Oracle之SQL优化专题02-稳固SQL执行计划的方法

Posted on November 5, 2018 by Alfred Zhao

首先构建一个简单的测试用例来实际演示：

create table emp as select * from scott.emp;
create table dept as select * from scott.dept;
create index idx_emp_empno on emp(empno);
create index idx_dept_deptno on dept(deptno);

测试过程中查看真实执行计划的方法：

set lines 1000 pages 1000
alter session set statistics_level = ALL;
Execute SQL;
select * from table(dbms_xplan.display_cursor(null,null,'allstats last'));

正常的SQL执行，执行计划会走相应的索引：

--good SQL: 39dv3d8jkzyuw
select a.empno, a.ename, b.dname, a.job, a.sal from emp a, dept b where a.deptno = b.deptno and empno = 7788;

--good xplan: 1725450077
SQL> select * from table(dbms_xplan.display_cursor(null,null,'allstats last'));

PLAN_TABLE_OUTPUT
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
SQL_ID  39dv3d8jkzyuw, child number 0
-------------------------------------
select a.empno, a.ename, b.dname, a.job, a.sal from emp a, dept b where
a.deptno = b.deptno and empno = 7788

Plan hash value: 1725450077

--------------------------------------------------------------------------------------------------------------------
| Id  | Operation                     | Name            | Starts | E-Rows | A-Rows |   A-Time   | Buffers | Reads  |
--------------------------------------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT              |                 |      1 |        |      1 |00:00:00.01 |       6 |      2 |
|   1 |  NESTED LOOPS                 |                 |      1 |      1 |      1 |00:00:00.01 |       6 |      2 |
|   2 |   NESTED LOOPS                |                 |      1 |      1 |      1 |00:00:00.01 |       5 |      2 |
|   3 |    TABLE ACCESS BY INDEX ROWID| EMP             |      1 |      1 |      1 |00:00:00.01 |       3 |      1 |
|*  4 |     INDEX RANGE SCAN          | IDX_EMP_EMPNO   |      1 |      1 |      1 |00:00:00.01 |       2 |      1 |
|*  5 |    INDEX RANGE SCAN           | IDX_DEPT_DEPTNO |      1 |      1 |      1 |00:00:00.01 |       2 |      1 |
|   6 |   TABLE ACCESS BY INDEX ROWID | DEPT            |      1 |      1 |      1 |00:00:00.01 |       1 |      0 |
--------------------------------------------------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

   4 - access("EMPNO"=7788)
   5 - access("A"."DEPTNO"="B"."DEPTNO")

糟糕的SQL执行，执行计划走全表扫描（这里实验直接利用使用hint强制不走索引来模拟这种情况）:

--bad SQL: dqd10y7wqrg7f
select /*+ no_index(a idx_emp_empno) no_index(b idx_dept_deptno)*/a.empno, a.ename, b.dname, a.job, a.sal from emp a, dept b where a.deptno = b.deptno and empno = 7788;

--bad xplan: 1123238657
SQL> select * from table(dbms_xplan.display_cursor(null,null,'allstats last'));

PLAN_TABLE_OUTPUT
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
SQL_ID  dqd10y7wqrg7f, child number 1
-------------------------------------
select /*+ no_index(a idx_emp_empno) no_index(b
idx_dept_deptno)*/a.empno, a.ename, b.dname, a.job, a.sal from emp a,
dept b where a.deptno = b.deptno and empno = 7788

Plan hash value: 1123238657

----------------------------------------------------------------------------------------------------------------
| Id  | Operation          | Name | Starts | E-Rows | A-Rows |   A-Time   | Buffers |  OMem |  1Mem | Used-Mem |
----------------------------------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT   |      |      1 |        |      1 |00:00:00.01 |       5 |       |       |          |
|*  1 |  HASH JOIN         |      |      1 |      1 |      1 |00:00:00.01 |       5 |  1214K|  1214K|  377K (0)|
|*  2 |   TABLE ACCESS FULL| EMP  |      1 |      1 |      1 |00:00:00.01 |       2 |       |       |          |
|   3 |   TABLE ACCESS FULL| DEPT |      1 |      4 |      4 |00:00:00.01 |       3 |       |       |          |
----------------------------------------------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

   1 - access("A"."DEPTNO"="B"."DEPTNO")
   2 - filter("EMPNO"=7788)

假设此时这些糟糕的SQL就是业务实际的SQL，且对应开发人员无法更改SQL文本（这里就是指无法去掉不走索引的hint），那么现在如何能将这些糟糕的SQL绑定成走索引的执行计划呢？
糟糕的SQL清单：

select /*+ no_index(a idx_emp_empno) no_index(b idx_dept_deptno)*/a.empno, a.ename, b.dname, a.job, a.sal from emp a, dept b where a.deptno = b.deptno and empno = 7788;

select /*+ no_index(a idx_emp_empno) no_index(b idx_dept_deptno)*/a.empno, a.ename, b.dname, a.job, a.sal from emp a, dept b where a.deptno = b.deptno and empno = 7900;

如何让其走索引？目前Oracle常见的2种稳固执行计划的方式：

1.SQL Profile稳固执行计划

适用于Oracle 10g及以上版本。
利用MOS文档215187.1提供的系列脚本中的coe_xfr_sql_profile.sql来稳固执行计划，只需要输入要调整SQL的SQL_ID和好的执行计划的plan_hash_value即可，脚本内容可参考：

在本次演示实验中，就是将sql_id=’dqd10y7wqrg7f’的SQL绑定好的plan_hash_value=1725450077，具体使用过程如下：

SQL> @coe_xfr_sql_profile.sql

Parameter 1:
SQL_ID (required)

Enter value for 1: dqd10y7wqrg7f


PLAN_HASH_VALUE AVG_ET_SECS
--------------- -----------
     1123238657        .095

Parameter 2:
PLAN_HASH_VALUE (required)

Enter value for 2: 1725450077

Values passed to coe_xfr_sql_profile:
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
SQL_ID         : "dqd10y7wqrg7f"
PLAN_HASH_VALUE: "1725450077"
...
Execute coe_xfr_sql_profile_dqd10y7wqrg7f_1725450077.sql
on TARGET system in order to create a custom SQL Profile
with plan 1725450077 linked to adjusted sql_text.


COE_XFR_SQL_PROFILE completed.

然后按照提示执行生成的coe_xfr_sql_profile_dqd10y7wqrg7f_1725450077.sql脚本即可；
需要特别注意的是：可以根据实际情况是否需要修改这个脚本中的force_match的值为true。
本次的例子，就是没有使用到绑定变量，而需求是不仅让empno = 7788的条件走索引，还要让其他输入值，比如empno = 7900也同样走索引，那就需要修改这个force_match的值为true。稳固执行计划的效果如下：

SQL> select /*+ no_index(a idx_emp_empno) no_index(b idx_dept_deptno)*/a.empno, a.ename, b.dname, a.job, a.sal from emp a, dept b where a.deptno = b.deptno and empno = 7788;

     EMPNO ENAME      DNAME          JOB              SAL
---------- ---------- -------------- --------- ----------
      7788 SCOTT      RESEARCH       ANALYST         3000

SQL> select * from table(dbms_xplan.display_cursor(null,null,'allstats last'));

PLAN_TABLE_OUTPUT
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
SQL_ID  dqd10y7wqrg7f, child number 0
-------------------------------------
select /*+ no_index(a idx_emp_empno) no_index(b
idx_dept_deptno)*/a.empno, a.ename, b.dname, a.job, a.sal from emp a,
dept b where a.deptno = b.deptno and empno = 7788

Plan hash value: 1725450077

-----------------------------------------------------------------------------------------------------------
| Id  | Operation                     | Name            | Starts | E-Rows | A-Rows |   A-Time   | Buffers |
-----------------------------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT              |                 |      1 |        |      1 |00:00:00.01 |       6 |
|   1 |  NESTED LOOPS                 |                 |      1 |      1 |      1 |00:00:00.01 |       6 |
|   2 |   NESTED LOOPS                |                 |      1 |      1 |      1 |00:00:00.01 |       5 |
|   3 |    TABLE ACCESS BY INDEX ROWID| EMP             |      1 |      1 |      1 |00:00:00.01 |       3 |
|*  4 |     INDEX RANGE SCAN          | IDX_EMP_EMPNO   |      1 |      1 |      1 |00:00:00.01 |       2 |
|*  5 |    INDEX RANGE SCAN           | IDX_DEPT_DEPTNO |      1 |      1 |      1 |00:00:00.01 |       2 |
|   6 |   TABLE ACCESS BY INDEX ROWID | DEPT            |      1 |      1 |      1 |00:00:00.01 |       1 |
-----------------------------------------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

   4 - access("EMPNO"=7788)
   5 - access("A"."DEPTNO"="B"."DEPTNO")

Note
-----
   - SQL profile coe_dqd10y7wqrg7f_1725450077 used for this statement

常用操作：
1)查询sql_profile
可以通过查询dba_sql_profiles来确认数据库中的sql_profile：

select * from dba_sql_profiles;

2)删除sql_profile
如果有一天不再需要这个sql_profile来稳固执行计划，可以这样删除sql_profile：
exec dbms_sqltune.drop_sql_profile(‘name’);

exec dbms_sqltune.drop_sql_profile('coe_dqd10y7wqrg7f_1725450077');

3)清除SQL执行计划
还可以清除共享池中指定SQL的执行计划：
exec sys.dbms_shared_pool.purge(‘address,hash_value’,’c’);

SQL> select sql_id, address, hash_value, plan_hash_value, sql_profile from v$sql where sql_id = 'dqd10y7wqrg7f';

SQL_ID        ADDRESS          HASH_VALUE PLAN_HASH_VALUE SQL_PROFILE
------------- ---------------- ---------- --------------- ----------------------------------------------------------------
dqd10y7wqrg7f 0000000076B909F8 4184587502      1123238657
dqd10y7wqrg7f 0000000076B909F8 4184587502      1123238657
dqd10y7wqrg7f 0000000076B909F8 4184587502      1725450077 coe_dqd10y7wqrg7f_1725450077

SQL> exec sys.dbms_shared_pool.purge('0000000076B909F8,4184587502','c');

PL/SQL procedure successfully completed.

SQL> select sql_id, address, hash_value, plan_hash_value, sql_profile from v$sql where sql_id = 'dqd10y7wqrg7f';

no rows selected

2.SPM稳固执行计划

适用于Oracle 11g及以上版本。
删除掉之前的sql_profile，尝试使用SPM来稳固执行计划，实际上，手工生成sql_plan_baseline的方式要更加灵活，但我实际用的比较少。
查看sql_plan_baselines：

select * from dba_sql_plan_baselines;
select sql_handle, plan_name, origin, enabled, accepted, fixed from dba_sql_plan_baselines;

SPM稳固执行计划方法：

var temp number
--1.bad: sql_id & plan_hash_value
exec :temp := dbms_spm.load_plans_from_cursor_cache(sql_id => '', plan_hash_value => );
--2.good: sql_id & plan_hash_value & sql_handle
exec :temp := dbms_spm.load_plans_from_cursor_cache(sql_id => '', plan_hash_value => , sql_handle => );
--3.drop bad plan_name
exec :temp := dbms_spm.drop_sql_plan_baseline(sql_handle => '', plan_name => '');

用上面的例子具体说明：

--1.bad: sql_id & plan_hash_value
SQL> var temp number
SQL> exec :temp := dbms_spm.load_plans_from_cursor_cache(sql_id => 'dqd10y7wqrg7f', plan_hash_value => 1123238657);
SQL>  select sql_handle, plan_name, origin, enabled, accepted, fixed from dba_sql_plan_baselines;

SQL_HANDLE                     PLAN_NAME                      ORIGIN         ENA ACC FIX
------------------------------ ------------------------------ -------------- --- --- ---
SQL_9c3626a309e5e8bd           SQL_PLAN_9sdj6nc4ybu5x96fd8705 MANUAL-LOAD    YES YES NO

--2.good: sql_id & plan_hash_value & sql_handle（上面查到的）
SQL> exec :temp := dbms_spm.load_plans_from_cursor_cache(sql_id => '39dv3d8jkzyuw', plan_hash_value =>1725450077, sql_handle => 'SQL_9c3626a309e5e8bd');

PL/SQL procedure successfully completed.

SQL>  select sql_handle, plan_name, origin, enabled, accepted, fixed from dba_sql_plan_baselines;

SQL_HANDLE                     PLAN_NAME                      ORIGIN         ENA ACC FIX
------------------------------ ------------------------------ -------------- --- --- ---
SQL_9c3626a309e5e8bd           SQL_PLAN_9sdj6nc4ybu5x2b78d17a MANUAL-LOAD    YES YES NO
SQL_9c3626a309e5e8bd           SQL_PLAN_9sdj6nc4ybu5x96fd8705 MANUAL-LOAD    YES YES NO

--3.drop bad plan_name
SQL> exec :temp := dbms_spm.drop_sql_plan_baseline(sql_handle => 'SQL_9c3626a309e5e8bd', plan_name => 'SQL_PLAN_9sdj6nc4ybu5x96fd8705');

PL/SQL procedure successfully completed.

SQL> select sql_handle, plan_name, origin, enabled, accepted, fixed from dba_sql_plan_baselines;

SQL_HANDLE                     PLAN_NAME                      ORIGIN         ENA ACC FIX
------------------------------ ------------------------------ -------------- --- --- ---
SQL_9c3626a309e5e8bd           SQL_PLAN_9sdj6nc4ybu5x2b78d17a MANUAL-LOAD    YES YES NO

验证稳固执行计划的效果：

SQL> select /*+ no_index(a idx_emp_empno) no_index(b idx_dept_deptno)*/a.empno, a.ename, b.dname, a.job, a.sal from emp a, dept b where a.deptno = b.deptno and empno = 7788;

     EMPNO ENAME      DNAME          JOB              SAL
---------- ---------- -------------- --------- ----------
      7788 SCOTT      RESEARCH       ANALYST         3000

SQL> select * from table(dbms_xplan.display_cursor(null,null,'allstats last'));

PLAN_TABLE_OUTPUT
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
SQL_ID  dqd10y7wqrg7f, child number 1
-------------------------------------
select /*+ no_index(a idx_emp_empno) no_index(b
idx_dept_deptno)*/a.empno, a.ename, b.dname, a.job, a.sal from emp a,
dept b where a.deptno = b.deptno and empno = 7788

Plan hash value: 1725450077

-----------------------------------------------------------------------------------------------------------
| Id  | Operation                     | Name            | Starts | E-Rows | A-Rows |   A-Time   | Buffers |
-----------------------------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT              |                 |      1 |        |      1 |00:00:00.01 |       6 |
|   1 |  NESTED LOOPS                 |                 |      1 |      1 |      1 |00:00:00.01 |       6 |
|   2 |   NESTED LOOPS                |                 |      1 |      1 |      1 |00:00:00.01 |       5 |
|   3 |    TABLE ACCESS BY INDEX ROWID| EMP             |      1 |      1 |      1 |00:00:00.01 |       3 |
|*  4 |     INDEX RANGE SCAN          | IDX_EMP_EMPNO   |      1 |      1 |      1 |00:00:00.01 |       2 |
|*  5 |    INDEX RANGE SCAN           | IDX_DEPT_DEPTNO |      1 |      1 |      1 |00:00:00.01 |       2 |
|   6 |   TABLE ACCESS BY INDEX ROWID | DEPT            |      1 |      1 |      1 |00:00:00.01 |       1 |
-----------------------------------------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

   4 - access("EMPNO"=7788)
   5 - access("A"."DEPTNO"="B"."DEPTNO")

Note
-----
   - SQL plan baseline SQL_PLAN_9sdj6nc4ybu5x2b78d17a used for this statement

可以看到SPM已经起作用了。但如果谓词条件换成7900，就会不起作用。我没有找到SPM中类似像sql_profile中force_match的参数，日常工作中也是使用sql_profile稳固执行计划多一些。
另外注意dba_sql_plan_baselines中记录的执行计划对应的ACCEPTED和ENABLE的值都为YES，才可能会被SQL使用。
常用操作：
select sql_handle, plan_name, origin, enabled, accepted, fixed from dba_sql_plan_baselines;
1)将ENABLE的值设为”YES” or “NO”

var temp number
exec :temp := dbms_spm.alter_sql_plan_baseline(sql_handle => 'SQL_9c3626a309e5e8bd', plan_name => 'SQL_PLAN_9sdj6nc4ybu5x96fd8705', attribute_name => 'ENABLED', attribute_value => 'YES');

var temp number
exec :temp := dbms_spm.alter_sql_plan_baseline(sql_handle => 'SQL_9c3626a309e5e8bd', plan_name => 'SQL_PLAN_9sdj6nc4ybu5x96fd8705', attribute_name => 'ENABLED', attribute_value => 'NO');

2)将ACCEPTED值设为”YES”

var temp clob
exec :temp := dbms_spm.evolve_sql_plan_baseline(sql_handle => 'SQL_9c3626a309e5e8bd', plan_name => 'SQL_PLAN_9sdj6nc4ybu5x96fd8705', verify => 'NO', commit => 'YES');

注：我这里测试(在11.2.0.4环境下)发现ACCEPTED值设为”YES”后，无法再设置成”NO”，而ENABLED的值可以自由设置为”YES” or “NO”。

This entry was posted in Oracle性能优化 and tagged , , . Bookmark the permalink.