Any way to force FTS without any query change

Question

Hi All,
One of my Oracle 11i apps database got one Cache buffer chain latch issue and it is causing high cpu spikes.
I found the culprit SQL and its one of the Index (Nested Loop operation) which is causing high CBC latch.
I am not in a position to put any HINT or modify the SQL. When i tested the sql with FULL hint, query gave results (100k records) in less than 1 min without any latch contention.
My apps database is running on 10.2.0.4.0 RHEL 5.4 64 bit.
SQL
select
int.rowid  ,  decode(int.SEGMENT1|| int.SEGMENT2|| int.SEGMENT3|| int.SEGMENT4|| int.SEGMENT5|| int.SEGMENT6|| int.SEGMENT7 , '',  
decode (ccid_cc.code_combination_id, NULL, NULL, replace(ccid_cc.SEGMENT4,'.',' ') || '.' || replace(ccid_cc.SEGMENT1,'.',' ') || '.' || replace(ccid_cc.SEGMENT2,'.',' ') || '.' ||
replace(ccid_cc.SEGMENT3,'.',' ') || '.' || replace(ccid_cc.SEGMENT5,'.',' ') || '.' || replace(ccid_cc.SEGMENT6,'.',' ') || '.' || replace(ccid_cc.SEGMENT7,'.',' ')  ),
replace(int.SEGMENT4,'.',' ') || '.' || replace(int.SEGMENT1,'.',' ') || '.' || replace(int.SEGMENT2,'.',' ') || '.' || replace(int.SEGMENT3,'.',' ') || '.' ||
replace(int.SEGMENT5,'.',' ') || '.' || replace(int.SEGMENT6,'.',' ') || '.' || replace(int.SEGMENT7,'.',' ') ) flexfield , nvl(flex_cc.code_combination_id, nvl(int.code_combination_id, -4))  ,
decode(int.SEGMENT1|| int.SEGMENT2|| int.SEGMENT3|| int.SEGMENT4|| int.SEGMENT5|| int.SEGMENT6|| int.SEGMENT7 , '', decode(ccid_cc.code_combination_id, null,
decode(int.code_combination_id, null, -4, -5), decode(sign(nvl(ccid_cc.start_date_active, int.accounting_date-1) - int.accounting_date), 1, -1,
decode(sign(nvl(ccid_cc.end_date_active, int.accounting_date +1) - int.accounting_date), -1, -1, 0)) +  decode(ccid_cc.enabled_flag, 'N', -10, 0) + 
decode(ccid_cc.summary_flag, 'Y', -100, decode(int.actual_flag, 'B', decode(ccid_cc.detail_budgeting_allowed_flag, 'N', -100, 0),  decode(ccid_cc.detail_posting_allowed_flag, 'N', -100, 0)))),
decode(flex_cc.code_combination_id, null, -4, decode(sign(nvl(flex_cc.start_date_active, int.accounting_date-1) - int.accounting_date), 1, -1,
decode(sign(nvl(flex_cc.end_date_active, int.accounting_date +1) - int.accounting_date), -1, -1, 0)) +  decode(flex_cc.enabled_flag, 'N', -10, 0) +  decode(flex_cc.summary_flag, 'Y', -100, 
decode(int.actual_flag, 'B', decode(flex_cc.detail_budgeting_allowed_flag, 'N', -100, 0),  decode(flex_cc.detail_posting_allowed_flag, 'N', -100, 0)))))   , int.user_je_category_name ,
int.user_je_category_name , 'UNKNOWN' period_name , decode(actual_flag, 'B'   , decode(period_name, NULL, '-1' ,period_name), nvl(period_name, '0')) period_name2  , currency_code  ,
decode(actual_flag   , 'A', actual_flag   , 'B', decode(budget_version_id        , 1244, actual_flag        , 1303, actual_flag        , 1344, actual_flag        , NULL, '1', '6')   , 'E',
decode(encumbrance_type_id        , 1000, actual_flag        , 1001, actual_flag        , 1020, actual_flag        , NULL, '3', '4'), '5') actual_flag  , '0' exception_rate ,
decode(currency_code   , 'JPY', 1   , 'STAT', 1   , decode(actual_flag, 'E', -8, 'B', 1      , decode(user_currency_conversion_type          , 'User',
decode(currency_conversion_rate, NULL, -1, currency_conversion_rate)         ,'Corporate',decode(currency_conversion_date,NULL,-2,-6)          ,'Spot',
decode(currency_conversion_date,NULL,-2,-6)          ,'Bloomberg Daily Rate',decode(currency_conversion_date,NULL,-2,-6)          , NULL, decode(currency_conversion_rate,NULL,
decode(decode(nvl(to_char(entered_dr),'X'),'X',1,2),decode(nvl(to_char(accounted_dr),'X'),'X',1,2),                  
decode(decode(nvl(to_char(entered_cr),'X'),'X',1,2),decode(nvl(to_char(accounted_cr),'X'),'X',1,2),-20,-3),-3),-9),-9))) currency_conversion_rate ,
to_number(to_char(nvl(int.currency_conversion_date,            int.accounting_date), 'J'))  , decode(int.actual_flag  , 'A',
decode(int.currency_code    , 'JPY', 'User'   , 'STAT', 'User'    , nvl(int.user_currency_conversion_type, 'User'))  , 'B', 'User', 'E', 'User'   ,
nvl(int.user_currency_conversion_type, 'User')) user_currency_conversion_type  , ltrim(rtrim(substrb(rtrim(substrb(int.reference1, 1, 50)) || ' ' || int.user_je_source_name || ' 10409378: ' || int.actual_flag || ' ' || int.group_id, 1, 100))) , rtrim(substrb(nvl(rtrim(int.reference2), 'Journal Import ' || int.user_je_source_name || ' 10409378:'), 1, 240)) ,
ltrim(rtrim(substrb(rtrim(rtrim(rtrim(substrb(int.reference4, 1, 25)) || ' ' || int.user_je_category_name || ' ' || int.currency_code || decode(int.actual_flag, 'E', ' ' || int.encumbrance_type_id,  'B', ' ' || int.budget_version_id,  '') || ' ' || int.user_currency_conversion_type || ' ' || decode(int.user_currency_conversion_type, NULL, '', 'User', to_char(int.currency_conversion_rate))) || ' ' || to_char(int.currency_conversion_date)) || ' ' || substrb(int.reference8, 1, 15) || int.originating_bal_seg_value, 1, 100))) , rtrim(nvl(rtrim(int.reference5), 'Journal Import 10409378:')) , rtrim(substrb(nvl(rtrim(int.reference6), 'Journal Import Created'), 1, 80)) , rtrim(decode(upper(substrb(nvl(rtrim(int.reference7), 'N'), 1, 1)),'Y','Y', 'N')) , decode(upper(substrb(int.reference7, 1, 1)), 'Y', decode(rtrim(reference8), NULL, '-1', rtrim(substrb(reference8, 1, 15))), NULL) , rtrim(upper(substrb(int.reference9, 1, 1))) , rtrim(nvl(rtrim(int.reference10), nvl(to_char(int.subledger_doc_sequence_value), 'Journal Import Created'))) , int.entered_dr , int.entered_cr , to_number(to_char(int.accounting_date,'J')) , to_char(int.accounting_date, 'YYYY/MM/DD')  , int.user_je_source_name , nvl(int.encumbrance_type_id, -1) , nvl(int.budget_version_id, -1) , NULL , int.stat_amount , decode(int.actual_flag   , 'E', decode(int.currency_code, 'STAT', '1', '0'), '0') , decode(int.actual_flag   , 'A', decode(int.budget_version_id     , NULL, decode(int.encumbrance_type_id, NULL, '0', '1')           , decode(int.encumbrance_type_id, NULL, '2', '3'))   ,
'B', decode(int.encumbrance_type_id     , NULL, '0', '4')   , 'E', decode(int.budget_version_id     , NULL, '0', '5'), '0') , int.accounted_dr ,
int.accounted_cr , nvl(int.group_id, -1) , nvl(int.average_journal_flag, 'N') , int.originating_bal_seg_value
from GL.GL_INTERFACE int,  GL.gl_code_combinations flex_cc,  GL.gl_code_combinations ccid_cc
where int.set_of_books_id = 43
and int.status != 'PROCESSED'
and accounting_date between to_date('2010/06/01', 'YYYY/MM/DD')
and to_date('2010/06/30', 'YYYY/MM/DD')
and (int.user_je_source_name,group_id) in (('Receivables', 74486))
and flex_cc.SEGMENT1(+) = int.SEGMENT1
and flex_cc.SEGMENT2(+) = int.SEGMENT2
and flex_cc.SEGMENT3(+) = int.SEGMENT3
and flex_cc.SEGMENT4(+) = int.SEGMENT4
and flex_cc.SEGMENT5(+) = int.SEGMENT5
and flex_cc.SEGMENT6(+) = int.SEGMENT6
and flex_cc.SEGMENT7(+) = int.SEGMENT7
and flex_cc.chart_of_accounts_id(+) = 101
and flex_cc.template_id(+) is NULL
and ccid_cc.code_combination_id(+) = int.code_combination_id
and ccid_cc.chart_of_accounts_id(+) = 101
and ccid_cc.template_id(+) is NULL
order by decode(int.SEGMENT1|| int.SEGMENT2|| int.SEGMENT3|| int.SEGMENT4|| int.SEGMENT5|| int.SEGMENT6|| int.SEGMENT7 , ''  ,rpad(ccid_cc.SEGMENT4,30) || '.' || rpad(ccid_cc.SEGMENT1,30) || '.' || rpad(ccid_cc.SEGMENT2,30) || '.' || rpad(ccid_cc.SEGMENT3,30) || '.' || rpad(ccid_cc.SEGMENT5,30) || '.' || rpad(ccid_cc.SEGMENT6,30) || '.' || rpad(ccid_cc.SEGMENT7,30)  , rpad(int.SEGMENT4,30) || '.' || rpad(int.SEGMENT1,30) || '.' || rpad(int.SEGMENT2,30) || '.' || rpad(int.SEGMENT3,30) || '.' || rpad(int.SEGMENT5,30) || '.' || rpad(int.SEGMENT6,30) || '.' || rpad(int.SEGMENT7,30)   ) , int.entered_dr, int.accounted_dr, int.entered_cr, int.accounted_cr, int.accounting_date
/Execution plan
| Id  | Operation                      | Name                    | Rows  | Bytes | Cost (%CPU)| Time     |
|   0 | SELECT STATEMENT               |                         |       |       |  2027 (100)|          |
|   1 |  SORT ORDER BY                 |                         |     1 |   374 |  2027   (1)| 00:00:25 |
|   2 |   NESTED LOOPS OUTER           |                         |     1 |   374 |  2026   (1)| 00:00:25 |
|   3 |    NESTED LOOPS OUTER          |                         |     1 |   315 |  2024   (1)| 00:00:25 |
|*  4 |     TABLE ACCESS BY INDEX ROWID| GL_INTERFACE            |     1 |   256 |  2022   (1)| 00:00:25 |
|*  5 |      INDEX RANGE SCAN          | GL_INTERFACE_N1         | 25375 |       |   132   (0)| 00:00:02 |
|*  6 |     TABLE ACCESS BY INDEX ROWID| GL_CODE_COMBINATIONS    |     1 |    59 |     2   (0)| 00:00:01 |
|*  7 |      INDEX RANGE SCAN          | GL_CODE_COMBINATIONS_N7 |  4211 |       |     2   (0)| 00:00:01 |
|*  8 |    TABLE ACCESS BY INDEX ROWID | GL_CODE_COMBINATIONS    |     1 |    59 |     2   (0)| 00:00:01 |
|*  9 |     INDEX UNIQUE SCAN          | GL_CODE_COMBINATIONS_U1 |     1 |       |     1   (0)| 00:00:01 |
Predicate Information (identified by operation id):
   4 - filter(("ACCOUNTING_DATE">=TO_DATE(' 2010-06-01 00:00:00', 'syyyy-mm-dd hh24:mi:ss') AND
              "INT"."STATUS"<>'PROCESSED' AND "ACCOUNTING_DATE"<=TO_DATE(' 2010-06-30 00:00:00', 'syyyy-mm-dd
              hh24:mi:ss')))
   5 - access("INT"."USER_JE_SOURCE_NAME"='Receivables' AND "INT"."SET_OF_BOOKS_ID"=43 AND
              "GROUP_ID"=74486)
   6 - filter(("FLEX_CC"."CHART_OF_ACCOUNTS_ID"=101 AND "FLEX_CC"."TEMPLATE_ID" IS NULL AND
              "FLEX_CC"."SEGMENT1"="INT"."SEGMENT1" AND "FLEX_CC"."SEGMENT2"="INT"."SEGMENT2" AND
              "FLEX_CC"."SEGMENT3"="INT"."SEGMENT3" AND "FLEX_CC"."SEGMENT4"="INT"."SEGMENT4" AND
              "FLEX_CC"."SEGMENT5"="INT"."SEGMENT5" AND "FLEX_CC"."SEGMENT6"="INT"."SEGMENT6"))
   7 - access("FLEX_CC"."SEGMENT7"="INT"."SEGMENT7")
   8 - filter(("CCID_CC"."CHART_OF_ACCOUNTS_ID"=101 AND "CCID_CC"."TEMPLATE_ID" IS NULL))
   9 - access("CCID_CC"."CODE_COMBINATION_ID"="INT"."CODE_COMBINATION_ID")SQL plan with FULL(flex_cc) hint
| Id  | Operation                      | Name                    | Rows  | Bytes | Cost (%CPU)| Time     |
|   0 | SELECT STATEMENT               |                         |     1 |   374 | 15377   (4)| 00:03:05 |
|   1 |  SORT ORDER BY                 |                         |     1 |   374 | 15377   (4)| 00:03:05 |
|   2 |   NESTED LOOPS OUTER           |                         |     1 |   374 | 15376   (4)| 00:03:05 |
|*  3 |    HASH JOIN OUTER             |                         |     1 |   315 | 15374   (4)| 00:03:05 |
|*  4 |     TABLE ACCESS BY INDEX ROWID| GL_INTERFACE            |     1 |   256 |  2022   (1)| 00:00:25 |
|*  5 |      INDEX RANGE SCAN          | GL_INTERFACE_N1         | 25375 |       |   132   (0)| 00:00:02 |
|*  6 |     TABLE ACCESS FULL          | GL_CODE_COMBINATIONS    |  1193K|    67M| 13342   (4)| 00:02:41 |
|*  7 |    TABLE ACCESS BY INDEX ROWID | GL_CODE_COMBINATIONS    |     1 |    59 |     2   (0)| 00:00:01 |
|*  8 |     INDEX UNIQUE SCAN          | GL_CODE_COMBINATIONS_U1 |     1 |       |     1   (0)| 00:00:01 |
Predicate Information (identified by operation id):
   3 - access("FLEX_CC"."SEGMENT1"(+)="INT"."SEGMENT1" AND
              "FLEX_CC"."SEGMENT2"(+)="INT"."SEGMENT2" AND "FLEX_CC"."SEGMENT3"(+)="INT"."SEGMENT3" AND
              "FLEX_CC"."SEGMENT4"(+)="INT"."SEGMENT4" AND "FLEX_CC"."SEGMENT5"(+)="INT"."SEGMENT5" AND
              "FLEX_CC"."SEGMENT6"(+)="INT"."SEGMENT6" AND "FLEX_CC"."SEGMENT7"(+)="INT"."SEGMENT7")
   4 - filter("ACCOUNTING_DATE">=TO_DATE(' 2010-06-01 00:00:00', 'syyyy-mm-dd hh24:mi:ss') AND
              "INT"."STATUS"<>'PROCESSED' AND "ACCOUNTING_DATE"<=TO_DATE(' 2010-06-30 00:00:00', 'syyyy-mm-dd
              hh24:mi:ss'))
   5 - access("INT"."USER_JE_SOURCE_NAME"='Receivables' AND "INT"."SET_OF_BOOKS_ID"=43 AND
              "GROUP_ID"=74486)
   6 - filter("FLEX_CC"."CHART_OF_ACCOUNTS_ID"(+)=101 AND "FLEX_CC"."TEMPLATE_ID"(+) IS NULL)
   7 - filter("CCID_CC"."CHART_OF_ACCOUNTS_ID"(+)=101 AND "CCID_CC"."TEMPLATE_ID"(+) IS NULL)
   8 - access("CCID_CC"."CODE_COMBINATION_ID"(+)="INT"."CODE_COMBINATION_ID")
Above execution is very quick without any CBC latch contention.Is there any way, i can force FTS on this flex_cc table. I did try altering 'GL_CODE_COMBINATIONS_N7' index stats but then CBO is using GL_CODE_COMBINATIONS_N6 index and result is same CBC latch contention.
Index statitics :
select index_name,NUM_ROWS,SAMPLE_SIZE,LAST_ANALYZED,CLUSTERING_FACTOR,BLEVEL,LEAF_BLOCKS,DISTINCT_KEYS from dba_indexes where
index_name='GL_CODE_COMBINATIONS_N7';
INDEX_NAME                  NUM_ROWS SAMPLE_SIZE LAST_ANALYZED        CLUSTERING_FACTOR     BLEVEL LEAF_BLOCKS DISTINCT_KEYS
GL_CODE_COMBINATIONS_N7      7236665     7236665 2010/06/07 11:00:33             185935          2       20751          2293INDEX DDL
CREATE INDEX "GL"."GL_CODE_COMBINATIONS_N7" ON "GL"."GL_CODE_COMBINATIONS" ("SEGMENT7")
  PCTFREE 0 INITRANS 2 MAXTRANS 255 COMPUTE STATISTICS
  STORAGE(INITIAL 131072 NEXT 131072 MINEXTENTS 1 MAXEXTENTS 2147483645
  PCTINCREASE 0 FREELISTS 1 FREELIST GROUPS 1 BUFFER_POOL DEFAULT)
  TABLESPACE "APPS_TS_TX_IDX"

Answer 1

Have you looked at stored outlines ? They are for plan stability.
Also you may want to look little deep on where your statistics are going wrong.
I would recommend using below hint and checking where is difference in estimated(E-Rows) and actual number(A-rows) of rows. This may point out some incorrect stats.
e.g.
select /*+ gather_plan_statistics */ * from dual;
D
X
select * from table(dbms_xplan.display_cursor(null,null,'ALLSTATS LAST'));
Plan hash value: 272002086
| Id | Operation | Name | Starts | E-Rows | A-Rows | A-Time | Buffers |
| 1 | TABLE ACCESS FULL| DUAL | 1 | 1 | 1 |00:00:00.01 | 3 |
------------------------------------------------------------------------------------