’ ela= 8916 file#=2 block#=329388 blocks=1 obj#=0 tim=23252790790688
WAIT #30: nam=’db file sequential read’ ela= 8318 file#=2 block#=329389 blocks=1 obj#=0 tim=23252790810226
WAIT #30: nam=’db file sequential read’ ela= 6725 file#=2 block#=329390 blocks=1 obj#=0 tim=23252790818804
WAIT #30: nam=’db file sequential read’ ela= 6541 file#=2 block#=329391 blocks=1 obj#=0 tim=23252790827108

Have you an idea about obj”#=0?
Thanks,