How To reduce UNDO records

Question

Scenario:
Table A has 40 million records and table B has 30 million records.
And we have standy Database so there is no option for using Nologging.
Using Join and where clause i need to insert around 15 million records to table C.
If we insert using select statement it will generate more UNDO records
Please provide me the solution so that i can avoid UNDO as much as possibe.
Regards
Deepak S

Answer 1

>>when u give create table as it will generate undo . my table consists of 15 million records . when we create table d as select * from tablename . It will generte more undo..
My suggession is give the query in toad save the output in .dat file . using sqlloader we can import the data. So we can reduce the undo ....
Hmmm, have you some clue about regular sqlloader generate less redo than CTAS ?
We're agree about regular term since you have a standby databasen nologging operation should be avoid.
Read the following test case. Note that I'm alone on the database, so all the redo generation are because my queries.
First, test with CTAS method :
SQL> select b.name, a.value
  2  from   v$sysstat a, v$statname b
  3  where  a.statistic#=b.statistic#
  4* and    b.name like '%redo size%'
SQL> /
NAME                                                                  VALUE
redo size 846112
SQL> select count(*) from all_objects;
  COUNT(*)
     41183
SQL> create table myobjects as select * from all_objects;
Table created.
SQL> select b.name, a.value
  2  from   v$sysstat a, v$statname b
  3  where  a.statistic#=b.statistic#
  4* and    b.name like '%redo size%'
SQL> /
NAME                                                                  VALUE
redo size 5672516
SQL> select (5672516-846112)/1024/1024 from dual
SQL> /
(5672516-846112)/1024/1024
                4,6028175CTAS generate around 4,6Mb redolog.
Now, with your method. Creation of a data file :
QL> set head off
SQL> set pagesize 0
SQL> set linesize 1000
SQL> set trims on
SQL> spool E:\Scripts\Sql\myobjects.dat
SQL> select OWNER          ||';'||
            OBJECT_NAME    ||';'||
            SUBOBJECT_NAME ||';'||
            OBJECT_ID      ||';'||
            DATA_OBJECT_ID ||';'||
            OBJECT_TYPE    ||';'||
            CREATED        ||';'||
            LAST_DDL_TIME  ||';'||
            TIMESTAMP      ||';'||
            STATUS         ||';'||
            TEMPORARY      ||';'||
            GENERATED      ||';'||
            SECONDARY     
     from   myobjects;
SYS;DUAL;;258;258;TABLE;30/08/05;30/08/05;2005-08-30:13:50:32;VALID;N;N;N
PUBLIC;DUAL;;259;;SYNONYM;30/08/05;30/08/05;2005-08-30:13:50:32;VALID;N;N;N
SYS;SYSTEM_PRIVILEGE_MAP;;311;311;TABLE;30/08/05;30/08/05;2005-08-30:13:50:34;VALID;N;N;N
PUBLIC;SYSTEM_PRIVILEGE_MAP;;313;;SYNONYM;30/08/05;30/08/05;2005-08-30:13:50:34;VALID;N;N;N
SYS;TABLE_PRIVILEGE_MAP;;314;314;TABLE;30/08/05;30/08/05;2005-08-30:13:50:34;VALID;N;N;N
SQL> spool offThen a controlfile to use with sqlloader :
E:\Scripts\Sql\myobjects.ctl
LOAD DATA
INFILE 'E:\Scripts\Sql\myobjects.dat'
BADFILE 'E:\Scripts\Sql\myobjects.bad'
DISCARDFILE 'E:\Scripts\Sql\myobjects.dsc'
INTO TABLE myobjects
FIELDS TERMINATED BY ';'
(OWNER           CHAR(30),
OBJECT_NAME     CHAR(30),
SUBOBJECT_NAME  CHAR(30),
OBJECT_ID       INTEGER EXTERNAL(22) NULLIF OBJECT_ID=BLANKS,
DATA_OBJECT_ID  INTEGER EXTERNAL(22) NULLIF DATA_OBJECT_ID=BLANKS,
OBJECT_TYPE     CHAR(19),
CREATED         DATE(8) "DD/MM/YY" NULLIF CREATED=BLANKS,
LAST_DDL_TIME   DATE(8) "DD/MM/YY" NULLIF LAST_DDL_TIME=BLANKS,
TIMESTAMP       CHAR(19),
STATUS          CHAR(7),
TEMPORARY       CHAR(1),
GENERATED       CHAR(1),
SECONDARY       CHAR(1))Now, we create the empty table and check the redo amount before sqlloader :
SQL> create table myobjects as select * from all_objects where 1=2;
Table created.
SQL> select b.name, a.value
  2  from   v$sysstat a, v$statname b
  3  where  a.statistic#=b.statistic#
  4  and    b.name like '%redo size%'
  5  /
NAME                                                                  VALUE
redo size 6236212Run the sqlloader :
E:\oracle\ora102\BIN>sqlldr scott/demo102 control=E:\Scripts\Sql\myobjects.ctl
Commit point reached - logical record count 40897
Commit point reached - logical record count 40961
Commit point reached - logical record count 41025
Commit point reached - logical record count 41089
Commit point reached - logical record count 41153
Commit point reached - logical record count 41183
Commit point reached - logical record count 41184Last, check the redo amount and compare with previous amount :
SQL> select b.name, a.value
  2  from   v$sysstat a, v$statname b
  3  where  a.statistic#=b.statistic#
  4  and    b.name like '%redo size%'
  5  /
NAME                                                                  VALUE
redo size 11196136
SQL> select (11196136-6236212)/1024/1024 from dual;
(11196136-6236212)/1024/1024
                  4,73015213
SQL> select count(*) from myobjects;
     41184
SQL> SQL*Loader generate around 4,7Mb.
To resume the redo amount generation :
1. CTAS ~ 4,6Mb
2. SQL*Loader ~ 4,7Mb
What you said in your last post :
My suggession is give the query in toad save the output in .dat file . using sqlloader we can import the data. So we can reduce the undo ...Are you still sure now ?
Nicolas.
Well, I'm sorry, you said about undo and me about redo...
Message was edited by:
N. Gasparotto