Technical Blog


Technical and Product News and Insights from Rackspace

Oracle online table partitions using the DBMS_REDEFINITION package


Starting with OracleĀ® 10g, you can partition tables online without any application downtime by using the DBMS_REDEFINITION package.

Use the following steps to change a non-partition table to a partition table by using DBMS_REDEFINITION. This example changes the non-partition table, TABLEA, to a range interval partition table.

Step 1: Backup the unpartition table

Run the following code to create a full export backup of the table, TABLEA.

expdp  \"/ as sysdba\" directory=EXPDP_DIR dumpfile=tableA_UNPAR.dmp logfile=tableA_UNPAR.log TABLES=TEST.TABLEA

expdp  \"/ as sysdba\"  directory=EXPDP_DIR dumpfile=tableA_metaunpar.dmp logfile=tableA_metaunpar.log TABLES=TEST.TABLEA content=metadata_only

Step 2: Examine the database objects

The following dependent (D) database objects can be dropped when a table is dropped:

  • CONSTRAINT (Constraints) D

  • INDEX (Indexes) D

  • MATERIALIZED_VIEW_LOG (Materialized view logs) D

  • OBJECT_GRANT (Object grants) D

  • TRIGGER (Triggers) D

Run the following SQL commands and save the output in a spool file, such as cons_trig_indx.txt:

set LINESIZE 500
set PAGESIZE 1000
SQL> spool cons_trig_indx.txt
SQL> select name, type, owner from all_dependencies where referenced_owner = 'TEST' and referenced_name = 'TABLEA';

NAME                TYPE              OWNER
--------------      --------------    -------
PROC_TABLEA         PROCEDURE         TEST
TABLEA_TRIGG        TRIGGER           TEST
PKG_TABLEA          PACKAGE BODY      TEST


SQL> select OWNER, INDEX_NAME, TABLE_OWNER, TABLE_NAME, STATUS, TABLESPACE_NAME
from dba_indexes where TABLE_OWNER='TEST' and TABLE_NAME='TABLEA';

OWNER   INDEX_NAME       TABLE_OWNER  TABLE_NAME   STATUS   TABLESPACE_NAME
---------------------------------------------------------------------------
TEST    TABLEA_IDX_ID01    TEST        TABLEA      VALID    TABLEA_TBL
TEST    TABLEA_IDX_ID04    TEST        TABLEA      VALID    TABLEA_TBL
TEST    TABLEA_IDX_PK      TEST        TABLEA      VALID    TABLEA_TBL


SQL> select STATUS, OBJECT_TYPE, OBJECT_NAME  from dba_objects
where OWNER='TEST' and OBJECT_TYPE = 'TRIGGER' and STATUS='INVALID';

no rows selected

SQL> select CONSTRAINT_NAME, CONSTRAINT_TYPE from dba_constraints
where TABLE_NAME='TABLEA' and owner='TEST';
SQL> spool off

CONSTRAINT_NAME     C
------------------  -----
SYS_C002004601      C
SYS_C002004602      C
SYS_C002004603      C
IDX_PK              P
FK01                R

Step 3: Capture the DDL of TABLEA

Run the following commands to capture the data definition language (DDL) of TABLEA and save scripts in spool file DEF_TABLEA.sql before you create the partition table:

set echo off
set feedback off
set linesize 160
set long 2000000
set pagesize 0
set trims on
column txt format a150 word_wrapped
SQL> spool DEF_TABLEA.sql
SQL> select DBMS_METADATA.GET_DDL('TABLE','TABLEA','TEST') txt FROM dual;
SQL> spool off

Step 4: Copy the DDL script

Run the following command to copy the DDL script that you created in step 3.

cp DEF_TABLEA.sql DEF_TABLEA_PAR.sql

Step 5: Review the dates in the non-partition table

Run the following command to find the dates in TABLEA:

SQL> select * from (select DT from TEST.TABLEA where rownum <15 order by DT DESC);

Step 6: Edit DEF_TABLEA_PAR.sql file

Edit DEF_TABLEA_PAR.sql to make the following changes:

  • Change all occurrences of TABLEA to TABLEA_PAR.

  • Delete all the constraints, like NOT NULL or any other constraints.

  • Insert the following command so that the table is created in a new tablespace:

      TABLESPACE "TABLEA_TBL_PAR" LOGGING
    
  • Insert the following commands to add the partition definition based on the dates identified in Step 5:

      PARTITION BY RANGE(DT)
      interval (numtoyminterval(1,'MONTH'))
      (partition TABLEA_2004  values less than  (to_date('01/01/2005','DD/MM/YYYY')),
       partition TABLEA_2005 values less than  (to_date('01/01/2006','DD/MM/YYYY')));
    

The DEF_TABLEA_PAR.sql file should now look like the following example:

CREATE TABLE "TEST"."TABLEA_PAR"
(    "ID" NUMBER(6,0),
     "CEID" NUMBER(6,0),
     "DT" DATE,
     "AMT" NUMBER(14,4),
     "RET" NUMBER(14,4),
     "CNT" NUMBER(4,0),
     "VCNT" NUMBER(4,0),
     "EXEDT" DATE,
     "LASTUPDBY" VARCHAR2(15),
     "VENUM" NUMBER(6,0),
     "LASTUPDDT" TIMESTAMP (6))
TABLESPACE "TABLEA_TBL_PAR" LOGGING
PARTITION BY RANGE(DT)
interval (numtoyminterval(1,'MONTH'))
(partition TABLEA_2004  values less than  (to_date('01/01/2005','DD/MM/YYYY')),
 partition TABLEA_2005  values less than  (to_date('01/01/2006','DD/MM/YYYY')));

Step 7: Create the partition table

Create the partition table by running the following steps to run the DEF_TABLEA_PAR.sql script:

SQL> spool DEF_TABLEA_PAR.outp.txt
SQL> @DEF_TABLEA_PAR.sql

Table Created.

SQL> spool off

Step 8: Verify the partition table

Run the following commands to verify the partition table and return the defined partitions:

SQL> spool verify_partition.txt
SQL> select partition_name from DBA_tab_partitions where table_name ='TABLEA_PAR' and table_owner = 'TEST';
SQL> spool off

PARTITION_NAME
-----------------
TABLEA_2004
TABLEA_2005

Step 9: Gather statistics on the non-partition table

Run the following commands to gather statistics on the non-partition table and save them to a spool file.

SQL> SPOOL gather_stats.txt
SQL> exec dbms_stats.gather_table_stats ('TEST', 'TABLEA',cascade => TRUE);
SQL> spool off

Step 10: Check redefinition feasibility

Note: The source table (non-partitioned) does not need a primary key before you use the redefinition package.

Run the following commands to see if redefinition is possible, and save the results to a spool file:

SQL> spool check_the_redefinition.txt
SQL> EXEC DBMS_Redefinition.can_redef_table ('TEST', 'TABLEA');
SQL> spool off

Step 11: Start redefinition

If no errors are listed in check_the_redefinition.txt, start the redefinition by using the following long-running command:

SQL> spool start_redef_table.txt
SQL>begin
    dbms_redefinition.start_redef_table
    (
     uname => 'TEST',
     orig_table => 'TABLEA',
     int_table => 'TABLEA_PAR');
     end;
   /
SQL> spool off

Step 12: Watch for tablespace errors during redefinition

The redefinition operation in step 11 might result in tablespace alerts like the following example:

ERROR at line 1:
ORA-12008: error in materialized view refresh path
ORA-01688: unable to extend table TEST.TABLEA_PAR
partition SYS_P42 by 1024 in tablespace TABLEA_TBL
ORA-06512: at "SYS.DBMS_REDEFINITION", line 52
ORA-06512: at "SYS.DBMS_REDEFINITION", line 1646
ORA-06512: at line 2

ERROR at line 1:
ORA-12008: error in materialized view refresh path
ORA-14400: inserted partition key does not map to any partition
ORA-06512: at "SYS.DBMS_REDEFINITION", line 52
ORA-06512: at "SYS.DBMS_REDEFINITION", line 1646
ORA-06512: at line 2

If you see tablespace errors similar to the preceding example, then you should take the following steps:

  1. Run the following command to stop the redefinition process.

     SQL> spool abort_redef_table.txt
     SQL> begin
          dbms_redefinition.abort_redef_table
          (
          uname => 'TEST',
          orig_table => 'TABLEA',
          int_table => 'TABLEA_PAR');
          end;
         /
     SQL> spool off
    
  2. Drop the partition table and the materialized view.

  3. Increase the size of tablespace. In this example, you should increase the size of the tablespace TABLEA_TBL.

  4. Rerun step 11.

Step 13: Check for redefinition errors

After the redefinition process completes successfully, run the following commands to check for any errors:

SQL> spool copy_table_dependents.txt
SQL> SET SERVEROUTPUT ON
     DECLARE
     l_num_errors PLS_INTEGER;
     BEGIN
       DBMS_REDEFINITION.copy_table_dependents(
           uname             => 'TEST',
           orig_table        => 'TABLEA',
           int_table         => 'TABLEA_PAR',
           copy_indexes      => DBMS_REDEFINITION.cons_orig_params, -- Non-Default
           num_errors        => l_num_errors);
           DBMS_OUTPUT.put_line('l_num_errors=' || l_num_errors);
     END;
/
SQL> spool off

If the redefinition was successful, you should see results similar to the following in the copy_table_dependents.txt file:

l_num_errors=0
PL/SQL procedure successfully completed.

Step 14: (Optional) Resynchronize the partition table

If you like, run the following commands to resynchronize the partition table with an interim name:

SQL> spool sync_interim_table.txt
SQL>
     BEGIN
       DBMS_REDEFINITION.sync_interim_table
       (
           uname => 'TEST',
           orig_table => 'TABLEA',
           int_table => 'TABLEA_PAR');
      END;
/
SQL> spool off

Step 15: Gather statistics on the partition table

Run the following commands to gather statistics on the partition table:

SQL> spool gather_statistics_par.txt
SQL> exec dbms_stats.gather_table_stats ('TEST', 'TABLEA_PAR',cascade => TRUE);
SQL> spool off

Step 16: Create constraint script

Run the following commands to prepare a script to enable the validate constraint.

SQL> spool constraint_enable_validate.txt
SET LINESIZE 500
SET PAGESIZE 1000

SQL> select 'alter table' ||' '||OWNER||'.'||TABLE_NAME||' enable validate constraint'||' '||CONSTRAINT_NAME||';' from dba_constraints where TABLE_NAME = 'TABLEA_PAR' and OWNER='TEST';

'ALTERTABLE'||''||OWNER||'.'||TABLE_NAME||'ENABLEVALIDATECONSTRAINT'||''||CONSTR
--------------------------------------------------------------------------------
alter table TEST.TABLEA_PAR enable validate constraint TMP$$_SYS_C002004601;
alter table TEST.TABLEA_PAR enable validate constraint TMP$$_SYS_C002004602;
alter table TEST.TABLEA_PAR enable validate constraint TMP$$_SYS_C002004603;
alter table TEST.TABLEA_PAR enable validate constraint TMP$$_IDX_PK;
alter table TEST.TABLEA_PAR enable validate constraint TMP$$_FK01;

SQL> spool off

Step 17: Enable the validate constraint

Run the script and commands produced by step 16, as shown in the following example:

SQL> spool constraint_enable_execute.outp.txt
SQL>@constraint_enable.sql

alter table TEST.TABLEA_PAR enable validate constraint TMP$$_SYS_C002004601;
alter table TEST.TABLEA_PAR enable validate constraint TMP$$_SYS_C002004602;
alter table TEST.TABLEA_PAR enable validate constraint TMP$$_SYS_C002004603;
alter table TEST.TABLEA_PAR enable validate constraint TMP$$_IDX_PK;
alter table TEST.TABLEA_PAR enable validate constraint TMP$$_FK01;

SQL> spool off

Step 18: Compare non-partition and partition tables

Compare the original, non-partition table with the new, partition table to verify that all attributes are the same.

Step 19: Rename the tables

Run the following commands to set the interim table as the real table to switch the table names:

SQL> spool finish_redef_table.txt
     BEGIN
       DBMS_REDEFINITION.finish_redef_table
      (
        uname => 'TEST',
        orig_table => 'TABLEA',
        int_table => 'TABLEA_PAR');
     END;
/

--------------------------------------------
@?/rdbms/admin/utlrp.sql
--------------------------------------------

SQL>spool off

Step 20: Compare the tables

Run the following commands to compare the record counts of both tables and make sure they match:

SQL> spool table_count.outp.txt
SQL> select count(*) from TEST.TABLEA;

 COUNT(*)
----------
  890540

SQL> select count (*) from TEST.TABLEA_PAR;

 COUNT(*)
----------
  890540

SQL> spool off

Step 21: Verify partition success

Run the following commands to verify that the partition process was successful:

SQL> spool check_partition.txt
SQL> select partitioned from dba_tables where table_name = 'TABLEA' and owner='TEST';

PAR
------
YES

SQL> select partition_name , SUBPARTITION_COUNT, TABLESPACE_NAME from dba_tab_partitions where table_name='TABLEA' and table_owner='TEST';
SQL> select table_name, partition_name, high_value, partition_position from DBA_tab_partitions where table_name='TABLEA' and table_owner='TEST';
SQL> spool off

Step 22: Reexamine the database objects

Run the following commands to examine the database objects and compare the results to step 2:

SET LINESIZE 500
SET PAGESIZE 1000
SQL> spool cons_indx_trigg.txt
SQL> select name, type, owner from all_dependencies where referenced_owner = 'TEST' and referenced_name = 'TABLEA';

NAME                TYPE              OWNER
----------------    ---------------   ------------
PROC_TABLEA         PROCEDURE         TEST
TABLEA_TRIGG        TRIGGER           TEST
PKG_TABLEA          PACKAGE BODY      TEST

SQL> select OWNER, INDEX_NAME, TABLE_OWNER, TABLE_NAME, STATUS, TABLESPACE_NAME from dba_indexes where TABLE_OWNER='TEST' and TABLE_NAME='TABLEA';

OWNER  INDEX_NAME       TABLE_OWNER TABLE_NAME  STATUS   TABLESPACE_NAME
------------------------------------------------------------------------
TEST   TABLEA_IDX_ID01  TEST        TABLEA      VALID    TABLEA_TBL
TEST   TABLEA_IDX_ID04  TEST        TABLEA      VALID    TABLEA_TBL
TEST   TABLEA_IDX_PK    TEST        TABLEA      VALID    TABLEA_TBL

SQL> select STATUS, OBJECT_TYPE, OBJECT_NAME  from dba_objects where OWNER='TEST' and OBJECT_TYPE = 'TRIGGER' and STATUS='INVALID';

no rows selected

SQL> select CONSTRAINT_NAME, CONSTRAINT_TYPE from dba_constraints where TABLE_NAME='TABLEA' and owner='TEST';

CONSTRAINT_NAME        C
-------------------		----------
SYS_C002004601         C
SYS_C002004602         C
SYS_C002004603         C
IDX_PK                 P
FK01                   R

12 rows selected.

SQL> spool off

Step 23: Rebuild the indexes

Run the following commands to rebuild the indexes on the new tablespace:

SQL> spool rebuild_indx.txt
SQL>@rebuild_index.sql

ALTER INDEX TEST.TABLEA_IDX_ID01 REBUILD TABLESPACE TABLEA_TBL_PAR ONLINE;
ALTER INDEX TEST.ITABLEA_IDX_ID04 REBUILD TABLESPACE TABLEA_TBL_PAR ONLINE;
ALTER INDEX TEST.TABLEA_IDX_PK REBUILD TABLESPACE TABLEA_TBL_PAR ONLINE;

SQL> spool off

Step 24: Validate the indexes

Run the following commands to verify that the status is valid and that the tablespace for all indexes is TABLEA_TBL_PAR:

SQL> spool verify_indx.outp.txt
SQL> select OWNER, INDEX_NAME, TABLE_OWNER, TABLE_NAME, STATUS, TABLESPACE_NAME from dba_indexes where TABLE_OWNER='TEST' and TABLE_NAME='TABLEA';

OWNER  INDEX_NAME       TABLE_OWNER  TABLE_NAME   STATUS   TABLESPACE_NAME
---------------------------------------------------------------------------
TEST   TABLEA_IDX_ID01  TEST         TABLEA       VALID   	 TABLEA_TBL_PAR
TEST   TABLEA_IDX_ID04  TEST         TABLEA       VALID   	 TABLEA_TBL_PAR
TEST   TABLEA_IDX_PK    TEST         TABLEA       VALID     TABLEA_TBL_PAR

SQL>spool off

Step 25: Drop original non-partition table

After the DBAs have confirmed that everything looks good, execute the following command to remove the original table, which now has the name of the interim table, TEST.TABLEA_PAR:

SQL> DROP table TEST.TABLEA_PAR cascade constraints;

Conclusion

The preceding steps used the interim table, TEST.TABLEA_PAR, to partition the table, TEST.TABLEA, into a range interval table without any application downtime.

Use the Feedback tab to make any comments or ask questions.

post avatar
Pradeep Rai

Share this information: