由于物化视图定义为on commit导致update更新基表慢的解决方案

由于物化视图定义为on commit导致update更新基表慢的解决方案

以下是模拟和解决测试过程：

(模拟update慢的过程)

1、首先基于基表创建物化视图日志:

create materialized view log on scott.emp with rowid;

2、首先基于scott用户下emp创建物化视图：

create materialized view mv_emp

REFRESH FAST on commit

as

select * from scott.emp;

3、通过oracle 10046 查看update语句执行过程:

(1)sql>alter session set sql_trace=true;

(2)sql>alter session set tracefile_identifier='lzq';

(3)sql>alter session set events '10046 trace name context forever, level 1';

(4)sql>update scott.emp set sal=1450 where empno=7934;

(5)sql>alter session set events '10046 trace name context off';

(6)sql>show parameter user_dump_dest

(7)cd 到user_dump_dest查看trace文件被标识为lzq的trace 文件.

(8)格式化trace文件方便查看，tkprof prod2_ora_8623_lzq.trc prod2_ora_8623_lzq.out

(9)查看生成的prod2_ora_8623_lzq.out文件并查找跟物化视图mv_emp执行相关的信息见如下：

(生产环境中update一条语句真正执行的时间为1分多，而其中此处merger into "SCOTT"."MV_EMP" 占了1分左右

，而物化视图真正是给日报、月报、年报来用的跟客户沟通不用更新基表的时候就立即更新物化视图，物化视图

可以抛弃on commit时候就刷新，可以采取定时更新即可，从而可以避免update更新基表慢的问题，

MERGE INTO "SCOTT"."MV_EMP" "SNA$" USING (SELECT CURRENT$."EMPNO",

CURRENT$."ENAME",CURRENT$."JOB",CURRENT$."MGR",CURRENT$."HIREDATE",

CURRENT$."SAL",CURRENT$."COMM",CURRENT$."DEPTNO" FROM (SELECT "EMP"."EMPNO"

"EMPNO","EMP"."ENAME" "ENAME","EMP"."JOB" "JOB","EMP"."MGR" "MGR",

"EMP"."HIREDATE" "HIREDATE","EMP"."SAL" "SAL","EMP"."COMM" "COMM",

"EMP"."DEPTNO" "DEPTNO" FROM "EMP" "EMP") CURRENT$, (SELECT DISTINCT

MLOG$."EMPNO" FROM "SCOTT"."MLOG$_EMP" MLOG$ WHERE "XID$$" = :1   AND

("DMLTYPE$$" != 'D')) LOG$ WHERE CURRENT$."EMPNO" = LOG$."EMPNO")"AV$" ON

("SNA$"."EMPNO" = "AV$"."EMPNO") WHEN MATCHED THEN UPDATE SET

"SNA$"."EMPNO" = "AV$"."EMPNO","SNA$"."ENAME" = "AV$"."ENAME","SNA$"."JOB" =

"AV$"."JOB","SNA$"."MGR" = "AV$"."MGR","SNA$"."HIREDATE" =

"AV$"."HIREDATE","SNA$"."SAL" = "AV$"."SAL","SNA$"."COMM" = "AV$"."COMM",

"SNA$"."DEPTNO" = "AV$"."DEPTNO" WHEN NOT MATCHED THEN INSERT

(SNA$."EMPNO",SNA$."ENAME",SNA$."JOB",SNA$."MGR",SNA$."HIREDATE",SNA$."SAL",

SNA$."COMM",SNA$."DEPTNO") VALUES (AV$."EMPNO",AV$."ENAME",AV$."JOB",

AV$."MGR",AV$."HIREDATE",AV$."SAL",AV$."COMM",AV$."DEPTNO")

call  count   cpu  elapsed   disk   query  current    rows

------- ------ -------- ---------- ---------- ---------- ---------- ----------

Parse    1   0.00   0.00     0     0     0     0

Execute   1   0.01   0.08     2    20     5     1

Fetch    0   0.00   0.00     0     0     0     0

------- ------ -------- ---------- ---------- ---------- ---------- ----------

total    2   0.01   0.08     2    20     5     1

(模拟update慢的解决方案：)

1、首先删除之前创建的物化视图(先做个备份)

drop MATERIALIZED VIEW mv_emp;

2、创建物化视图

CREATE MATERIALIZED VIEW mv_emp

AS SELECT * FROM scott.emp;

3、建立一个用来刷新物化视图的存储过程:

CREATE OR REPLACE PROCEDURE auto_refresh_mview_job_proc

AS

BEGIN

dbms_mview.REFRESH('mv_emp');

END;

4、用ORACLE 10g的scheduler每天12:00和19:00定期刷新物化视图(时间可以根据需求定义)

BEGIN

DBMS_SCHEDULER.CREATE_JOB(

job_name => 'auot_refresh_mview_job',

job_type => 'STORED_PROCEDURE',

job_action => 'scott.auto_refresh_mview_job_proc',

start_date => SYSDATE,

repeat_interval => 'FREQ=DAILY; BYHOUR=12,19',

enabled => TRUE,

comments => 'Refresh materialized view mv_emp'

);

END;

5、通过oracle 10046 查看update语句执行过程:

(1)sql>alter session set sql_trace=true;

(2)sql>alter session set tracefile_identifier='lzq';

(3)sql>alter session set events '10046 trace name context forever, level 1';

(4)sql>update scott.emp set sal=1450 where empno=7934;

(5)sql>alter session set events '10046 trace name context off';

(6)sql>show parameter user_dump_dest

(7)cd 到user_dump_dest查看trace文件被标识为lzq的trace 文件.

(8)格式化trace文件方便查看，tkprof prod2_ora_8623_lzq.trc prod2_ora_8623_lzq.out

(9)查看生成的prod2_ora_8623_lzq.out文件并查找报告中是否还存在MERGE INTO "SCOTT"."MV_EMP"更新物化视图的信息，

此时因为已经定时为定时刷新，从而可以提高update语句的时间，从而优化过程完成，由于物化视图定义为on commit导致update更新基表慢的解决方案》(https://www.unjs.com)。