Gather Schema Statistics fails with Ora-20001 errors after 11G database upgrade

Gather Schema Statistics" program reported following errors in request log files :

Error #1: ERROR: While GATHER_TABLE_STATS: 
object_name=GL.JE_BE_LINE_TYPE_MAP***ORA-20001: invalid column name or duplicate columns/column groups/expressions in method_opt*** 
Error #2: ERROR: While GATHER_TABLE_STATS: 
object_name=GL.JE_BE_LOGS***ORA-20001: invalid column name or duplicate columns/column groups/expressions in method_opt*** 
Error #3: ERROR: While GATHER_TABLE_STATS: 
object_name=GL.JE_BE_VAT_REP_RULES***ORA-20001: invalid column name or duplicate columns/column groups/expressions in method_opt***

Error #4: ERROR: While GATHER_TABLE_STATS:  object_name=FII.FII_FIN_ITEM_HIERARCHIES***ORA-20001: invalid column name or duplicate columns/column groups/expressions in method_opt***

There are two reasons for that error message:

1 ) There are duplicate rows on FND_HISTOGRAM_COLS table for JE_BE_LINE_TYPE_MAP table.
Because of this problem, FND_STATS tries to gather histogram information using wrong command and
it fails with ora-20001 errors.

Following SQL should have returned one row , not two.

SQL> select a.column_name, nvl(a.hsize,254) hsize
from FND_HISTOGRAM_COLS a
where table_name = 'JE_BE_LINE_TYPE_MAP'
order by column_name;

COLUMN_NAME HSIZE
------------------------------ ----------
SOURCE 254
SOURCE 254

2) Column does not exist on the table but still listed in FND_HISTOGRAMS_COL table.

Solution:
Find out all duplicates and/or obsolete rows in FND_HISTOGRAM_COLS and delete one of them.
Remember to take backup of the FND_HISTOGRAM_COLS table before deleting any data.

-- identify duplicate rows
select table_name, column_name, count(*)
from FND_HISTOGRAM_COLS
group by table_name, column_name
having count(*) > 1;


-- Use above results on the following SQL to delete duplicates
delete from FND_HISTOGRAM_COLS
where table_name = '&TABLE_NAME'
and  column_name = '&COLUMN_NAME'
and rownum=1;


-- Use following SQL to delete obsoleted rows
delete from FND_HISTOGRAM_COLS
where (table_name, column_name) in 
  (
   select hc.table_name, hc.column_name
   from FND_HISTOGRAM_COLS hc , dba_tab_columns tc
   where hc.table_name  ='&TABLE_NAME'
   and hc.table_name= tc.table_name (+)
   and hc.column_name = tc.column_name (+)
   and tc.column_name is null
  );


Cheers!!!

Related Posts by Categories