Uncommitted transactions

Similar Questions

• Transaction in stored procedure not treated as Atomic?

  Hello

  I have a stored procedure that inserts a line in a table called reservations and also decrement the places available for an event.

  For example: at an event, there are 100 places available. After you add a reservation, it will be left only 99 seats.

  I also have a Java program that sends 100 concurrent transactions to the database.

  After 100 transactions, there are 100 reservations, but there are still about 60 free places.

  So, I guess that the transaction he has not treated as atomic. I don't know why some transactions fail?

  The code of the stored procedure:

  create or replace PROCEDURE procedureOne (s_price, NUMBER, NUMBER, of s_customer)

  s_event NUMBER, DATE, s_movie NUMBER of s_dat)

  IS

  s_ev NUMBER;

  s_free NUMBER: = 0;

  BEGIN

  commit;

  Set transaction read write;

  Start

  SET TRANSACTION ISOLATION LEVEL READ COMMITTED;

  Select Event_ID, Free_seats EN s_ev, s_free

  events

  where ID_Movie = s_movie;

  If s_free > 0 then

  Start

  update events set Free_seats = s_free-1;

  insert into values of reservations (seqa_seq. NEXTVAL, s_dat, s_customer, s_event, s_price);

  commit;

  end;

  on the other

  Start

  Rollback;

  end;

  end if;

  end;

  commit;

  end procedureOne;

  invatacelul wrote:

  So, I guess that the transaction he has not treated as atomic. I don't know why some transactions fail?

  You said: "I also have a Java program that sends 100 concurrent transactions to the database. Atomicity applies to a single transaction - it says transaction executes all or nothing. Is what you run in isolation. If you want isolation unless you serialize, transaction 2 sees no uncommitted transaction 1 results because the READ COMMITTED isolation level indicates that a transaction can read only the data that has been committed to the database. And since you wrote your code as SELECT + UPDATE you get results you never expected. Use:

  UPDATE OF EVENTS

  SET FREE_SEATS = FREE_SEATS - 1

  WHERE ID_Movie = s_movie

  AND FREE_SEATS > 0;

  SY.

• Restore the transaction committed

  I am trying to execute the procedure in Oracle 12 c using SQL Developer

  create table Parent
  (
  ID number(5),
  Name Varchar2(50)
  );
  create table Child
  (
  ID number(5),
  Name Varchar2(50)
  );
  Procedure ParentProc
  begin
  //Doing the Transactions
  insert into Parent values(1,'Parent');
  //calling child proc
  ChildProc;
  rollback;
  end ParentProc;
  
  
  
  
  Procedure ChildProc
  begin
  //Doing the Transactions
  insert into child values(1,'Child');
  commit;
  end ParentProc;
  

  
  

  The value in the child table persists even after that the restoration is used in the parent process.

  There is a solution where the validation of the procedure of the child can be eliminated, but as child procedure is called by other sources also directly, if validation is necessary here.

  And also, I don't want to use a variable to separate he was called by any other process, or if called directly.

  I just want to know if I use the valid approach, I am submitting commit or rollback of the calling program

  that I learned from this https://blogs.oracle.com/dev2dev/entry/write_recovery_code_with_transaction

  This article shows everything we said.

  1. the DML is run by the customer

  2. then the command also committed by the customer if the DML succeeded

  Just a question will be the oracle database GET two calls from the calling program

  (1) the first Call When I Sumit the query.

  (2) second call when I submit Commit / Rollback.

  Yes - there are two statements and each of them will be sent to the server separately.

  During this period the uncommitted transaction will be lock on the tables that are in the DML statements. or it is a hidden life?

  Yes - No validated DML may have line and/or table locks. These locks will not come UNTIL the end of transactions - that is, until a commit or rollback is performed.

• Capture Capture stream contains transactions and spills

  I wanted to clarify something on the subject of capture streams. I saw a weird behavior that broadcasts continuously if I do a big update on a huge table, then roll back the transaction, starts to spread the settlement transaction. My question is why is stream capture messages and the reversals of them when they are not committed? Is there any config setting I'm missing?
  
  Thank you
  Kevin

  An uncommitted transaction in a streams environment is also well captured and propagated but not enforced. The settlement does not apply to a transaction to stop. In case of a cancellation of a transaction, the capture process creates ADR for do it again generated and place these ADR according to the rules specified for the capture process. The ADR are propagated to the destination site according to the propagation rules. Apply it treat assmebles the ADR in a transaction and then applies them in the correct order when the COMMIT message is received. For transactions ending in a restoration, imputation does not apply the changes in the ADR for this transaction.

• necessary to enhance Network.http.Max - connections-per-server to javascript

  I have an AJAX web application which is very sensitive the number of connections allowed by the browser. Bad things happen when the limit is set at the bottom: frozen screens, the uncommitted transactions, deadlocks, etc..

  To avoid this, I need to check the current value of the configuration parameter "Network.http.Max - connections-per-server" program (i.e. javascript code) and to reject a request if the limit is too low with the instructions appropriate for the user.

  I guess that's possible because "subject: config" restores all values of settings and even allows for user to change. However, the code-behind for the page is very complicated, so I can not understand how to accomplish my task. All I need is a function or a method returns the current value of a given configuration parameter.

  Any help will be greatly appreciated.

  Anatoly

  You do not have access to the pref the subject: config page via a Web site.
  
  In theory it is possible, but the user get a pop-up and must confirm each time.

  See also http://www.mozilla.org/projects/security/components/signed-scripts.html

• Restore SQL DB with AppAssure

  Has anyone tried to restore a database using AppAssure replacing everything simply?  I find a lot of documentation on a separate copy and then using SQL to copy on the tables or lines from the backup, but I just want to restore a copy of the database from a recovery point in AppAssure. My experience with SQL is limited, so I'll try to do it is as easy as possible.   The database is a small database of MS Dynamics SL and is only about 2G in size. It is not used after that hours so it seems that I could simply detach and move the DB direct current and attach the copy of the .mdf file backup and recovery point .ldf AppAssure.   Will this work?  Am I missing something?

  Thanks for any input.

  Yes, you want both the database and the log. Attach the database while creating the new journal of a.a. is a solution of "last resort", as the journal may contain uncommitted transactions.

• Why do we need to restore?

  Hello

  I use plain jdbc and connection of Mattie pool. I mean I use the data source. In a try block I put autocommit to false and after executing the update statement, I commit the transaction by Connection.Commit (), but I do not write back any roll in the catch block. My intention is that if an exception occurs it must work because it is not committed. So why should we declare rollback in the catch block? Is - mandatory to do? Is there any problem if we do not?

  My intention is that if an exception occurs it must work because it is not committed.

  Then you should put a ROLLBACK so that your 'intent' becomes a reality to the EXACT PLACE, he should be here.

   So why should we declare rollback in catch block?
  

  Because it is the EXACT POINT in time when we know with certainty that a RESTORE is the proper thing to do.

  Your code must perform EXPLICITLY all the necessary actions for the realization of your intention. You should NEVER let Oracle (any DB) or Java implicit operations.

  Is - mandatory to do?

  No - it is NOT mandatory.

  Is there any problem if we do not?

  Yes - it there will BE problems - if someone needs to maintain, correct or improve your code later.

  With no explicit ROLLBACK the restoration will occur ONLY when the connection object is closed or destroyed. That may not happen until your application has actually ends then leaves Java.

  During this time your uncommitted transaction may have LOCKED data in the database, which prevents other users to work with these data. You are also always other database resources (e.g. space temp, memory) that you really do need more.

  BEST PRACTICE is to ALWAYS release resources when you need is no longer. Which includes using a ROLLBACK as soon as you know that is what is needed.

• The validation process hung for a table only!

  Hi all

  There is a table contains 79 columns (with a cloumn CLOB) with about 100 000 record, sometimes a problem occurs only with this table (and only these days) when you attempt to insert, update, delete from this table, the validation process hung! Why? and how to solve? I can't say why this problem occurs.

  We renamed the table and re-create using EXPDP, IMPDP orders, re-create triggers on it and repeat grants thereon to the roles that the previous, this procedure has solved the problem for maybe one day only, but the problem returns!

  Notes:

  This table is on a table space users, otherwise I use DB Oracle 11 g R2

  > The question is: What does it mean the blocking locks? and why this problem occurs? and how to avoid it from happening again?

  Have you read this:

  https://docs.Oracle.com/CD/E18283_01/server.112/e16508/consist.htm

  For example, this part:

  Written in conflict read-committed Transactions

  In a read committed transaction, a write conflict occurs when the transaction attempts to change an update by a simultaneous uncommitted transaction line, sometimes called a blocking transaction.

  The read committed transaction awaits the transaction that blocks at the end and release his lock of line. The options are the following:

  -If the blocking transaction is cancelled, the pending transaction then proceeds to change the locked row previously as if the other transaction never existed.

  -If the blocking transaction is committed and frees its locks, then the transaction pending proceeds to update scheduled for the newly modified line.

  > Where can I find this log file please?

  For example:

  C:\app\oracle\diag\rdbms\oracle\orcl\trace\alert_orcl.log

  Kind regards

  Zlatko

• Period close 11i and R12

  Dear all,

  Latest version of Oracle Applications: 11.5.10.2

  We are in 11i and 11i pertaining to the close period limits is as below:

  All business units that share a set of books also share the same period status. When we update the State time to open in an operating unit, this period is open for all units operating within the whole books. When attempting to update the State closed in an operating unit, the system checks the uncommitted transactions in all units operating within all of the books. If there are uncommitted transactions in one of the operating units, you are asked to solve for these before operating units uncommitted transactions can close the period for all the books.

  At the above limitation is here in R12, if yes could you please explain in detail.

  Thanks and greetings

  Srinivas

  Hi Srinivas,

  This restriction exists as well in R12. Periods gl are always defined by all of the books, so, if you open (or close) the period within an organization that uses all of the books, it is opened (or closed) in all agencies as well.

  Cheryl

• What should I Actions she undertakes to buffer log regarding

  Hello Experts,

  When a user commit its transaction, the complete log buffer (which can contain uncommitted transactions, other users) will also push to log online?

  Or just in the entrance of himself?

  Thank you and best regards,

  Tong Ning

  Hello

  Yes.All journal the log buffer flused online. Also look at the following link

  https://asktom.Oracle.com/pls/Apex/f?p=100:11:0:P11_QUESTION_ID:621023586146

  Yasin.

• daily batch

  If I have a daily prizes that need to pick up the records that have been modified in the last 24 hours and proceed with a treatment, there is the issue of timing.

  If I run:

  insert into TARGET_TABLE select * From SOURCE_TABLE where MODIFIED_DATE >= to_date(to_char(sysdate - 1, 'YYYYMMDD') || '070000', 'YYYYMMDDHHMISS')
  and MODIFIED_DATE < to_date(to_char(sysdate , 'YYYYMMDD') || '070000', 'YYYYMMDDHHMISS')
  

  If I have a column MODIFIED_DATE which is updated by a trigger whenever there is a change, how can I make sure I pick up all the records? Because of the consistency of reading and a transaction may take a few minutes to commit a file can get the updated MODIFIED_DATE up-to-date in the range above after the statement above is executed, so it won't be picked up. How to solve the question of the coherence of reading?

  Can expand you on the two options above?

  Course

  1 accept the risk

  Many orgs do excerpts at the end of the morning (for example 01:00 to 03:00) when it WOULD normally not watch uncommitted transactions. Their extract can also occur AFTER other maintenance operations who have already dealt with uncommitted transactions by ending sessions that have them.

  In these situations, it is would have little risk of DML uncommitted could cause a problem of extract.

  For an example of a situation where there is a problem using a MODIFIED_DATE value to retrieve data see my last reply in this thread from April

  https://forums.Oracle.com/thread/2527467

  Is Oracle, possible to check which are uncommitted by table transactions?

  Ask the view DBA_DML_LOCKS (and/or view DDL)

  http://docs.Oracle.com/CD/E11882_01/server.112/e25513/statviews_3157.htm

  Select * from DBA_DML_LOCKS

  SESSION_ID, OWNER, NAME, MODE_HELD, MODE_REQUESTED, LAST_CONVERT, BLOCKING_OTHERS

  10, SCOTT, EMP_READ_ONLY, Row - X (SX), no, 1163, does not

  69, SCOTT, EMP_READ_ONLY, Row - X (SX), no, 1648, does not

  This shows that two session held in exclusive line on the EMP_READ_ONLY TABLE lock. As a test, I inserted a line from two different sessions, but he has NOT yet committed.

  So when you do not want to accept the risk of putting you the table in READ-ONLY mode, perform extract and then put the table in READ WRITE mode.

  ALTER TABLE EMP_READ_ONLY IN READ ONLY

  You will get an exception if there is remaining DML operations. You can find them and deal with them.

  How would a materialized view using would solve the issue of the calendar? How that would solve the question of an operation which began, let's say at 06:58 and gets committed the day before at 07:04. If I update the view materialized to 7 today, the record would not be captured because at 7, the transaction is not committed. I have the same problem? I don't know I'm missing something, please provide details.

  I said usage log a MV. I did not use a materializled view.

  You can create a log of MV on the table so that

  Oracle will save all changes on the table for the NEWSPAPER.

  See my response detailed, dated May 23, 2013 13:37 in this thread

  https://forums.Oracle.com/forums/thread.jspa?MessageID=11034366�

  My answer includes examples of code and results that show the information captured in a newspaper of the MV and how you can use it.

  If possible, you must use LogMiner to simply capture changes from the REDO log files.

• A question about ROLLBACK and in stored subprograms unhandled Exceptions

  Hello

  Oracle version:  Enterprise Edition Release 11.2.0.1.0 - 64 bit

  OS :   Linux Fedora Core 17 (X86_64)

  In Chapter 11 of the PL/SQL online, here's what oracle says about The unhandled Exceptions :

  If a stored subprogram stops with an unhandled exception, PL/SQL does not restore of database changes made by the subprogramme.

  I tried the following code:

  code

  HOST clear; SET SQLBLANKLINES ALTER SESSION SET PLSQL_WARNINGS = ' ENABLE: ALL '; DROP TABLE tmptab; CREATE TABLE tmptab (n NUMBER); CREATE or REPLACE PACKAGE pkg1_pkg DEFINE AUTHID AS EXCEPTION MYEXCEPTION. PRAGMA EXCEPTION_INIT (MYEXCEPTION,-20001); END; / DISPLAY ERRORS; -It is a subroutine that exists -with an exception unhandled CREATE OR REPLACE PROCEDURE mytestProc DEFINE AUTHID IS BEGIN INSERT INTO tmptab (n) VALUES (100); -Unhandled exception does a ROLLBACK -which annuls the previous INSERT statement

  RAISE pkg1_pkg. MYEXCEPTION;

  END mytestProc;

  /

  DISPLAY ERRORS;

  
  

  
  

  BEGIN

  mytestProc();

  END;

  /

  
  

  
  

  SELECT * FROM tmptab;

  And here is the result

  BEGIN * ERROR on line 1: ORA-20001: ORA-06512: at "TRAINING. MYTESTPROC', line 6 ORA-06512: at line 2 no selected line

  The fact that SELECT * FROM tmptab gives: no selected line indicates that the exception unhandled in the stored routine myTestProc is doing a ROLLBACK for the INSERT statement in the tmptab table.

  So I do not understand why in the documentation noted that no ROLLBACK is not done.

  Could someone kindly do some clarification on this?

  Thanks in advance,

  Kind regards

  Dariyoosh

  Hello

  If a stored subprogram exits with an unhandled exception, PL/SQL does not restore changes to the database made by the subprogramme.

  What is happening here, it is an anonymous block (which is not stored) stops with an unhandled exception, and who restores the uncommitted transactions.

  I created a procedure to see how many lines is in tmptab:

  CREATE OR REPLACE PROCEDURE tmptab_cnt (in_txt IN VARCHAR2)
  DEFINE AUTHID
  AS
  n PLS_INTEGER;
  BEGIN
  SELECT COUNT (*)
  N
  OF tmptab;

  dbms_output.put_line (n
  || "lines of tmptab in.
  || in_txt
  );
  END tmptab_cnt;
  /

  and called it of your procedure with your anonymous block:

  CREATE OR REPLACE PROCEDURE mytestProc
  DEFINE AUTHID
  IS
  BEGIN
  INSERT INTO tmptab (n) VALUES (100);
  tmptab_cnt ("mytestProc");

  RAISE pkg1_pkg. MYEXCEPTION;
  END mytestProc;
  /
  DISPLAY ERRORS;

  BEGIN
  mytestProc();
  EXCEPTION
  WHILE OTHERS THEN
  tmptab_cnt (anonymous"block" ");
  LIFT;
  END;
  /

  The result of this is:

  ...

  Created procedure.

  No errors.
  1 lines in tmptab in mytestProc

  
  1 lines in block anonymous tmptab
  BEGIN
  *
  ERROR on line 1:
  ORA-20001:
  ORA-06512: at line 6

  no selected line

  As you can see, there is 1 row in the table after that the procedure back to the anonymous block.

  Thanks for posting such a complete and clear test script; that really helps.

  How put you in the form code, with green background and title?

• How to prohibit this restoration?

  Hi, all the experts:
  How to prohibit this restoration? the hidden or using events parameter?
  
  
  Thanks in advance.

  >
  Roll forward is the application of commited transactions recorded in the journals of restoration by progression, for the data files before a crash.
  >
  Isn't that partially correct - before rolling process also applies "uncommitted" transactions from the logs of recovery. These "committed," operations are then 'cancelled' during 'restore '.

  See the section 'Crash and recovery proceedings' in the doc of database Concepts
  http://docs.Oracle.com/CD/B28359_01/server.111/b28318/startup.htm#sthref1431
  >
  Crash recovery and instance involves two distinct operations: roll forward of the current data files online using the operations committed and uncommitted contained in online redo records and then back change in transactions not posted to their original state.
  >
  The reason that uncommitted transactions must also be applied is explained in this section.
  >
  Two potential problems can occur if an instance failure occurs:

  Blocks of data modified by a transaction could not be written in time the validation data files and may only appear in the redo log. Therefore, the redo log contains changes must be redone to the database during recovery.

  After the roll before the phase, data files could contain changes that had not committed at the time of the failure. These uncommitted changes must be cancelled in order to ensure the consistency of the transactions. These changes were either recorded in the files of data before the failure, is introduced in the phase before rolling.
  >
  The foregoing explains how data files can contain uncommitted changes.
  >
  To resolve this dilemma, two distinct steps are typically used by the Oracle database for a successful recovery from a failure of the system: roll forward with the redo log (recovery of cache) and roll back with lower prices or cancel segments (recovery of the transaction).
  >
  'Cancel' the unsaved changes in files of data Oracle to use rollback/undo segments. The problem is that when you get the information you need not necessarily be in these segments. But this information IS in the REDO logs.

  So part of this "roll forward" phase works also on segments rollback/cancellation to the same point in time as the data files... Then the phase of "restore" can "undo" transactions not posted from data files using that info restored by the restore/undo segments that present the advanced phase of roll.

• GG replicat & excerpt problem 2013-03-08 13:06:10 ERROR OGG-00446.

  I use the door Golden for oracle to mysql guide.
  
  use oracle 11g
  
  GoldenGate for oracle
  
  Oracle GoldenGate for Oracle shell
  OGGCORE_11.2.1.0.1_PLATFORMS_120423.0230_FBO version 11.2.1.0.1
  
  GoldenGate for mysql
  
  Oracle GoldenGate interpreter for MySQL
  Version 11.2.1.0.1 OGGCORE_11.2.1.0.1_PLATFORMS_120423.0230
  
  extract from report of eora1:
  
  
  ...
  
  2013-03-08 13:05:38 WARNING OGG-01423 n valid default destination directory archive newspaper found for thread 1.
  
  2013-03-08 13:05:38 INFO OGG-00546 default thread stack size: 8388608.
  
  2013-03-08-13:05:38 INFO OGG-01513 positioning (Thread 1) sequence 7, ARI 32085520, RCS 0.1045884.
  
  2013-03-08 13:05:38 INFO OGG-01516 positioned (Thread 1) sequence 7, RBA 32085520, RCS 0.1045884, March 8, 2013 13:04:08.
  
  2013-03-08 13:05:38 INFO OGG - 01517 Position of the first record transformed for threads 1, 7, RBA 32085520, RCS 0.1045878 sequence, 8 March 2013 13:04:08.
  
  2013-03-08 13:05:38 INFO OGG-00732 found markers of recovery thread crash #1 on sequence RBA 32119312 7. Abandonment of the uncommitted transactions.
  
  2013-03-08 13:05:43 size of buffer of INFORMATION OGG-01226 taken 27985 value (RAS size 27985).
  
  2013-03-08 13:05:43 initialization INFO OGG-01055 recovery completed for the file target. / dirdat/t1000004, at 1120 RBA.
  
  2013-03-08 13:05:43 INFO OGG-01478 output file ./dirdat/t1 using the format VERSION 11.2.
  
  2013-03-08 13:05:43 INFO OGG-01026 reversal remote file. / dirdat/t1000005.
  
  2013-03-08 13:05:43 INFO OGG-01053 recovery completed for the target file. / dirdat/t1000005, RBA 1060.
  
  2013-03-08 13:05:43 INFO OGG-01057 recovery completed for all targets.
  
  ***********************************************************************
  * Run time Messages *.
  ***********************************************************************
  
  replicat rmsq1 report:
  
  
  ...
  
  Database version:
  MySQL
  Server version: 5.5.28 - 0ubuntu0.12.04.2 - log
  Client version: 6.0.0
  Host connection: Localhost via UNIX socket
  Protocol version: 10
  
  Context of the source:
  SourceModule: [er.common]
  SourceID: [scratch/aime1/adestore/views/aime1_adc4150256/oggcore/OpenSys/src/app/er/common.cpp]
  SourceFunction: [extract_start_point]
  SourceLine: [2090]
  ThreadBacktrace: [7] elements
  : [home/bw/wangchao/goldengate_mysql/libgglog.so(CMessageContext::AddThreadContext()+0x26) [0x7f06c34e3bc6]]
  : [home/bw/wangchao/goldengate_mysql/libgglog.so (CMessageFactory::CreateMessage (CSourceContext *, unsigned int,...) + 0 x 366) [0x7f06c34dd8b6]]
  : [home/bw/wangchao/goldengate_mysql/libgglog.so (_MSG_ERR_CHECKPOINT_GENERIC (CSourceContext *, const char *, CMessageFactory::MessageDisposition) + 0x42) [0x7f06c34b6102]]
  : [Home/bw/wangchao/goldengate_mysql/replicat (extract_start_point (time_elt_def *, time_elt_def *) + 0 x 177) [0x53b687]]
  : [home/bw/wangchao/goldengate_mysql/replicat(main+0xa6b) [0x57ec0b]]
  : [lib/x86_64-linux-gnu/libc.so.6(__libc_start_main+0xed) [0x7f06c070476d]]
  : [home/bw/wangchao/goldengate_mysql/replicat(__gxx_personality_v0+0x282) [0x4b678a]]
  
  2013-03-08 13:06:10 ERROR OGG-00446.
  
  2013-03-08 13:06:10 ERROR OGG-01668 PROCESS ABENDING.
  
  
  the journal of replicat confuses me...

  $. / ogg oggerr 446
  00446, 00000, '{0} '.
  * {0}: errorText (String)
  * Cause: An error occurred during the processing of the checkpoint file.
  * Action: Contact the Oracle Support.

  You're checkpoint file is damaged, which is very rare. Sounds like a test system and you have probably something old around you didn't clean or upgrade correctly. I have just start over, make sure that nothing is dirchk and start adding new processes.

  Good luck
  -joe

• Archiving log / nologging / direct path insert

  Could you please confirm if the following are true or correct me if my interpretation is wrong:
  
  Log mode archive 1) and logging is necessary to deal with the media recovery; It was not necessary for the recovery of the instance.
  
  (2) IF the insert is no. APPEND mode, redo is generated, even if the table is in nologging mode and database is in log noachive mode. This recovery is necessary for recovery of the instance.
  
  (3) direct path insert ignore cancel the generation and can jump redo generation if the object is in nologging mode.
  
  Thank you.
  
  In case if it is relevant, I'm using Oracle 11.2.0.3.

  (1) Yes, archive logs is required for media recovery.
  2 and 3) even if the table is in nologging mode, it generates little recovery data dictionary and index maintenance. After a reboot of a failure - Oracle reads the online redo logs and replay any what transaction it finds in there. It's the little "roll forward." The binary redo information are used to re-read everything that did not get written to the data files. This review included refreshing the information for CANCELLATION (UNDO is protected by roll forward).

  After the restoration has been applied, the database is generally available for use now - and restore begins. For any transaction that was being processed when the instance failed - we have to revert his changes, roll back. We do this by treating the cancellation of all uncommitted transactions.

  The database is now fully recovered.

  Also read he following link
  http://docs.Oracle.com/CD/B19306_01/server.102/b14220/startup.htm
  http://asktom.Oracle.com/pls/asktom/f?p=100:11:0:P11_QUESTION_ID:5280714813869