How to Create primary key index with duplicate rows.

Question

Hi All,
While rebuilding an index on a table , I am getting error that there are duplicate rows in a table.
Searching out the reason led me to an interesting observation.
Please follow.
SELECT * FROM user_ind_columns WHERE table_name='SERVICE_STATUS';
INDEX_NAME     TABLE_NAME     COLUMN_NAME     COLUMN_POSITION     COLUMN_LENGTH     CHAR_LENGTH     DESCEND
SERVICE_STATUS_PK     SERVICE_STATUS     SUBSCR_NO_RESETS     2     22     0      ASC
SERVICE_STATUS_PK     SERVICE_STATUS     STATUS_TYPE_ID     3     22     0     ASC
SERVICE_STATUS_PK     SERVICE_STATUS     ACTIVE_DT     4     7     0     ASC
SERVICE_STATUS_PK     SERVICE_STATUS     SUBSCR_NO     1     22     0     ASC
SELECT index_name,index_type,table_name,table_type,uniqueness, status,partitioned FROM user_indexes WHERE index_name='SERVICE_STATUS_PK';
INDEX_NAME     INDEX_TYPE      TABLE_NAME     TABLE_TYPE     UNIQUENESS     STATUS     PARTITIONED
SERVICE_STATUS_PK     NORMAL     SERVICE_STATUS     TABLE     UNIQUE     VALID     NO
SELECT constraint_name ,constraint_type,table_name,status,DEFERRABLE,DEFERRED,validated,index_name
FROM user_constraints WHERE constraint_name='SERVICE_STATUS_PK';
CONSTRAINT_NAME     CONSTRAINT_TYPE     TABLE_NAME      STATUS     DEFERRABLE     DEFERRED     VALIDATED     INDEX_NAME
SERVICE_STATUS_PK     P     SERVICE_STATUS     ENABLED     NOT DEFERRABLE     IMMEDIATE VALIDATED     SERVICE_STATUS_PK
1. Using index scan:
SELECT COUNT (*)
FROM (SELECT subscr_no, active_dt, status_type_id, subscr_no_resets
FROM service_status
GROUP BY subscr_no, active_dt, status_type_id, subscr_no_resets
HAVING COUNT (*) > 1) ;
no rows returned
Explain plan:
Operation     OBJECT Name     ROWS     Bytes     Cost     OBJECT Node     IN/OUT     PStart     PStop
SELECT STATEMENT Optimizer MODE=CHOOSE          519 K          14756                     
FILTER                                        
SORT GROUP BY NOSORT          519 K     7 M     14756                     
INDEX FULL SCAN     ARBOR.SERVICE_STATUS_PK     10 M     158 M     49184                     
2. Using Full scan:
SELECT COUNT (*)
FROM (SELECT /*+ full(s) */ subscr_no, active_dt, status_type_id, subscr_no_resets
FROM service_status s
GROUP BY subscr_no, active_dt, status_type_id, subscr_no_resets
HAVING COUNT (*) > 1) ;
71054 rows returned.
Explain Plan:
Operation     Object Name     Rows     Bytes     Cost     Object Node     In/Out     PStart     PStop
SELECT STATEMENT Optimizer Mode=CHOOSE          1           24123                     
SORT AGGREGATE          1                               
VIEW          519 K          24123                     
FILTER                                        
SORT GROUP BY          519 K     7 M     24123                     
TABLE ACCESS FULL     ARBOR.SERVICE_STATUS     10 M     158 M     4234                     
Index SERVICE_STATUS_PK is a unique and composite primary key VALID index. And the constraint is ENABLED and VALIDATED still having duplicate rows in table.
How it is possible?
Is it an Oracle soft Bug??
Regards,
Saket Bansal

Answer 1

saket bansal wrote:
Values are inserted as single rows inserts through an GUI interface.And you still claim to have over 71K duplicate records, without the GUI getting any kind of errors?
That does not add up and can only be explained by a "bug".
I tried inserting a duplicate record but failed.
SQL> insert into service_status (select * from service_status where rownum <2);
insert into service_status (select * from service_status where rownum <2)
ERROR at line 1:
ORA-00001: unique constraint (ARBOR.SERVICE_STATUS_PK) violatedAre you really sure there is no other way data in this table is populated/manipulated in bulk?