The documentation and installation guides can be found here. 

The Advanced Tutorials Guide contains up-to-date How-Tos including:

• Create a Tabular Form
• Create a Parameterized Report
• Create a Drill Down Report
• Control Form Layout
• Work with Check Boxes
• Implement a Web Service
• Create a Stacked Bar Chart
• Upload and Download Files in an Application
• Incorporate JavaScript into an Application
• Build an Access Control Page
• Review a Packaged Application
• Create a Master Detail PDF Report
• Design an Issue Tracking Application
• Build and Deploy an Issue Tracking Application

The new features are documented here.

I will now update my courseware with the new features and deliver my courses as of August with the new version.

I was not aware that they are already online.

Congratulations!!!

=;-)

LH

It is possible to request a workspace now.

The documentation is still not available yet, instead the online docs for APEX 3.2 are linked.

Let’s test it.

=;-)

LH

It is a very fun class and I have some very interesting people in the course.

 Mark Anderson
is a highly skilled trainer and delivers the Microsoft SQL Server 2008 for the Experienced Oracle Database Administrator course for internal Microsoft staff and is responsible for the high end Microsoft customers. This course is definitely one I would like to attend!

He is about to publish his latest  book on SQL SERVER for ORACLE DBAs shortly. I am pretty shure that this will be another candidate for my BOOKS WORTH THEIR MONEY category  and I will definitely get a copy ASAP..

=;-)

sysdba

Use this link to request a workspace as hosted environment:

I will post about my first impressions soon.

=;-)

Scenario is 10.2.0.1  Express Edition on Win32.

I have a table called job_history in the HR scheme which has a composit primary key : and there is an index for the primary key:

SQL> select index_name from user_indexes where table_name=’JOB_HISTORY’;

INDEX_NAME

Does the following info help to explain?

——————————
JHIST_EMP_ID_ST_DATE_PK
JHIST_JOB_IX
JHIST_EMPLOYEE_IX
JHIST_DEPARTMENT_IX

This is my constraint in the data dictionary:

SQL> select column_name , position
  2   from user_cons_columns
  3  where constraint_name=’JHIST_EMP_ID_ST_DATE_PK’;

COLUMN_NAME      POSITION
———————–    ———————————
EMPLOYEE_ID        1

START_DATE          2

Now I drop the primary key constraint

SQL> alter table job_history drop constraint
  2  JHIST_EMP_ID_ST_DATE_PK including indexes;

Table altered.

And here is the surprise:

SQL> select index_name from user_indexes where table_name=’JOB_HISTORY’;

INDEX_NAME
——————————
JHIST_EMP_ID_ST_DATE_PK
JHIST_JOB_IX
JHIST_EMPLOYEE_IX
JHIST_DEPARTMENT_IX

I do not see the point here. Why is the index still there?

Evern though I have explicitly written INCLUDING INDEXES.

BUG or Feature?!

Can anyone explain?

I have tried to run a  top level anonymous block in which assigned a value to a BIND variable and got kicked out of my session with an ORA-03113.

SQL> VAR lname VARCHAR2(4)

SQL> <<my_lable>>
 BEGIN
    :lname:=’LUTZ’;
   end;
/

ERROR:
ORA-03114: not connected to ORACLE
 <<my_lable>>
*
ERROR at line 1:
ORA-03113: end-of-file on communication channel

THIS IS A BUG which exists in 10gR2.

Only top level blocks are affected by this BUG.
 After I wrapped the block into another block it worked.

Here is the workaround:

BEGIN
  <<my_lable>>
 BEGIN
 :lname:=’LUTZ’;
 END;
END;
/

PL/SQL procedure successfully completed.

 I have tested only WIN 32 and LINUX32 =>  XE and Enterprise Edition on Win32.  It has been fixed in the meentime, but I am not sure with which patch.
In EE 11gR1  it works.

The Database Machine runs Oracle Database 11g and Real Application Clusters, Oracle Enterprise Linux, and includes all the required Infiniband infrastructure and related hardware.

This is the first hardware product ever Oracle marketed under its own name.
It integrates Oracle Database Software with highly suffisticated hardware using Linux and ASM to build a very fast database maschine.

You can find more detailed information on Kevin Closson’s blog where he maintains an FAQ page.
Kevin is Performance Architect in Oracle’s Systems Technology Group which is a part of Server Technologies.

A few impressions from OOW by Tim Hall can be found on his blog.

I will deliver a series of 11g Administration Courses for Oracle again.

One day after I arrived the there was a a big fire which burned down the house of the maglis ash-shura,  consultative councel    completely.

I was lucky enough to have my new camera with me to take some exciting shots.

It looked to me as if there were multiple fires burning at the same time.

Plus a sweater with the SELECT Journal Contributer logo:

Additionally I was granted a one year free membership for Independent Oracle User Group (IOUG) for my article on DETECTING AND REPAIRING CORRUPT BLOCKS IN ORACLE DATABASE 11g WITH THE DATA RECOVERY ADVISER which is included in the curernt issue Volume 15 Number 3/2008.

Thanks IOUG and SELECT, this was a nice surprise.

=;-)

I am very happy that this has become such a successfull forum here.

Thanks for reading
I’ll keep on posting in the future!

   Oracle Unbreakable Linux.

 Before this the clusterware needed at least one Standard Edition License.

Now you can download the clusterware and use it as an extension of your Linux to make virtually any application highly available even without having an Oracle Database Server.

For more info on this and ready made scripts see my previouse post on how to protect a single instance database with the clusterware.

 This is really cool!

=;-) 

 yearly conference of the Swiss Oracle Usergroup.

The topics were :

- Real Application Testing with Oracle 11g
     REPLAY DATABASE and
     SQL PERFORMANCE ADVISOR

- SQL Plan Management and Automatic SQL Tuning in 11g

I have used my new courseware which
I have produced in the last few weeks for 11g New Features.

I offer to to deliver special courses for topics such as

- High Availability with Flashback and other 10g and 11g features
- Backup 10g & 11g
- Memory Management 10g & 11g
- Upgrade Issues 10g & 11g
- 11g New Features Overview

If you are interested in these topics please contact me by email:

info@sysdba.ch

I have a completely packed class with 21 students.

 Most of the students are Oracle Trainers, Support Agents and Consultants.

 Muhammad is an internationally recognized artist and has been painting since more than 25 years now. He had exhibitions in Paris, Hamburg, Berlin, Zurich and many other places.

He lives in a very poor area near the great basar Khan Al-Khalili.

In April he will have an exhibition in Talaat Harb Streeet in Downtown Cairo and I hope  that he will sell a great number of his beautiful naive paintings again.

I will deliver a number of Oracle 11g trainings here.

This is the first time I am in Egypt after 12 years.
In the middel of the ninetees I had been living here for a couple of longer times.
I am very interested in the developments in Egyptian society since then.

Hopefully I will meet a lot of interesting people. 

Yesterday I had organized a little dinner with Steven Feuerstein, my former collegue Christoph Burandt and his lovely wife Silvia as well as my dear friend Jenny.

Steven had delivered a two day seminar for Oracle University in Munich. Although he was very sick in the last days, he had caught  a chest cold and swollowed an enormaouse amount of remedies in order to be able to deliver anyway, he preferred not to hang around in his room alone and come for our little dinner.

Thanks for coming Steven! Hope you’re feeling better already.

We talked about everything but PL/SQL and had a lot of fun.
Christoph is the one who had introduced me to PL/SQL in 2000 and I was very happy to meet him once again.
I think that he is one of the best instructors in this field you can get in Germany.
His wife is from Peru and Steven adressed her in Spanish.

Everybody could speak Spanish except me!

=;-)

 This patchset not only contains bugfixes but also comes with a number of New Features.

One of the highlights is the possibilty to capture a workload in a 10.2.0.4 database and ship it to an 11g database for testing.

Only with this feature in place the 11g Real Application Testing Feature DATABASE REPLAY makes real sense.
Before this patchset it was only possible to capture a workload in an 11g database and run it for  testing in an also 11g database.

For more detailed evaluation please see my chapter on CHANGE MANAGEMENT in the forthcoming 11g New Features book.

 It will be shipping aroung middle of the year.

Now customers can check in front what an upgrade would bring for them.

Here is a brief desription of the new 1og Feature:

LUTZ AS SYSDBA @ orcl10g SQL> SELECT * FROM v$version;

BANNER
—————————————————————-
Oracle Database 10g Enterprise Edition Release 10.2.0.4.0 – Prod
PL/SQL Release 10.2.0.4.0 – Production
CORE    10.2.0.4.0      Production
TNS for Linux: Version 10.2.0.4.0 – Production
NLSRTL Version 10.2.0.4.0 – Production

In 10.2.0.4 a new paramter is used to enable workload capturing:

LUTZ AS SYSDBA @ orcl10g SQL> show parameter pre_11

NAME                                 TYPE        VALUE
———————————— ———– ——————————
pre_11g_enable_capture               boolean     FALSE
Note:
The PRE_11G_ENABLE_CAPTURE initialization parameter can only be used with Oracle Database 10g Release 2 (10.2). This parameter is not valid in subsequent releases. After upgrading the database, you will need to remove the parameter from the server parameter file (spfile) or the initialization parameter file (init.ora); otherwise, the database will fail to start up.

Oracle even has created the script $ORACLE_HOME/rdbms/admin/wrrenbl.sql which can be used to modity the parameter.  The script to disable workload capture is $ORACLE_HOME/rdbms/admin/wrrdsbl.sql.  

The package DBMS_WORKLOAD_CAPTURE has been added to the 10g RDBMS. This is the interface to workload capturing and filtering:

SYS AS SYSDBA @ orcl10g SQL> desc dbms_workload_capture

PROCEDURE ADD_FILTER
 Argument Name                  Type                    In/Out Default?
 —————————— ———————– —— ——–
 FNAME                          VARCHAR2                IN
 FATTRIBUTE                     VARCHAR2                IN
 FVALUE                         VARCHAR2                IN
PROCEDURE ADD_FILTER
 Argument Name                  Type                    In/Out Default?
 —————————— ———————– —— ——–
 FNAME                          VARCHAR2                IN
 FATTRIBUTE                     VARCHAR2                IN
 FVALUE                         NUMBER                  IN
PROCEDURE DELETE_CAPTURE_INFO
 Argument Name                  Type                    In/Out Default?
 —————————— ———————– —— ——–
 CAPTURE_ID                     NUMBER                  IN
PROCEDURE DELETE_FILTER
 Argument Name                  Type                    In/Out Default?
 —————————— ———————– —— ——–
 FNAME                          VARCHAR2                IN
PROCEDURE EXPORT_AWR
 Argument Name                  Type                    In/Out Default?
 —————————— ———————– —— ——–
 CAPTURE_ID                     NUMBER                  IN
PROCEDURE FINISH_CAPTURE
 Argument Name                  Type                    In/Out Default?
 —————————— ———————– —— ——–
 TIMEOUT                        NUMBER                  IN     DEFAULT
 REASON                         VARCHAR2                IN     DEFAULT
FUNCTION GET_CAPTURE_INFO RETURNS NUMBER
 Argument Name                  Type                    In/Out Default?
 —————————— ———————– —— ——–
 DIR                            VARCHAR2                IN
FUNCTION REPORT RETURNS CLOB
 Argument Name                  Type                    In/Out Default?
 —————————— ———————– —— ——–
 CAPTURE_ID                     NUMBER                  IN
 FORMAT                         VARCHAR2                IN
PROCEDURE START_CAPTURE
 Argument Name                  Type                    In/Out Default?
 —————————— ———————– —— ——–
 NAME                           VARCHAR2                IN
 DIR                            VARCHAR2                IN
 DURATION                       NUMBER                  IN     DEFAULT
 DEFAULT_ACTION                 VARCHAR2                IN     DEFAULT
 AUTO_UNRESTRICT                BOOLEAN                 IN     DEFAULT

And the data dictionary views DBA_WORKLOAD_CAPTURES and DBA_WORKLOAD_FILTERS can be used to monitor the captures.

 As in 11g the packages captures all incomming requestes into external files which need to be shipped to a testing system and pre-processed on the testing system. Those files contain logical information to replay client requests in the appropriate sequence and with all contention and waiting as it happened on the capture site. On the testing site we need to start replay clients to run the test workload.

This is the interface for REPLAY DATABASE in the OEM GUI:

The executable $ORACLE_HOME/bin/wrc is only available in 11g.

 So we can now capture in 10g and replay in 11g.

 I am just downloading the latest patchset for RDBMS Server 10.2.0.4 for Linux x86.

It is documented in Metalink NOTE: 316900.1

 Components of This Patch Set

The following are the updated components of this patch set:

• Up to January 2008 CPU is included in this patch set.
• Platform-specific bug fixes as listed in the “Fixed Platform-Specific Bugs” section
• Generic bug fixes as listed in the List of Bugs fixed document
• Oracle Configuration Manager 10.2.7.1

I am very curiose about the possibility to capture a workload in 10g and run for Real Application Testing in 11g.

It should be included in this patchset.

I will try it and post about it here ASAP.

=;-)

All you need to do is threaten them!

They have removed the links to my content.

We’re Sorry – We Were Unable to Locate The Article You Requested.It’s possible that this article is no longer available, has been reported as a bad link or is currently being reviewed/edited by its author.

We apologize for the inconvenience! Please use the menu system or the search boxes to locate other items that may be of interest.

If you feel there has been an error, please contact us and let us know about the issue and we’ll get on it immediately!

This is the second time already that this strange organization steals my posts and demands money from their “MEMBERS” to look at it. I had them remove my posts already in 2006.

This so call usergroup does not have any publications of its own and there are no meetings of the members (if they had any, a lot of people would show up there and yell at their chairman I think). Their members seem to be just people who are not able to search the web by themselves and pay for a compilation of links to stolen content they could have for free in the web.

 It seems to be  basically just one person behind this organization:

I think that the expression usergroup should be a protected expression by law and that it should be possible to screw such crooks for making money with the intellectual property of other people.

I followed the link and found a website where someone had a blog which justs consists of stolen posts. He just had added my post to his blog and did not even mention me. His name is, oh no! I am not even going to mention it here because this would only increase his daily hits!

I contacted him by email and asked him to remove my post right away – nothing ever happened since more than a week now.

Theft of intellectual property seems to become a more and more fashionable practice.

Here is a link to a list of some recent popular cases of this kind of vandalism.

I am not sure if I want to continue blogging in the future.

Let’s publish books again
and let people pay for it!

 I had a look at his profile and found a very interesting link there, which I want to share here.
It is a website from Oracle Server Technologies. It gives very intersing information about how to protect against SQL injection, one of the most serios threats for DBAs.

Have a look here! You will like it!

 =;-)

Lutz

 We chatted about God and the world and we had lots of fun.

Amoung others we talked about things in Oracle 11g which do not really work yet and¦or are undocumented or documented wrong.
I was lucky to be able to tell him about things which he did not know about yet
like the not yet documented parameter DDL_LOGGING, see my post here.  I think that Oracle development just forgot to document it.

It is hard to catch up with Laurent sometimes!

I am the reviewer of Laurent’s yet to be published book on 11g SQL for advanced developers and DBAs and he demonstrated some features he had filed a BUG for in the last days.

Of course Oracle 11g was a big topic and we have found out some interesting things about system partitioning ( it is more that he found out than me, altough I teach the stuff to people in my New Features courses !!!!)

He  talked about something in one of the chapters in his book which he had found about system partitioning in 11g and which did not work. 

I got suspicious because I was so sure that it was possible, since it was something I teach to my students in my 1day 11g New Features Overview classes.
But I had to back down and confess again that I did not really verify every single bit of code of the Oracle Coureseware.

This is why I want to share the lecture I had to take tonight (luckily with a couple of beers!):

Oracle has introduced a number of new partitioning algorithms with release 11g.  One of these new features is called SYSTEM PARTITIONING. This is very handy if you want to partition a table but cannot make uo your mind about a partitioning key for some reason. Maybe there is only one column holding xml data in the table or what so ever.

The current version of the courseware for the 11g New  Features Courses says that the following syntax is valid and Laurent showed me that it is not:

– I create a table which is system partitioned:

SYS AS SYSDBA @ orcl SQL> create table lutz.t3 (col1 number, col2 number)
  2  partition by system
  3  (partition p1);
Table created.

– I create a bind varable (first time a number second time a string)
– which I want to use to tell Oracle which partition to user for an INSERT
–and assign it the value 1 and try an INSERT:

SYS AS SYSDBA @ orcl SQL> var par number
SYS AS SYSDBA @ orcl SQL> exec :par:=1
SYS AS SYSDBA @ orcl SQL> insert into lutz.t3 partition (:par) values (1,2);
insert into lutz.t3 partition (:par) values (1,2)
                               *
ERROR at line 1:
ORA-14108: illegal partition-extended table name syntax

– I try another time with a bid variable of type string:

SYS AS SYSDBA @ orcl SQL> var par varchar2(2)
SYS AS SYSDBA @ orcl SQL> exec :par:=’p1′
SYS AS SYSDBA @ orcl SQL> insert into lutz.t3 partition (:par) values (1,2);
insert into lutz.t3 partition (:par) values (1,2)
                               *
ERROR at line 1:
ORA-14108: illegal partition-extended table name syntax

– I try it for a last time and now I try to use the partition number directly:

SYS AS SYSDBA @ orcl SQL> insert into lutz.t3 partition (1) values (1,2);

insert into lutz.t3 partition (1) values (1,2)
                               *
ERROR at line 1:
ORA-14108: illegal partition-extended table name syntax

The same error pops up!

And here is the explanation of the error:

[oracle@rhas4 ~]$ oerr ora 14108
14108, 00000, “illegal partition-extended table name syntax”
// *Cause:  Partition to be accessed may only be specified using its name.
//          User attempted to use a partition number or a bind variable.
// *Action: Modify statement to refer to a partition using its name

You can only reference a system partition with its name not by its number or even via a bind variable! At least not in 11.0.6!!!

Sorry people but I just could not yet test all of the more than 500 New Features in the courseware. But I will keep on hitting the road, for sure!

=;-)

In his courses he always encourages his students to send their comments to Bryn Llewellyn the project manager for PL/SQL at Oracle.

A while ago Steven already has launched a PL/SQL forum in OTN where people can directly ask him questions.

Now Steven has put together a website to make it really easy for developers to communicate their concerns and priorities to Bryn (with his support, of course).

 Seems to become an interesting place and I will add it to my blogrole. 

I was especially pleased to see that almost all of the delegates had already attented at least one or two trainings from me last year, so we were already familiar with each other.

Since I did not have the chance to attend the UK Oracle Usergroup meeting it was nice to have Luca Canali again in the class. He had delivered a speach about the ASM internal  in Birmingham last December and I would have really loved to listen to it. He is one of the few persons I know who can really tell about the limits of ASM from real life experience.

Also I was so happy to meet Christopher Lambert again. He just seems to know everything about Oracle Enterprise Manager and I always learn from him.

Here are a few snapshots from the classroom I have taken with my cellphone.

 I hope that I was able to meet their expectations once again and will be invited again.

It is a great honor for me to deliver for you guys!

On the way back we had a stop over at Istanbul airport where

  Tonguç Yilmaz waited for me
to chat and have a coffee. Can you immagine that there is somebody waiting for you at 6:15h in the morning just to hug you and have a chat at the airport.
This is real friendship and hospitality!

He updated me about the latest gossip from the bloggers community and we talked about OOW and UKOUG meeting and who we have met and who is new to the scene. 
 Thanks for coming Tonguç, you are such a nice guy!

=;-)

 Swiss Oracle Usergroup’s Newsletter 1/2008.

It contains my article on some 11g New Features related to DDL.
I had posted about them a while ago in my 11g Top New Features Series.

You can read about DDL LOGGING and DDL LOCKING in 11g in my 11g posts.

Thank you SOUG!

=;-)

for download on Edelivery

This patchset does not only contain bugfixes but also a number of NEW FEATURES .

Here is an overview of the new features if you upgrade from 10.2.0.3:

• Provides data masking, that is the capability to replace sensitive data with false but realistic looking data on test and development databases.
• Allows usage of commonly-used privileged delegation software, such as sudo or pbrun, to fit into the existing credential subsystem in Enterprise Manager.
• Allows bidirectional flow of events between Enterprise Manager and third-party products.
• Provides users through the Service Level Management Pack the capability to record transactions on Forms 6i-based applications and periodically replay these transactions to monitor performance and availability.
• Provides DHTML transaction recording and monitoring features that allow you to record and playback DHTML-based applications. The features help you to capture mouse and keyboard actions, in addition to the HTTP requests.
• Allows you to monitor and manage Oracle BI Suite Enterprise Edition (Maui release). You can discover Oracle BI EE servers, view configuration information, monitor performance, and so on.
• Provides, through the Diagnostics Pack for Non-Oracle Middleware, Java Applications Diagnostics Expert (JADE) that allows you to diagnose performance problems in production Java applications with minimal performance overhead and without shutting down the applications.
• Provides improved SOA management capabilities that enable you to automate the provisioning of BPEL Processes using the deployment procedures. Also ofers configuration management of BPEL Process Manager.
• Provides out-of-box plug-in support for Microsoft Exchange targets. Allows you to monitor performance, availability, usage statistics, and administration of Exchange Servers.
• Offers support for the following plug-ins:
• • EMC CLARiiON System: You can view storage, configuration, and performance information for laying out best practices, analyzing performance bottlenecks, and resolving issues.
  • VMware ESX Server: You can monitor configuration and performance data of VMware ESX Server 3.
  • Microsoft SQL Server: You can access core administration capabilities.
  • IBM DB2 Database: You can monitor IBM DB2 Release 9.
  • EMC Symmetrix DMX System: Provides fixes for certain performance-related issues found in Reports and enhances its capability to function better in a larger environment that may consist of over 3000 disks.

You can download the versions for
- WIN 32 bit
- Linux x86

The documentation is here.

Also there is a new version of the agent available for download for WIN 32 and Linux x68.

as well as a number of new PLUGINs for the OEM.

=;-)

This software validates a checksum which Oracle created for a buffer before it started writing it to disk. If the checksum of the block which NetApp is to write to the storage is not the same than the one Oracle has produced Snapvalidator creates an error message and prevents the data from being written to the storage.

Only with such a check we can be sure that exactly what we sent from the buffer cache is what arrives on disk in the end.

Let’s test it!

=;-)

Here are a few impressions:

  Elzbieta Gajewska 
is an outstandingly fast learning young computer scientist from Poland. She was already in my RAC course a few months ago.  And again she asked me brilliant questions, some of  them could be good for challenging the Oaktable Network.

Some of the delegates are already taking part in the second class from me. It is always fun to deliver a course for CERN since it is never boring since these people are so highly motivated.

And again I met Anton Topurov
who is already testing with 11g in CERN and has also the Beta Version of the yet to come soon 10.2.0.4. patch already. But he does not talk about it yet. 
He also was in my RAC class recently.

I am very greateful that he has managed to get me into a group of people who can visit the CMS experiment.

Download sources

Oracle Virtualization Homepage

Official Discussion Forum for VM

Documentation for VM

VM Wiki

Pre Configured Templates

FAQ

Podcast with Kurt Hackel (Oracle VM Architect)

I have asked Wim Coekaerts  during his Meet the Experts session
in the OTN lounge last week at the OOW about trainings on Oracle VM.  He said that there will be new courses on this available soon from Oracle University.

the latest DOAG NEWS   4/2007 in my mailbox.
It contains the German version of my article on 11g New Features of the Cost Based Optimizer.

The DOAG (Deutsche Oracle Anwender Gruppe) is the German Oracle user group which celebrates its 20 aniversary this year.

Congratulations DOAG!

I have recceived a number of comments on this post by Peter about the AUTO mode of the parameter who asked if it really works because he had tewsted and did not find the result he had expected. This made me test as well and I had the same experience as Peter.

So I went to the demo ground here at the OOW yesterday and asked the guys from Server Technologies directly what the algorithm used is and received a surprising answer:

They told me that only MANUAL and FORCE are supported settings for the parameter. AUTO is neither documented nor supported!!!!

I asked them why it is possible to set the parameter to AUTO in the Enterprise manager and also there is an error message telling me this:

SYS AS SYSDBA @ ocrl11 SQL> show parameter result_cache

NAME                                 TYPE        VALUE
———————————— ———– ——————————
client_result_cache_lag              big integer 3000
client_result_cache_size             big integer 0
result_cache_max_result              integer     5
result_cache_max_size                big integer 1280K
result_cache_mode                    string      MANUAL
result_cache_remote_expiration       integer     0
 

SYS AS SYSDBA @ ocrl11 SQL> alter system set result_cache_mode=nonsense;
alter system set result_cache_mode=nonsense
*
ERROR at line 1:
ORA-00096: invalid value NONSENSE for parameter result_cache_mode, must be from among FORCE, MANUAL, AUTO

And as Peter has found out: IT DOES NOT WORK WITH AUTO obviousely!
Very good catch Peter!

The guys from Server Technologies told me that it will probably never work and will never be supported with AUTO. And if I think it over it is not bad to manually control which statements are cached and which not.

=;-)

http://www.oracle.com/technology/tech/virtualization/index.html?rssid=rss_otn_news

Virtualization Technology Center Virtualization is the process of abstracting computing resources such that multiple operating system and application images can share a single physical server, bringing significant cost-of-ownership and manageability benefits. Through its Oracle VM product, Oracle offers scalable, low-cost server virtualization for heterogeneous applications.

Topic: How to migrate a DB from file system in to ASM storage
Time: 5:00 pm
Location: uncoference overlook II @ Moscone West Level 3 (next to OTN lounge)

 It will be life demo with open end Q&A.

Borders bookstore just provided me with two copies of the new book on ASM
which I will raffle at my little presentation today.

Oracle Automatic Storage Management  by Nitin Vengurlekar, Murali Vallath & Richard Long

Good luck!

Looking forward to meeting you all there

=;-)

Here is one which is a MUST READ for everybody who is interested how Automatic Storage Management (ASM) really works:
Oracle Automatic Storage Management  
What makes this book especially interesting is the fact that the authors are the Oracle developers of ASM themselves, Nitin Vengurlekar and  Richard Long, and Murali Vallath., the president of the RAC SIG,  plus the architect of the technology Bill Bridge who wrote the foreword for the book .

This book keeps its promises and covers the 10g and 11g features of ASM in detail.
And it is an easy reader with a sence of humor:

“The ACD is a log that allows ASM to make atomic changes to multiple data structures. It is similar to the redo logs used by the Oracle RDBMS. Each instance gests a 42 MB chunk of the ACD. (As Douglas Adams explained in Hitchiker’s Guide to the Galary, “42″ is the answer to the ultimate question of life, the Univese and Everything.) …

 This is the way of reading I like!!

Thank you guys, I love it

patchset 10.2.0.4 which will be available very soon!!!

which will allow to mask sensitive data in a captured workload 
so it is not possible to read the data in plain in the testing environment, but still run the workload for testing anyway. This will definitively make it more attractive for a number of customers to use Database Replay, one of the Real Application Testing (RAT) features in 11g.

Also it will be possible to capture a workload on a 10.2.0.4 database and run it on an 11g database for testing!!!

I want to share two of them:   The first is SQL Real Time Monitoring:
With OEM 11g it is possible to monitor every sql statement in real time while it is running.

 We will soon have a graphical interface which we can use to monitor sql statements while they are running and can drill down into the statement’s steps     

  and see which step takes how much DB-TIME and what does it consume: is it CPU is it I/O and so forth. This makes it very easy to find out which step needs tuning exactly. In the demo yesterday they executed a long runnig parallel select and showed how easy it is to find out which full table scan need tuning in order to improve the performance of teh entire statement.

This is a very cool feature!

My next post will be about Data masking for testing.
=,-)

Microsoft Windows (32-bit) (1.7 GB) 
Microsoft Windows (x64) (1.7 GB)
Linux x86 (1.7 GB)
Linux x86-64 (1.8 GB)
Solaris (SPARC) (64-bit) (1.9 GB)
AIX (PPC64) Disk 1, Disk 2 (2.3 GB) 
HP-UX Itanium Disk 1, Disk 2 (2.3 GB)

Wim told me that the software for the OVS will be
available for download latest Thursday this week (Nov. 15th. 2007).

I asked him whether OVS is zertified for RAC and he said
that it is not right now but it will be soon, sometime beginning next year.

The software is not downloadable yet as far as I understood but will around soon as well.

Here are a few slides from the presentation:

With this technology it will be possible to store virtually everything in ASM storage and it is even possible to migrate running live VMs from one physical maschine to another one within minutes.

I cannot wait to test it!

who is right now recording a Pod Cast with Laurent Schneider.

OTN handed out T-shirts,. have a look, can you believe it? It is not even executable!!

 ”Its hearting my eyes” Laurent says

There are a number of very intersting people here some of them under cover as it looks like  Tom will deliver an expert session this afternoon at the OTN lounge.

Here are a few impression of the Moscone Center from today:
  

Today in the evening Larry Ellison will deliver a key note and
I hope that he will announce some intersting new features.

This time I had chosen US Airways and UA which was much better then my last trip which was with AA and Delta and which was a torture.

When I was on the way from the airplane to the baggage all of a sudden my cellphone was ringing.
  It was Laurent Schneider who called me 
and told me that he was waiting for me at the exit with his coleague Daniel from LC SYSTEMS to pick me up and take me downtown for a nice dinner. What a nice surprise!

Tomorrow I will take part in the 11g Exam Cram and
in the evening I will go to the ACE dinner.

In previous releases a DDL statement always errors out respectively hangs in a hold if an exclusive lock (X) on a required resource is not available.

Let’s have a look at a little example:

The user HR looks a table in shared) mode by acquiring ROW LOCKS (RS) on 0 rows:

HR@orcl SQL> select * from employees where 1=2 for update;

no rows selected

User SYSTEM tries to acquire an EXLUSIVE LOCK (X) on the table:
SYSTEM@orcl SQL> LOCK TABLE hr.employees IN EXCLUSIVE MODE;

Here SYSTEM will wait until the shared table lock is released by HR although there are no rows locked explicitly!!! The SELECT FOR UPDATE does not only lock rows explicitly but also requires a shared table lock.

Already in 10g it was possible to specify NOWAIT in order not to hang in a hold situation but to let the statement error out immediately:

SYSTEM@orcl SQL> LOCK TABLE hr.employees IN EXCLUSIVE MODE NOWAIT;

*ERROR at line 1:ORA-00054:
resource busy and acquire with NOWAIT specified or timeout expired 

As of Oracle 11g it is possible to specify a timeout period for which a lock request should wait before erroring out. This can be very helpful because it is very likely that just a few seconds later the EXCLUSIVE ENQUEUE might be available. We now have the chance to wait for it before the DDL statement fails:

SYSTEM@orcl SQL> LOCK TABLE hr.employees IN EXCLUSIVE MODE WAIT 10;

ERROR at line 1:ORA-00054: resource busy and acquire with NOWAIT specified or timeout expired

Here the DDL statement errors out only after 10 seconds if the requested lock is still not available!

Forthermore Shared exclusive locks (SRX) have been improved in Oracle database 11g.
They are used now instead of Exclusive locks (X) for the following operations:
-          CREATE INDEX ONLINE
-          CREATE MATARIALIZED VIEW LOG
-          ALTER TABLE ENABLE CONSTRAINT  NOVALIDATE
In 11 g transactions can be active and hold implicit (RX) or explicit (RS) row locks on a segment which is undergoing one of the above mentioned operations. This can reduce wait times in highly concurrent configurations for example for rebuilding online indexes.

Thank you very much SOUG!
I will definitely keep on publishing…

…and blogging

=;-)

Linux x86 (as of August 14th. 2007)
Linux x86-64bit ( October 30th. 2007)
Windows 32 bit ( October 25th. 2007)

I have doublechecked again about this statement and have contacted Oracle Pre-Sales in Switzerland in order to find out what is the truth.
A Pre-Sales manager in Switzerland asured me that it is not critical from where you down  load and it is not relevant for support where you got the software from, OTN or edelivery. Nevertheless, there is not always the same software available on both sites.

O.K., this relativates the first statement and I can sleep more comfortable again.

 This is what OTN tells us about the downloads from there:
On OTN software downloads are free, and each comes with a Development License that allows you to use full versions of the products only while developing and prototyping your applications (or for strictly self-educational purposes). 

Read it carefully!

=;-) 

Obviousely wordpress.com is banned by the censor of the Turkish state.

This week I will deliver a course on Oracle 11g New Features for Administrators at the site of a delivery partner of Oracle Education Turkey.

As of Oracle database 11g we can have loggin information for DDL operations in the alert log files. I say files in plural because we have two different alert log files now in 11g:
one is the traditional alert_<sid>.log as we had it ever since,
the other one is an log.xml file which is located in the diagnostic directory in the alert folder. The xml file gives you a lot more information than the traditional alert log file.

The diagnostic directory is specified by the parameter DIAGNOSTIC_DEST  in 11g which depricates all the *_DUMP_DEST parameters from older releases. It defaults to $ORACLE_BASE/diag/

This feature can be very handy because it shows a logging entry for every DDL statement with a timestamp in the two alert log files. You will like it if you must find out when the wrong table was dropped or truncated ecactly in order to flash it back.

It is not enabled by default and you must change the parameter to TRUE first in 11gR1. 

Here is a little demo:

ALTER SYSTEM SET enable_ddl_logging=TRUE SCOPE=BOTH; SYS AS SYSDBA @ orcl SQL> show parameter ddlNAME                                 TYPE        VALUE
———————————— ———– ——————————
ddl_lock_timeout                     integer     0
enable_ddl_logging                   boolean     TRUE  

[oracle@edw2r8p0 ~]$ tail -f /u01/app/oracle/diag/rdbms/orcl/orcl/alert/log.xml

<txt>drop table hr.tt 
</txt></msg>
<msg time=’2007-10-16T15:29:58.343+02:00′ org_id=’oracle’ comp_id=’rdbms’ msg_id=’opiexe:3933:4222364190′ client_id=” type=’NOTIFICATION’ group=‘schema_ddl’ level=’16′ module=’SQL*Plus’ pid=’27158′> <txt>drop table t
 </txt>
</msg> Here is the same tail for the traditional alert_<sid>.log: Tue Oct 16 15:29:58 2007drop table tNice one, eh?
=;-) 

Günther Stürner is Vice President Sales Consulting and head of the Business Unit Server Technologies of ORACLE Germany. He is responsible for the entire database and middleware business in Germany.
Günther Stürner is one of the Oracle employees who were already inside the company from the very beginning in Germany. I have published this interview in the SOUG Newsletter 4/2007 and want to share it here:

sysdba:

Oracle 11g will be released shortly and we will take another giant step into the direction of the Self Managing Database. Tasks a DBA in Oracle 10g had to perform manually will now be automatically managed by the server. One of the highlights in 11g surely is Automatic SQL Tuning.
The optimizer now has the chance to automatically create SQL-profiles and also automatically implement them if the calculated COSTs are lower than those of previous executions of the same statement.Does this meen for us DBAs that Statement Tuning is not necessary at all with Oracle 11g?

G.S.

Questions like these have already been asked when Oracle 10g was released. I clearly remember a 10g meeting  where one of the delegates after my speech said :’… I have the impression that you want to make quite a number of us unemployed…’.   This was a very understandable reaction if you think of what we were claiming.
It is only that reality is quite different: in real life the number of production systems grows fast and at the same time the complexity of the applications and systems dramatically increases. Also the field of competences of the DBAs extends largely.This is why the number of DBA does not shrink, only the typical work of a DBA changes a lot.It is exactly the same with tuning: the new possibilities just make things easier for us and also make new options possible. Nowadays it is almost impossible to manually tune complex SQL-statements for example in analytical contexts. Without the support of the system it would be very difficult to troubleshoot things, even for very experienced DBAs.To get back to the question: we will still have to tune only that we now have an even better and more enhanced support by the server. This will lead us to dramatically improved tuning results with at the same time much less effort we will have to make to get there.  To my mind there should be no Oracle Systems running without OEM-Diagnostic- and without EM-Tuning-Pack any more.

sysdba:

As you mention the OEM packages I would like to ask you a question about licensing policies: In the past there was a lot of fuzz about the OEM packages. I am talking especially about Diagnostics Pack, Tuning Pack and Change Management Pack.
These packages are only combinable with Enterprise Edition but they are available in all Oracle versions. It is only that we may not use them without a license for Enterprise Edition plus an additional license for each of these packages. Oracle itself uses them for self diagnostic tasks internally in all editions.
What many DBAs do not know is that you they these packages implicitly not only when they utilize certain wizards in the Enterprise Manager but also if they only query certain V$-Views or by running certain scripts  from $ORACLE_HOME/rdbms/admin or call some procedures with certain parameter values.
A lot of customers do not even know about these facts and I think that quite a number of them are permanently in danger of being under licensed by using the packages implicitly.  There is even a petition to Oracle available  in the web which has been joined by a large number of highly reputed people already.
Will the Enterprise Manager Packages be available for all editions in the future?

This topic has a long history of discussion and has been addresse also by us again and again. Actually there there is no plan to make these packages available also for SE oder SE1.

sysdba:

G.S.

‘Never touch a running system’ is what people always say after a change to the system has caused problems. But IT-systems are subject to permanent changes and those changes are not an exceptional case but the normal procedure! We permanently have to apply changes not only the operation systems but also to the database systems, the hardware configuration or the components of the infrastructure.At the same time the number of systems which are constraint by hard SLAs increases dramatically.
It is almost impossible for anyone to change a production system without extensive testing in front. The number of simulation tests after the implementation of changes to a database system has increases enormously in the past two years.
This is where Real Application Testing comes in very handy – which is a new 11g-option by the way. It enables us to run and analyze realistic workloads on a production system and re-run them on a new target platform without the need of additional software for the simulation. We can now compare and analyze the deltas for performance metrics for the same workload on different systems. Changes to critical  systems should never be allowed without testing under real conditions.

sysdba:

With respect to storage there are a lot of new features in Oracle 11g.
ASM can now handle multiple extent sizes not only for different datafiles, the extent size can even vary within files.
Also ASM can perform automatic defragmentation and open files much faster than in 10g.
Other 11g highlights are Preferred Mirror Readin order to prevent a permanent shifting of ASM extents across the interconnect in a  stretched-cluster and Fast Mirror Resync which is used to prevent that ASM disks get automatically gedropped if they are temporarily not availableWill we be able to store other file types in ASM in the future? I think of the Cluster Registry, Voting Disk, the Oracle binaries ($ORACLE_HOME) and maybe even other files. There have been rumours about a Posix File System in ASM around in the past.

G.S.

There will be a number of very interesting new features in the field of storage in the near future. ASM new features in 11g as well as other new trends about storage in general.
We will publish further details on these in the next weeks and months.
New storage features in 11g which are non ASM features are among others the compression-Option in 11g which will certainly be very interesting for many of our customers because now  compression is not only a feature for data warehouses any more but is usable for any application profile:
less disk usage, faster compression-algorithms, more efficient memory management, de-duplication for files stored inside the database, compression for RMAN-Backups and compression in Data Guard contexts make this option highly interesting for many customers.

sysdba:

Thank you very much for this interview Günther.
We are all looking forward to the results of the first real life tests

It contains two contributions from me:
- an interview with Oracle Vice President Guenther Stuerner from Germany about 11g which I have made on the 20th. aniversary of SOUG this year in Luzern (not posted here yet, but soon).
- an article about some new features for the Cost Based Optimizer in 11g which I have already published here on my blog a while ago.

 Hope you like it,

=,-)

I will deliver this class and
I am looking foreward to meeting you there.

=;-)

I think it will be a busy week and I hope to meet some people who had promised to come.

I will not deliver a speach about 11g Change Management as I had planned. Somehow the European Usergroup did not find the topic interesting enough. So let’s see what they have in place instead.

=;-)

So I will definitely deliver five days of Oracle 11g New Features for Adminstrators in:

Brussels October 15th. -19th

Istanbul October 22nd. – 26th.

Looking forward to seeing you there!

=;-)

The training takes place at the in the
  World Trade Center in Geneva
where Oracle Education still has some facilities.

At the end of the course Philipe,
 one of these hackers surprised me with

SIX different kinds of Belgium beer
which he had organized to convince me that there also is good beer
produced in his home country.
And I had to admitt that he was right which I did not know before!!!

So we had a little party at the end of the class…
I just love these little socializing events with my students when work is done!
This was also a good preparation for my 11g course in Brussels in three weeks.

It is titeled RMAN RECIPES and is published by APRESS.

 These authors have already proven their expert competency by publishing a lot of highly interesting white papers in OTN and other books in the past. (Just search for their names in OTN!).

Here are the details of the book:

Here are the contents as announced by APRESS:

Looks like asksteven.oracle.com

=;-) 

I will be online from time to time and then you can chat with me directly on my blog.
You are also free to enter my public chatroom

Also I have added archives and recent comments into the sidebar.

Hope you like it.

=;-)

One of the real highlights in Oracle 11g is the possibility to cache the results of queries as well as the return values of functions.

I want to discuss only the SQL Query Result Cache here and will post about the PL/SQL result cache later on.

Before Oracle 11g there was only one way to communicate between sessions – it was the built in package DBMS_PIPE we had to use.

Just imagine you are using a DB_LINK to read data from a remote database. Wouldn’t it be nice to have teh result for your query already in the cache over there instead of having to parse and cause physical reads from disk and all these things?

This is possible now with Oracle 11g!

As of Oracle 11gR1 it is possible to cache results of queries in a special section of the SHARED POOL called RESULT CACHE.

Let’s see how it works: 

There are three different modes you can use and they are managed with the initialization parameter RESULT_CACHE_MODE. This parameter is dynamic and can be changed with ALTER SYSTEM and ALTER SESSION. The default is MANUAL and it is also possible to use AUTO and FORCE. With force every result is stored in the result cache. With Auto the optimizer makes the descision what to cache.
And there are few more parameters which can be used to control caching of queries and memory utilization of the result cache of the instance.

 SYSTEM  @ orcl11 SQL> show parameter result 
NAME                                 TYPE                             VALUE
———————————— ——————————– ——————————
client_result_cache_lag              big integer                      3000c
lient_result_cache_size             big integer                      0
result_cache_max_result              integer                          5
result_cache_max_size                big integer                      768K
result_cache_mode                    string                           MANUAL
result_cache_remote_expiration       integer                          0 
 

Let’s have a look at the memory utilization in the SHARED POOL:  

SYSTEM  @ orcl11 SQL> SELECT *
     FROM  v$sgastat  2 
      WHERE lower(name) like ‘%result%’; 
POOL         NAME                            BYTES
———— ————————– ———-
shared pool  Result Cache: State Objs         2852
shared pool  Result Cache: Memory Mgr          200
shared pool  Result Cache: Chunk Ptrs           96
shared pool  Result Cache: Bloom Fltr         2056
shared pool  Result Cache: Cache Mgr          8304  

Let’s also have a look at a package which you can use to get a memory utilization report for the result cache and to flush the query results : 

SYSTEM  @ orcl11 SQL> desc dbms_result_cache
PROCEDURE BYPASS 
Argument Name                  Type                    In/Out Default? 
—————————— ———————– —— ——– 
BYPASS_MODE                    BOOLEAN                
INFUNCTION FLUSH RETURNS BOOLEAN 
Argument Name                  Type                    In/Out Default?
 —————————— ———————– —— ——– 
RETAINMEM                      BOOLEAN                 IN     DEFAULT 
RETAINSTA                      BOOLEAN                 IN     DEFAULT
PROCEDURE FLUSH 
Argument Name                  Type                    In/Out Default? 
—————————— ———————– —— ——– 
RETAINMEM                      BOOLEAN                 IN    
DEFAULT RETAINSTA                      BOOLEAN                 IN    
DEFAULT
FUNCTION INVALIDATE RETURNS NUMBER 
Argument Name                  Type                    In/Out Default?
 —————————— ———————– —— ——– 
OWNER                          VARCHAR2                IN 
NAME                           VARCHAR2                IN
PROCEDURE INVALIDATE 
Argument Name                  Type                    In/Out Default? 
—————————— ———————– —— ——– 
OWNER                          VARCHAR2                IN 
NAME                           VARCHAR2                IN
FUNCTION INVALIDATE RETURNS NUMBER 
Argument Name                  Type                    In/Out Default? 

—————————— ———————– —— ——– 
OBJECT_ID                      BINARY_INTEGER          IN
PROCEDURE INVALIDATE 
Argument Name                  Type                    In/Out Default? 
—————————— ———————– —— ——– 
OBJECT_ID                      BINARY_INTEGER          IN
FUNCTION INVALIDATE_OBJECT RETURNS NUMBER 
Argument Name                  Type                    In/Out Default?
 —————————— ———————– —— ——– 
ID                             BINARY_INTEGER          IN
PROCEDURE INVALIDATE_OBJECT 
Argument Name                  Type                    In/Out Default? 
—————————— ———————– —— ——– 
ID                             BINARY_INTEGER          IN
FUNCTION INVALIDATE_OBJECT RETURNS NUMBER 
Argument Name                  Type                    In/Out Default? 
—————————— ———————– —— ——– 
CACHE_ID                       VARCHAR2                IN
PROCEDURE INVALIDATE_OBJECT 
Argument Name                  Type                    In/Out Default? 
—————————— ———————– —— ——– 
CACHE_ID                       VARCHAR2                IN
PROCEDURE MEMORY_REPORT 
Argument Name                  Type                    In/Out Default?
 —————————— ———————– —— ——– 
DETAILED                       BOOLEAN                 IN     DEFAULT
FUNCTION STATUS RETURNS VARCHAR2    
 

SYSTEM  @ orcl11 SQL> set serveroutput on
SYSTEM  @ orcl11 SQL> exec dbms_result_cache.memory_report
R e s u l t   C a c h e   M e m o r y   R e p o r t[Parameters]
Block Size          = 1024 bytes
Maximum Cache Size  = 786432 bytes (768 blocks)
Maximum Result Size = 38912 bytes (38 blocks)[Memory]
Total Memory = 13508 bytes [0.010% of the Shared Pool]…
Fixed Memory = 10656 bytes [0.008% of the Shared Pool]…
State Object Pool = 2852 bytes [0.002% of the Shared Pool]…
Cache Memory = 0 bytes (0 blocks) [0.000% of the Shared Pool] 
PL/SQL procedure successfully completed. 

Also there are a number of dynamic performace views which can be used to monitor the result cache:

SYSTEM  @ orcl11 SQL> desc v$result_cache_objects

 Name                                                              Null?    Type 
—————————————————————– ——– ——————————————– 
ID                                     NUMBER TYPE                         VARCHAR2(10) 
STATUS                                                                               VARCHAR2(9) 
BUCKET_NO                                                                    NUMBER 
HASH                                                                              NUMBER 
NAME                                                                       VARCHAR2(128) 
NAMESPACE                                                                  VARCHAR(5) 
CREATION_TIMESTAMP                                              DATE 
CREATOR_UID                                                                NUMBER 
DEPEND_COUNT                                                               NUMBER 
BLOCK_COUNT                                                                NUMBER 
SCN                                                                        NUMBER 
COLUMN_COUNT                                                               NUMBER 
PIN_COUNT                                                                  NUMBER 
SCAN_COUNT                                                                NUMBER 
ROW_COUNT                                                                NUMBER 
ROW_SIZE_MAX                                                               NUMBER 
ROW_SIZE_MIN                                                              NUMBER 
ROW_SIZE_AVG                                                              NUMBER 
BUILD_TIME                                                                 NUMBER 
LRU_NUMBER                                                                 NUMBER 
OBJECT_NO                                                                  NUMBER 
INVALIDATIONS                                                              NUMBER 
SPACE_OVERHEAD                                                             NUMBER 
SPACE_UNUSED                                                               NUMBER 
CACHE_ID                                                                   VARCHAR2(93) 
CACHE_KEY                                                                  VARCHAR2(93)  

SYSTEM  @ orcl11 SQL> select * from v$result_cache_objects; 
no rows selected  

Now let’s cache a query result:
There is a new result_cache hint which you can use to manually cache a query result in odrder to override the default setting of the parameter in the session. 
(There also is a no_result_cache hint)
 

SYSTEM  @ orcl11 SQL> SELECT /*+ RESULT_CACHE */ 
2  count(*) , owner 
3   FROM lutz.t GROUP BY OWNER;  
COUNT(*) OWNER
———- ——————————        
4 LUTZ        12
TSMSYS     94256
PUBLIC        36
OUTLN        48
FLOWS_FILES         8
HR      2064
SYSTEM        32
ORACLE_OCM      1212
EXFSYS       220
DBSNMP     13136
SYSMAN     3136
XDB      6112
FLOWS_030000   
SYS      117224
WMSYS 1260  15 rows selected. 

#  let’s take a look at the explain plan!
# I just run the statement again

 SYSTEM  @ orcl11 SQL> set autotrace trace explain statistics
SYSTEM  @ orcl11 SQL> / 
15 rows selected.  
Execution Plan
———————————————————-
Plan hash value: 47235625 
————————————————————————————————–
| Id  | Operation           | Name                       | Rows  | Bytes | Cost (%CPU)| Time    
|————————————————————————————————–
|   0 | SELECT STATEMENT    |                            |   295K|  4907K|  1028   (7)| 00:00:13 |
|   1 |  RESULT CACHE       | 3dvq9p812snk55sb94f98ss805 |       |       |            |          |
|   2 |   HASH GROUP BY     |                            |   295K|  4907K|  1028   (7)| 00:00:13 |
|   3 |    TABLE ACCESS FULL| T                          |   295K|  4907K|   976   (2)| 00:00:12
|————————————————————————————————– 
Result Cache Information (identified by operation id):——————————————————   
1 – column-count=2; dependencies=(LUTZ.T); parameters=(nls); name=”SELECT /*+ RESULT_CACHE */count(*) , owner FROM lutz.t GROUP BY OWNER”  
Note—–  
- dynamic sampling used for this statement  

Statistics
———————————————————-         
5  recursive calls          
0  db block gets        
84  consistent gets         
0  physical reads         
0  redo size       
734  bytes sent via SQL*Net to client       
416  bytes received via SQL*Net from client         
2  SQL*Net roundtrips to/from client          
0  sorts (memory)         
0  sorts (disk)        
15  rows processed  SYSTEM  @ orcl11 SQL> set autotrace off 
SYSTEM  @ orcl11 SQL>  SELECT id, type, name, scn, row_count 
2* FROM  v$result_cache_objects;        
ID TYPE       NAME                                  SCN  ROW_COUNT
———- ———- —————————— ———- ———-         
0 Dependency LUTZ.T                            1267310          0        
1 Result     SELECT /*+ RESULT_CACHE */        1267310         15
                      count(*) , owner 
                      FROM lutz.t
                     GROUP BY OWNER  

SYSTEM  @ orcl11 SQL> desc v$result_cache_statistics

 Name                                                              Null?    Type 
—————————————————————– ——– ——————————————– ID                                                                         NUMBER 
NAME                                                                       VARCHAR2(128) 
VALUE                                                                      NUMBER 

SYSTEM  @ orcl11 SQL> SELECT * FROM v$result_cache_statistics ;        
ID NAME                                VALUE
———- —————————— ———-        

1 Block Size (Bytes)                   1024        
2 Block Count Maximum                   768        
3 Block Count Current                    32        
4 Result Size Maximum (Blocks)           38         
5 Create Count Success                    1        
6 Create Count Failure                    0        
7 Find Count                              1        
8 Invalidation Count                      0        
9 Delete Count Invalid                    0       
10 Delete Count Valid                      0 

10 rows selected.   

What I have described up to here is the server side query result cache.
If you are using an OCI application you can also configure a client side query result cache.

Again it is a set of parameters which are used to manage the OCI query result cache:

 SYSTEM  @ orcl11 SQL> show parameter client_result
NAME                                 TYPE                             VALUE
------------------------------------ -------------------------------- ------------------------------
client_result_cache_lag              big integer                      3000
client_result_cache_size             big integer                     

On the client side you can use an optional configuraton file to manage the client side settings in order to override the server side client settings.

When I arrived it had just stopped raining for a while and iI took the chance to stoll around a little and have a look at the surroundings.

 It is a nice city which also is very crouded with tourists.

In a nice passage
I found  a cinema with a hanging skulpture which I would like to know more about.
 Can anyone of you tell me what this is?

This week I have delivered my first
11g New Features Overview Seminar
for Oracle University in Switzerland. 

Here is my 11g teaching schedule so far:
- September 07th. Zurich (Overview)
- September 12th. in Prague (Overview)
- October 15th. Brussels ? (5 days hands on)
- October 22nd. Istanbul (5 days hands on)

For questions about these and other events
and bookings
please send me an email:
info@sysdba.ch

 =;-)

1. As of Oracle 10g the server collects optimizer statistics on all objects which do not have any object statistics in the data dictionary at all and for those object for which the statistics have become stale.

This is implemented by an automatically created scheduler job called GATHER_STATS_JOB which runs every night and every weekend by default and belongs to the AUTO_TASKS_JOB_CLASS. This mechanism has been changed in 11g by the way and we now use a different auto-task framework.Oracle uses table monitoring for this feature which is enabled by DEFAULT for all tables in 10g. This means that Oracle monitors all tables and for those where more than 10% of the values have changed since the last job run the statistics will get a STALE-flag in the data dictionary and the job run would compute new statistics for them. So far so good. A lot of manual work was removed the DBA shoulders in 10g but we had to find out that for some objects the threshold of 10% was suboptimal.This issue was addressed in Oracle 11gR1. Now we can set individual thresholds on table level with DBMS_STATS and there are new data dictionary views for these thresholds.

Let’s see how it works:

SYSTEM  @ orcl11 SQL> desc  dba_tab_stat_prefs

Name                                         Null?                  Type
—————————– ————– ——————-
OWNER                                   NOT NULL      VARCHAR2(30)
TABLE_NAME                      NOT NULL       VARCHAR2(30)
PREFERENCE_NAME                                    VARCHAR2(30)
PREFERENCE_VALUE                                   VARCHAR2(1000)

SYSTEM  @ orcl11 SQL> exec dbms_stats.set_table_prefs(‘SYSTEM’,'T’, ‘STALE_PERCENT’, ’15′)
PL/SQL procedure successfully completed. SYSTEM  @ orcl11 SQL>  select *
                                           2     from dba_tab_stat_prefs
                                          3     where owner=’SYSTEM’
                                          4       and table_name=’T'; OWNER   TABLE_NAME     PREFERENCE_NAME     REFERENCE_VALUE
———     ——————     —————————     ————————-
SYSTEM     T                                STALE_PERCENT              15     

2.  Another very nice new feature is deferred publication of newly gathered statistics. Now a DBA can control when new statistics are visible to the optimzer. This can be very handy for examply if you want to make sure that all statistics for all dependent objects have been collected and that not the new table statistics are used before the collection for the dependent indexes and partitions and subpartitions has been completed. They will be published all at the same time.Furthermore there is a procedure DBMS_STATS.RESUME_GATHER_STATS which can be used to catch up again, in case for some reason the statistics collection fails from comleting. This procedure continues exactly where the previouse run stopped.These new still unpublished statistics could be shipped to a testing system to test the impact on the workload before they would be published to the production system. The unpublished new statistics are stored in special tables and are called PENDING STATISTICS  in contrast to CURRENT STATISTICS which are the ones the optimzer can use.There are a lot of  *_PENDING_STATS views for the data dictionary.
After testing you can use DBMS_STATS.PUBLISH_PENDING_STATS to make the PENDING STATISTICS current.

3. The next highlight in 11g is MULTI COLUMN STATISTICS: now Oracle can gather statistics on groups of correlated columns and store them. This addresses a number issues which were around in 10g where Oracle had introduced the possibilty for the optimzer to take into account correlations for columns in a few limited cases. Jonathan Lewis had an entire chapter in his COST BASED ORACLE FUNDAMENTALS book. So let’s wait what the master of the optimizer will find out with his test cases about this! 

4. Incrementally updated global statistics on partitioned tables are the next feature I want to discuss here:Oracle 10g collects statistics at global- (table), partition- and sub-partition-level for partitioned tables. In other words table monitoring works on partitions as well and is enabled by default in 10g. Very good!The problem in 10g is that the collection of the global level statistics can cause a dramatic overhead because a full table scan is necessary while the partition level statistics are computed per partition and and are only gathered fpr those partitions in which more than 10% of the values have changed. In most cases only a few partitions undergo changes normally when we have a range base partitioning . The others are stable.In Oracle database 11g a new technology called synopses is used for certain global stsistics such as the number of distinct values. Oracle can now only scan those partitions which have been significantly modified and for the other partitions make use of the old stastistics in order to calculate the global stastistics on table level. These synopses need additional diskspace and are stored in the SYSAUX tablespace.  This is the drawback! We have additional I/O in advance of faster statistics collection.Histograms are not incrementally maintained. For more information see the online Performance Tuning Guide for 11g.

5. Last but not least let me mention EXTENDED STATISTICS on expressions:if you have a WHERE CLAUSE like this:
WHERE UPPER(last_name) = ‘King’
the optimizer uses 1% as the value for SELECTIVITY if there is no function based index available, which in many cases leads to the wrong descision and we end up with a bad execution plan. In 10g the optimzer could already handle some built in functions like TO_NUMBER.

As of 11g we can do this:

 exec DBMS_STATS.GATHER_TABLE_STATS(‘HR’, ‘EMPLOYEES’, METHOD_OPT => ‘for all columns size 1 columns (upper(last_name)) size 5‘)

In the Licensing Information Volume of the online documentation for Oracle Database 11g we now get the full information about the Management Packages. These packages are again fully installed by default. They are optional but extra cost options. In the 10g docu it was not possible to find out which functionalities use these packages implicitly and a lot of customers were under licensed without even knowing. For more information see my posts in the licensing section of the sidebar.

So far so good! Big improvement!

And here is the downside:

The packages are still only combinable with enterprise edition!

Recently I have made an interview with Guenther Stuerner, Vice President Oracle Germany, and asked him about Oracle’s policy with regard to the demand for the packages for all editions. He said that there are no plans to make them available for SE and SE1. The interview will be published in the next SOUG Newsletter.

These people come from all over the world to work for CERN and it was a pleasure to have them as students again.
Yesterday we had a nice dinner at an Italian restaurant nearby and anyway I think that we all had a lot of fun this week. Aint’t it so?

It was such a pity that I forgot to take a picture in time

  so this picture only shows a subset of the group since some had already left when I got my camera out of the bag.

At the end of the class they even surprised me with a present. I was really so much moved that I could not open it right away.
And when I came home in the night I found this:

It is a collection of five very well smelling Habana Cigars:
- a Montecristo
- a Cohiba
- a Romeo Y Julieta
- a Partagás
- and a Hoyo de Monterrey
You are so lovely, people! Thanks a lot. I will smoke one of them next week with my old neigbor who will be 92 on Friday and tell her about you!

Hope to see you again and have a nice weekend!

=;-)

See more details on the book here!

We are a team of four authors working on it. The others are:
- John Garmany, Steve Karam, and V.J. Jain

Table of Contents:

John Garmany

Steve Karam

V.J. Jain

Lutz Hartmann

Next week I will also be in Geneva at the World Trade Center and teach a bunch of DBAs working for CERN again.

This is going to be fun I guess. We will squeeze 10 days of normal training into five days which is really challenging since these guys use to be so advanced in dealing with really large amounts of data.

When CERN will start its experiments some time later this year they will have a throughput of terabytes/second comming in and their database will grow to a size of 2 digit petabytes fastly.

I am really looking foreward to having a look at the environment there.

=;-)

here

Oracle University is pleased to bring you a preview of

Oracle Database 11g

through three brand new courses given by

Lutz Hartmann – Oracle ACE 

 Oracle Database 11g: New Features Overview Seminar (1 day)

Oracle Database 11g: Change Management Overview Seminar (1 day)
or for more in depth insight plus hands-on

Oracle Database 11g: New Features for Administrators (5 days)

Please read the below outlines of these courses
and enroll directely from this post: 

• Managing storage
• Ensuring high availability
• Improving scalability and performance
• Strengthening security
• Managing data warehouses

The instructor lecture is supported by recorded product demonstrations. By attending this seminar, you can more effectively plan your use of Oracle Database 11g Release 1.  

Date	Location	Price	Enrollment
27 August 2007	Baden	SFR 850	enroll now
31 August 2007	Geneva	SFR 850	enroll now

Oracle Database 11g: Change Management Overview Seminar (1 day)

Although change can bring improvement, it also can produce disruption. This course focuses on Oracle Database 11g Release 1 innovations that enable you to mitigate the chaos that database and application upgrades can cause.You learn about planned features you can use to record production workloads and then play them back in a test environment so that you can determine and eliminate potential issues before deploying changes to your production environment.The instructor lecture is supported by recorded product demonstrations. By attending this course, you can more effectively realize the positive benefits of database and application upgrades while minimizing the negative side effects.  

Date	Location	Price	Enrollment
29 August 2007	Bern	SFR 850	enroll now

Oracle Database 11g: New Features for Administrators (five days) (5 days)
This course gives you the opportunity to learn about – and practice with – the planned new change-management features and other key enhancements in Oracle Database 11g Release 1.
You will examine the benefits and use of

• Planned new features in managing change
• Diagnosing and recovering from problems
• Ensuring high availability
• Improving scalability and performance
• Strengthening security
• Other areas that concern database administrators

The hands-on practice sessions help reinforce your understanding of these planned new capabilities in Oracle Database 11g Release 1 so that you will be ready to apply them when you return to your office.  

Date	Location	Price	Enrollment
17-21 September 2007	Baden	SFR 4250	enroll Now
24-28 September 2007	Geneva	SFR 4250	enroll Now
05-09 November 2007	Zürich	SFR 4250	enroll now
12-16 November 2007	Geneva	SFR 4250	enroll now
17-21 December 2007	Baden	SFR 4250	enroll now

On July 13th. John Leyden published a post on THEREGISTER in which Oracle U.K. is accused to host computers from which SSH brute force password attacks are launched against servers which run SSH software since May 3rd. 2007.

…Here is Oracle’s response as quoted on Register’s website:

“In response to Register inquiries, Oracle supplied a statement saying that an ongoing investigation is yet to confirm whether its systems have been misused or not.

“Security is a matter Oracle takes seriously and the company’s first priority is meeting customer needs and reducing their risk. As soon as Oracle became aware of the situation an investigation began, which is ongoing, but to date the company has found no evidence for any SSH brute-force attack originating from the Oracle owned machine currently listed on the DenyHosts website.” ®” …

=?-(

I cannot believe it!

Can you?

After talking about the SQL Analyzer (Sql Replay) in part I of this series I will continue with another Change Management functionality Oracle database 11g ships with. It is called Replay Database. 
This is one of the very big highlights of Oracle database 11g. And this is what many prodution DBAs have been hoping for since a long time.

Any change to a database system, such as an upgrade or changes to the configuration like new hardware or changed parameters, can have dramatic impacts on the performance of the application. It is necessary to test and validate these changed conditions under real life conditions in order to get around possible pitfalls before applying such changes to prodution. This is often a not easy to manage challenge for a DBA.

Oracle database 11g gives us the possiblity to capture a workload of a production system and run it in a testing environment. With this we can identify possible negative impacts of changes and analyze them under real conditions by comparing performance benchmarks for workload periodes in different systems.

How does it work?
On the prodution system we record all client activities for a typical workload periode with a new built in struture in the RDBMS. In fact it is a new built in package named DBMS_WORKLOAD_CAPTURE which is used . The captured worklaod consists of binary  files which hold metadata information for the workload and as well as replay files which are stored in a special directory. It is possible to use filters for certain session which should not be captured. The recording user as well as the replaying user needs at least SYSOPER privileges because it is necessary to restart the database before starting the capture process.

After capturing the workload data we need a testing system which should be axactely the same state as our prodution system before the capture process.  Now we start one or multiply replay clients (a binary called wrc) to run the pre processed workload against the testing system. Internally another new built in package named DBMS_WORKLOAD _REPLAY is use for this.

After the replay phase we can use analytic tools like ADDM to analyze the workload performance and compare it against teh original wokload performance.

There are a number of new DBA_WORKLOAD_* as well as V$WORKLOAD_* views that can be used to manage and manage the different steps (capture and replay).

Also Oracle Enterprise Manager comes with a very handy interface for the capture process as well as for the replay and analyze steps.

 Some good scenarios for database replay could be:
-  testing a migration from single instance to RAC
-  reproducing of bugs for debugging 

I will go for the Oracle Cerified Master challenge tomorrow and the day after tomorrow. This time Oracle will have a 10gR2 environment for the exam the first time. Until June 30th. Oracle had used 10gR1 for the master exams. So I will be the test bunny again I guess!

Hopefully the exam will not be cancelled again in the last minute. This is my third attempt to get an exam date from Oracle. Two times I had enrolled for the exam and it was cancelled.

   Right now I am relaxing at my hotel thinking some things over which might pop up as tasks in one of the exam sessions. After these two days of exam I will get back to my 11g new features series here, promised!

 I will first describe the new Features in CHANGE MANAGEMENT because I think that these are really revolutionary:

 1.  SQL Performance Analyzer (sql replay):
              With this feature we can predict the impact of system changes on a workload. Using this we can forecast changes in response times for SQL after a change. This can be very handy in cases like parameter changes, schema changes, hardware changes, os changes, db upgrades…Any change that influence sql plans is a good candidate for thîs feature.

How does it work?
 We use so called SQL TUNING SETS (STS) to hold workload information such as sql text, binds, execution plans plus statistics on executions. The STS is populated from the cursor cache over a certain periode of time. We capture teh increments for all these informations periodically over time. Only the desired statements are captured the rest of the workload is filtered out. After we have created a staging table which we can ship to a test system with the changed configuration using datapump, we can capture the desired workload and load it into the staging table. There we can unpack the STS from the staging table on the test system and run the workload there. Then we can compare the sql exection statistics of the runs for the workload on the two different systems.

We can also use the SQL Performance Analyzer to run the same workload on the same system multiple times wit different parameter settings, for instance with different optimzer versions enabled, in order to find out differences in excution plans after upgrades.

Sources for the workload to capture can, aside the cursor cache,  also be AWR snapshots and AWR baselines. 

The technology behind this feature is a new package called DBMS_SQLPA. Here is an overview for the procedures of this package: 

CANCEL_ANALYSIS_TASK Procedure Cancels the currently executing task analysis of one or more SQL statements
CREATE_ANALYSIS_TASK Function Creates an advisor task to process and analyze one or more SQL statements
DROP_ANALYSIS_TASK Procedure  Drops a SQL analysis task
EXECUTE_ANALYSIS_TASK Function & Procedure Executes a previously created analysis task
INTERRUPT_ANALYSIS_TASK Procedure Interrupts the currently executing analysis task
REPORT_ANALYSIS_TASK Function Displays the results of an analysis task
RESET_ANALYSIS_TASK Procedure Resets the currently executing analysis task to its initial state
RESUME_ANALYSIS_TASK Procedure Resumes a previously interrupted analysis task that was created to process a SQL tuning set.
SET_ANALYSIS_TASK_PARAMETER Procedures Sets the SQL analysis task parameter value
SET_ANALYSIS_DEFAULT_PARAMETER Procedure Sets the SQL analysis task parameter default value

There is an interface to the SQL Performance Analyzer in the enterprise manager in the Advisor Central. 
A number of new to DBA_ADVISOR_* views have been added in 11g which we can use to display information from the SQL Performance Advisor. Also a number of DBA_ADVISOR_* views which were alredy available in 10g have new columns, such as DBA_ADVISOR_TASKS and DBA_ADVISOR_FINDINGS

=;-)

Hope to see you in one of my 11g seminars

A number of interesting speakers were scheduled for laudatio speeches such as Nenad Rozic, Sr. Director EMEA Server Technologies & Global DBA support. I had a very interesting chat with him about quality Oracle’s support and the management cultur within the company.

Also very interesting was the speech of Gunter Dueck, an top manager of IBM who talked about innovation processe in the industry and showed in a very easy to understand way what the blocking factors for innovation within large companies are. He is very impressing personality and has very interesting ideas about the cultur in large companies. I think that I will some of his books since I have the impression that I understand what he wants to say and that it is important to change something about the culture in management.

The most interesting speaker from the technical point of view was
Guenter Stuerner, the head of Oracle pre-sales in Germany.

He talked about Oracle 10g Time Ten database and 11g New Features.
and also showed some interesting old photographs from the first days of Oracle,
this one here shows

Larry Ellison, Bruce Scott (the master of a cat named TIGER), Bob Miner, Ed Oates even before Oracle was named Oracle.

Also I met some of my customers

whome I hope to see again in one of my new seminars and trainings.

Today I received this below mesage from Oracle University Switzerland
and I am happy to post it here just slightly adjusted:

Oracle University is pleased to bring you
Oracle Database 11g New Features with Lutz Hartmann – Oracle ACE
Oracle Database 11g: New Features Overview Seminar (1 day)

Date	Location	Price	Enrolment
27 August 2007	Baden	SFR 850	Enrol Now
29 August 2007	Digicomp Bern	SFR 850	Enrol Now
31 August 2007	Digicomp Geneva	SFR 850	Enrol Now

In this seminar, you are introduced to the planned key new capabilities of Oracle Database 11g Release 1 and gain an understanding of how to take advantage of them. Topic areas covered include new features for
· Managing storage
· Ensuring high availability
· Improving scalability and performance
· Strengthening security
· Managing data warehouses The instructor lecture is supported by recorded product demonstrations.
By attending this seminar, you can more effectively plan your use of Oracle Database 11g Release 1. or for more indepth insight plus hands-on
Oracle Database 11g: New Features for Administrators (5 days)

Date	Location	Price	Enrolment
24-28 September 2007	Geneva	SFR 4250	Enrol Now
05-09 November 2007	Zürich	SFR 4250	Enrol Now
12-16 November 2007	Geneva	SFR 4250	Enrol Now
17-21 December 2007	Baden	SFR 4250	Enrol Now

This course gives you the opportunity to learn about – and practice with -
the planned new change-management features and other key enhancements in Oracle Database 11g Release 1.You will examine the benefits and use of
· Planned new features in managing change
· Diagnosing and recovering from problems
· Ensuring high availability
· Improving scalability and performance
· Strengthening security
· Other areas that concern database administrators The hands-on practice sessions help reinforce your understanding of these planned new capabilities in Oracle Database 11g Release 1 so that you will be ready to apply them when you return to your office. » connect Lutz Hartmann as sysdba

I had always wanted to ride in a Stretch Limo 
   and it was lots of fun!

We went to the Golden Gate Bridge …to the Coit tower  from where you get a beautiful view over the city  .. we dashed through
China Town walked Lombard St. ..

and at the end of the tour we went to OLDEST ITALIAN RESTAURANT OF AMERICA the Fior d’Italia which is a must for every lover of Italian cusine who stays in San Francisco! Altough there are some Italians who say that it is impossible to eat pasta outside Italy. I would recommend them to try it. I had my best Saltimbocca ever there. It was just devine!

After dinner we all went downtown again with the obligatory   cable car. It was so much fun and I realy miss some of these people aready.

And there is not only the paper, but “…THC further releases practical tools to sniff and crack the password of an oracle database within seconds…” all you need to know is how to use them.

 Yesterday Hansen an Oracle instructor from Shanghai
explained the
Chinese  Oracle logo   Jia Gu Wen
to Jean-Francois.
He really gave his best to get the message through to him  …but

I am not sure if everything really got through to J.F.  ..he seemed to be a little puzzled to me!

 As you all know there is an auto task job which is automatically created with every Oracle 10g database called gather_stats_job which automatically collects statistics every night and on the weekend by default for all object for which more of 10% of the values have changed. This behaviour is enabled by default in 10g and is implemented by TABLE MONITORING which is enabled for all tables by default.

In Oracle 10g the 10% threshold is a hard coded value and cannot be adjusted. This is a particularly painful with columns with skewed data in large tables.  Here you would have to lock the statistics and gather them manually when needed with an adjusted sample size of probably more than 10%, because otherwise you would get terrible execution plans possibly. 

 In Oracle database 11g there is a new procedure in the DBMS_STATS package we can use to even set individual thresholds on segment level.
Here is  how it works:

lutz@ora11gBETA5 _SQL > exec dbms_stats.set_table_prefs(‘HR’, EMPS’, ‘STALE_PERCENT’, ’15′)

The previous statement adjusts the threshold for the table EMPS owned by HR to 15%.  

All togehter there are three new parameters in DBMS_STATS:
- STALE_PERCENT which overwrites the default threshold of 10%
- INCREMENTAL for statistics on partitions incrementally gathered
- PUBLISH in order to populate the data dictionary with the newly gathered statistics or not, whcih can conme in handy if you want to test with the new statistics on another system before really making them effective on production.

I think that this is exactely what a lot of people have been waiting for since a long time

On Sunday I took a day off:
no burea, no Oracle, no telefone conferences

… and walked along the bay  …where I found these sea lions relaxing on pontons in the harbour. I found out that they first apeard after directly the earth quake. Their populatin has grown up to about 300 int the summer and about 600 during the winter time.

This week I will take part here in the first
Oracle 11 New Features for Administrators course ever

 with my article on
Fast incremental backups with block change tracking in Oracle 10g.
This article is also available on my blog.

I am not so happy with this publication since it has been quite a while that I had writen it and I have modified some things in between.

This version of the article has a misleading formulation in it about incremental levels in 10g. In 10g it is still possible to make incrementals of multiple levels, but Oracle only uses incremental levels 0 & 1 for the suggested backup strategy.  But the incremental level 1-4 are still available with RMAN!

I will travel to San Francisco next week and take part in the first pilot course at Oracle University and hopefully deliver the first 11g courses sometime not very far in the future.
I will let you know about details ASAP!

  COUNT(*) ———-       1227

SQL> create table t2 (col1 number);

Table created.

SQL> select count(*) from tab$;

  COUNT(*) ———-       1228

SQL> select NUM_ROWS from dba_tables where table_name=’TAB$’;

  NUM_ROWS ———-       1213

SQL> exec dbms_stats.gather_database_stats;

PL/SQL procedure successfully completed.

SQL> select NUM_ROWS from dba_tables where table_name=’TAB$’;   NUM_ROWS ———-       1228 – IT DOES! – and now let’s see if the job does also: SQL> create table t3 (col1 number);

Table created.

SQL> create table t4 (col1 number);

Table created.

SQL> select NUM_ROWS from dba_tables where table_name=’TAB$’;

  NUM_ROWS ———-       1228

– gather_stats_job run manually from DATABASE CONTROL !!! SQL> select NUM_ROWS from dba_tables where table_name=’TAB$’;

  NUM_ROWS ———-       1230

– and IT ALSO DOES! — even though there were not even 0.1% of the values changed it did! So when should we gahter statistics for the data dictionary manually? Oracle recommends to collect them when a significat nnumber of changes were applied to the data dictionary, like dropping significant numbers of partions and creating new ones dropping tables, indexes, creating new ones and so on. But this only if it is a signifcant number of changes and you cannot wait for the next automatically scheduled job run.

I felt like I was newly born afterwards.

 After this I took a lst coffee at the Mandabatmaz Kahvesi
which is an absolute MUST for every fan of sofisticated coffee.
    I just had to congratulate the kahveçi for his brilliant coffee. The bull will not sink in the cup, this is for sure! One more very good reason to come back definitely!

I introduced them to the infrastructure of ASM and explained Oracle’s unique locking concept.

After this  Tonguç and his brothers invited me for a drink and dinner downtown and we had some intersting discussions.

One interesting questions we talked about was what happens exactly with a SHUTDOWN IMMEDIATE – is there a checkpoint before shutdown or is it just a flushing of the redolog buffer to the logfiles?

Here is the proof for the CHECKPOINT happening before shutdown:

SYS AS SYSDBA @ lutz1 SQL> create table t as select * from all_objects;

Table created.

SYS AS SYSDBA @ lutz1 SQL> select current_scn from v$database;

CURRENT_SCN
———–
     202799

SYS AS SYSDBA @ lutz1 SQL> select file#, checkpoint_change# from v$datafile;

     FILE# CHECKPOINT_CHANGE#
———- ——————
         1             197655
         2             197655
         3             197655
         4             197655

SYS AS SYSDBA @ lutz1 SQL> select file#, checkpoint_change# from v$datafile_header;

     FILE# CHECKPOINT_CHANGE#
———- ——————
         1             197655
         2             197655
         3             197655
         4             197655

SYS AS SYSDBA @ lutz1 SQL> shutdown immediate
Database closed.
Database dismounted.
ORACLE instance shut down.
SYS AS SYSDBA @ lutz1 SQL> startup mount
ORACLE instance started.

Total System Global Area  507510784 bytes
Fixed Size                  1220264 bytes
Variable Size             155189592 bytes
Database Buffers          348127232 bytes
Redo Buffers                2973696 bytes
Database mounted.
SYS AS SYSDBA @ lutz1 SQL> select file#, checkpoint_change# from v$datafile_header;

     FILE# CHECKPOINT_CHANGE#
———- ——————
         1             203071
         2             203071
         3             203071
         4             203071

SYS AS SYSDBA @ lutz1 SQL> alter database open;

Database altered.

SYS AS SYSDBA @ lutz1 SQL>  select file#, checkpoint_change# from v$datafile_header;

     FILE# CHECKPOINT_CHANGE#
———- ——————
         1             203072
         2             203072
         3             203072
         4             203072

SYS AS SYSDBA @ lutz1 SQL>

#######here is the content of the alertlog for the startup
####### as you can see there is no instance recovery!!!
####### there was a checkpoint before shutdown!!

Completed: ALTER DATABASE   MOUNT
Fri May 18 01:13:36 2007
alter database open
Fri May 18 01:13:36 2007
Thread 1 opened at log sequence 12
  Current log# 3 seq# 12 mem# 0: +DATA/lutz1/onlinelog/group_3.286.622819249
  Current log# 3 seq# 12 mem# 1: +FRA/lutz1/onlinelog/group_3.259.622819253
Successful open of redo thread 1
Fri May 18 01:13:36 2007
MTTR advisory is disabled because FAST_START_MTTR_TARGET is not set
Fri May 18 01:13:36 2007
SMON: enabling cache recovery
Fri May 18 01:13:38 2007
Successfully onlined Undo Tablespace 1.
Fri May 18 01:13:38 2007
SMON: enabling tx recovery
Fri May 18 01:13:38 2007
Database Characterset is WE8ISO8859P1
replication_dependency_tracking turned off (no async multimaster replication found)
Starting background process QMNC
QMNC started with pid=18, OS id=4284
Fri May 18 01:13:44 2007
Completed: alter database open

     in Istanbul

I would like to say thank you very much to Leyla Alpaslan & Gökhan Uğuroğlu
from Oracle University Turkey for inviting me to Istanbul and organizing this event.
I had not expected to find such such a well organized event in such a brilliant location
- they had produces  real Oracle Courseware of my sysdba slides  

- there was always coffee and deliciouse paistries available
It was only that at the end it got a little hot in the room, but this was my fault because I had told the hotel staff that it is too cold instead of too warm in Turkish.
I must polish up my Turkish again. Forgot so much over the years. 

As you might know a PL/SQL procedure runs with the privileges of the developer by default which has ever been this way.

Since Oracle 8i it is possible to use AUTHID CURRENT_USER to create programs runnig with the AUTHORITY of the USER of the program.

Steven demonstrated  how a procedure A running under AUTHORITY of the AUTHOR calls another procedure B running under AUTHORITY of the USER, which again calls a procedure C, running under AUTHORITY of the AUTHOR again.

Strangely enough we learned that once procedure B is invoked also procedure C uses the privileges of the CURRENT USER, because CURRENT USER “is set to the owner of the definer rights program..”
.. which is in other words, that as soon as you run a program running with USERS s AUTHORITY calling other programs running with the DEFAULT ,AUTHOR’s rights, you can end up with surprises waiting for you there.

When he was waiting in th lobby of the hotel he ran into a comrade he had studied togeter with and whom he had not met since 10 years. Both of them were very surprised to meet each other after such a long time, standing in front of the lift waiting for an Oracle guy to come down. And the best thing was that they were waiting for different people, one for me and the other one for Paul Dorsey, an Oracle expert from Iselin NJ in the U.S.
What an incident! Two friends meet in a hotel by accident ,
both waiting in the same hotel in front of a lift for two different people,
who also are both linked with Oracle!

 Istanbul must be a village I thought! Incredible!

We first went to Dolmabahçe, from where we took a taxi to Taksim place, which is the center of the European side of Istanbul. We strolled down Istiklal Caddesi until we came to location of the German Goethe Institut, we a very fancy restaurant is also located on the roof of the building.
It is called Litera , from where you get a
wounderful view 
across the Golden Horn to the old city of Istanbul.
We had a cold drink and after this we went down to a tunnel from where we took a kind of tram to Karaköy from where we took a ferry to the Anatolian side of Istanbul.
Tonguç showed me the train station HaydarpaŞa
from where the famous Orient Express starts all the way to Teheran,
and explained geography, history and strategic situation of Istanbul to me.  

On the way to the busstop we needed Tonguç asked a taxi driver for directions and to my surprize he said ” …you are walking into the wrong direction. Jump right into my taxi and I will give you a free ride to the busstop”, can you believe this?  
We took a middibus to Fenerbahçe, passing along the footballstadion.
Tonight is an historical football match and if the Fenerbahçe team wins against Trabzonspor there will be large party going on in all the town, because the team will be winner of the championship and this in their 100th. year! 
In Fenerbahçe I had tavuk göğsü, a very special Turkish desert which I like very much. And I guess that Tonguç took me all the way to very best place to have it.

Tonguç,  you are a very good tourist guide!
And I appreciated it very much that you spent your valuable time at the weekend to introduce me to your city!
Thank you very much for the wonderfull day!

I always like to find out what normal life is like in other countries,
not only what is shown to tourists .
I experienced  that Istanbul people are generally very friendly and kind people.

I had a first kebab  which was very good.

My hotel is on the European side of the Bosporus.
After my seminar I will have a look at the oriental side of the city.