Oracle11g包dbms_parallel_execute在海量数据处理过程中的应用

实验环境介绍

11.2.0.1 RHEL6.5

本文简介

一个朋友own_my要处理批量数据，但是脚本跑的太慢了，于是网上搜到了dbms_parallel_execute这个包，用完后给我说这个包非常强大，于是我也学习学习，关于优化一直是我喜欢的内容，在参考了大神realkid4 的blog后，我自己也做了做实验，感觉很强大，记录在此。

实验部分

实验目标

测试dbms_parallel_execute包在海量数据处理过程中的应用。

实验过程

[oracle@etlhost206 ~]$ sqlplus / as sysdba

SQL*Plus: Release 11.2.0.1.0 Production on Wed Jun 3 13:40:34 2015

Copyright (c) 1982, 2009, Oracle. All rights reserved.

Connected to:

Oracle Database 11g Enterprise Edition Release 11.2.0.1.0 - 64bit Production

With the Partitioning, OLAP, Data Mining and Real Application Testing options

SQL> CONN LHR/lhr

Connected.

SQL> CREATE TABLE T AS SELECT * FROM DBA_OBJECTS;

Table created.

SQL> insert into t select * from t;

76369 rows created.

SQL> insert into t select * from t;

152738 rows created.

SQL> insert into t select * from t;

305476 rows created.

SQL> COMMIT;

Commit complete.

SQL> insert into t select * from t;

610952 rows created.

SQL> insert into t select * from t;

1221904 rows created.

SQL> insert into t select * from t;

2443808 rows created.

SQL> insert into t select * from t;

4887616 rows created.

SQL> COMMIT;

Commit complete.

SQL> insert into t select * from t;

9775232 rows created.

SQL> COMMIT;

Commit complete.

SQL> insert into t select * from t;

19550464 rows created.

SQL> COMMIT;

Commit complete.

SQL> select bytes/1024/1024 from dba_segments a where a.segment_name='T';

BYTES/1024/1024

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

4341

SQL> SELECT COUNT(1) FROM T;

COUNT(1)

----------

39100928

SQL> show parameter job

NAME TYPE VALUE

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

job_queue_processes integer 1000

SQL> show parameter cpu

NAME TYPE VALUE

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

cpu_count integer 8

parallel_threads_per_cpu integer 2

resource_manager_cpu_allocation integer 8

SQL> set timing on

SQL> set time on;

15:50:01 SQL>

15:50:02 SQL> show parameter job

NAME TYPE VALUE

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

job_queue_processes integer 1000

15:50:09 SQL> select bytes/1024/1024 from dba_segments a where a.segment_name='T';

BYTES/1024/1024

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

4341

Elapsed: 00:00:00.41

15:50:31 SQL> declare

15:50:39 2 vc_task varchar2(100);

15:50:39 3 vc_sql varchar2(1000);

15:50:39 4 n_try number;

15:50:39 5 n_status number;

15:50:39 6 begin

15:50:39 7 --Define the Task

15:50:39 8 vc_task := 'Task 1: By Rowid'; --Task名称

15:50:39 9 dbms_parallel_execute.create_task(task_name => vc_task); --手工定义一个Task任务；

15:50:39 10

15:50:39 11 --Define the Spilt

15:50:39 12 dbms_parallel_execute.create_chunks_by_rowid(task_name => vc_task,

15:50:39 13 table_owner => 'LHR',

15:50:39 14 table_name => 'T',

15:50:39 15 by_row => true,

15:50:39 16 chunk_size => 10000); --定义Chunk

15:50:39 17

15:50:39 18 vc_sql := 'update /*+ ROWID(dda) */ t set DATA_OBJECT_ID=object_id+1 where rowid between :start_id and :end_id';

15:50:40 19 --Run the task

15:50:40 20 dbms_parallel_execute.run_task(task_name => vc_task,

15:50:40 21 sql_stmt => vc_sql,

15:50:40 22 language_flag => dbms_sql.native,

15:50:40 23 parallel_level => 4); --执行任务，确定并行度

15:50:40 24

15:50:40 25 --Controller

15:50:40 26 n_try := 0;

15:50:40 27 n_status := dbms_parallel_execute.task_status(task_name => vc_task);

15:50:40 28 while (n_try \< 2 and n_status != dbms_parallel_execute.FINISHED) loop

15:50:40 29 dbms_parallel_execute.resume_task(task_name => vc_task);

15:50:40 30 n_status := dbms_parallel_execute.task_status(task_name => vc_task);

15:50:40 31 end loop;

15:50:40 32

15:50:40 33 --Deal with Result

15:50:40 34 dbms_parallel_execute.drop_task(task_name => vc_task);

15:50:40 35 end;

15:50:40 36 /

PL/SQL procedure successfully completed.

Elapsed: 00:03:50.78

15:58:05 SQL>

15:58:06 SQL> create index idx_t_id on t(object_id) nologging parallel 4;

Index created.

Elapsed: 00:01:35.12

16:00:05 SQL> alter index idx_t_id noparallel;

Index altered.

Elapsed: 00:00:00.07

16:00:15 SQL>

16:02:51 SQL> declare

16:02:52 2 vc_task varchar2(100);

16:02:52 3 vc_sql varchar2(1000);

16:02:52 4 n_try number;

16:02:52 5 n_status number;

16:02:52 6 begin

16:02:52 7 --Define the Task

16:02:52 8 vc_task := 'Task 2: By Number Col';

16:02:52 9 dbms_parallel_execute.create_task(task_name => vc_task);

16:02:52 10

16:02:52 11 --Define the Spilt

16:02:52 12 dbms_parallel_execute.create_chunks_by_number_col(task_name => vc_task,

16:02:52 13 table_owner => 'LHR',

16:02:52 14 table_name => 'T',

16:02:52 15 table_column => 'OBJECT_ID',

16:02:52 16 chunk_size => 100000); --定义chunk

16:02:53 17 16:02:53 18 vc_sql := 'update /*+ ROWID(dda) */ t set DATA_OBJECT_ID=object_id+1 where object_id between :start_id and :end_id';

16:02:53 19 --Run the task

16:02:53 20 dbms_parallel_execute.run_task(task_name => vc_task,

16:02:53 21 sql_stmt => vc_sql,

16:02:53 22 language_flag => dbms_sql.native,

16:02:53 23 parallel_level => 4);

16:02:53 24

16:02:53 25 --Controller

16:02:53 26 n_try := 0;

16:02:53 27 n_status := dbms_parallel_execute.task_status(task_name => vc_task);

16:02:53 28 while (n_try \< 2 and n_status != dbms_parallel_execute.FINISHED) loop

16:02:53 29 dbms_parallel_execute.resume_task(task_name => vc_task);

16:02:53 30 n_status := dbms_parallel_execute.task_status(task_name => vc_task);

16:02:53 31 end loop;

16:02:53 32

16:02:53 33 --Deal with Result

16:02:53 34 dbms_parallel_execute.drop_task(task_name => vc_task);

16:02:53 35 end;

16:02:53 36 /

\^Cdeclare

*

ERROR at line 1:

ORA-01013: user requested cancel of current operation

ORA-06512: at "SYS.DBMS_LOCK", line 201

ORA-06512: at "SYS.DBMS_PARALLEL_EXECUTE", line 44

ORA-06512: at "SYS.DBMS_PARALLEL_EXECUTE", line 390

ORA-06512: at "SYS.DBMS_PARALLEL_EXECUTE", line 417

ORA-06512: at line 20

Elapsed: 00:07:12.08

16:11:36 SQL>

16:11:36 SQL> EXEC dbms_parallel_execute.drop_task(task_name => 'Task 2: By Number Col');

PL/SQL procedure successfully completed.

Elapsed: 00:00:00.11

16:31:53 SQL> declare

16:32:05 2 vc_task varchar2(100);

16:32:05 3 vc_sql varchar2(1000);

16:32:05 4 vc_sql_mt varchar2(1000);

16:32:05 5 n_try number;

16:32:05 6 n_status number;

16:32:05 7 begin

16:32:05 8 --Define the Task

16:32:05 9 vc_task := 'Task 3: By SQL';

16:32:05 10 dbms_parallel_execute.create_task(task_name => vc_task);

16:32:05 11

16:32:05 12 --Define the Spilt

16:32:05 13 vc_sql_mt := 'select distinct object_id, object_id from t';

16:32:05 14 dbms_parallel_execute.create_chunks_by_SQL(task_name => vc_task,

16:32:05 15 sql_stmt => vc_sql_mt,

16:32:05 16 by_rowid => false);

16:32:05 17

16:32:05 18 vc_sql := 'update /*+ ROWID(dda) */t set DATA_OBJECT_ID=object_id+1 where object_id between :start_id and :end_id';

16:32:05 19 --Run the task

16:32:05 20 dbms_parallel_execute.run_task(task_name => vc_task,

16:32:05 21 sql_stmt => vc_sql,

16:32:05 22 language_flag => dbms_sql.native,

16:32:05 23 parallel_level => 4);

16:32:05 24

16:32:05 25 --Controller

16:32:05 26 n_try := 0;

16:32:05 27 n_status := dbms_parallel_execute.task_status(task_name => vc_task);

16:32:05 28 while (n_try \< 2 and n_status != dbms_parallel_execute.FINISHED) loop

16:32:05 29 dbms_parallel_execute.resume_task(task_name => vc_task);

16:32:05 30 n_status := dbms_parallel_execute.task_status(task_name => vc_task);

16:32:05 31 end loop;

16:32:05 32

16:32:05 33 --Deal with Result

16:32:05 34 dbms_parallel_execute.drop_task(task_name => vc_task);

16:32:05 35 end;

16:32:05 36 /

\^Cdeclare

*

ERROR at line 1:

ORA-01013: user requested cancel of current operation

ORA-06512: at "SYS.DBMS_PARALLEL_EXECUTE_INTERNAL", line 634

ORA-06512: at "SYS.DBMS_PARALLEL_EXECUTE", line 163

ORA-06512: at line 14

Elapsed: 00:01:09.08

16:33:14 SQL> EXEC dbms_parallel_execute.drop_task(task_name => 'Task 3: By SQL');

PL/SQL procedure successfully completed.

相关字典视图查询

create_chunks_by_rowid过程

SELECT * FROM DBA_PARALLEL_EXECUTE_TASKS;

SELECT * FROM DBA_PARALLEL_EXECUTE_CHUNKS;

SELECT count(1) FROM DBA_PARALLEL_EXECUTE_CHUNKS;

select status, count(*) from user_parallel_execute_chunks group by status;

select D.owner,D.job_name,D.JOB_STYLE,D.JOB_TYPE,D.JOB_ACTION from dba_scheduler_jobs d where d.owner='LHR';

告警日志：

Wed Jun 03 15:53:48 2015

Archived Log entry 1202 added for thread 1 sequence 2669 ID 0x6779dfc4 dest 1:

Thread 1 advanced to log sequence 2671 (LGWR switch)

Current log# 4 seq# 2671 mem# 0: /app/oracle/flash_recovery_area/CNYDB/onlinelog/o1_mf_4bpxd8g7v.log

Wed Jun 03 15:53:49 2015

Archived Log entry 1203 added for thread 1 sequence 2670 ID 0x6779dfc4 dest 1:

Wed Jun 03 15:53:57 2015

Thread 1 advanced to log sequence 2672 (LGWR switch)

Current log# 5 seq# 2672 mem# 0: /app/oracle/flash_recovery_area/CNYDB/onlinelog/o1_mf_5bpxdbwdz.log

Wed Jun 03 15:53:58 2015

Archived Log entry 1204 added for thread 1 sequence 2671 ID 0x6779dfc4 dest 1:

Thread 1 advanced to log sequence 2673 (LGWR switch)

Current log# 1 seq# 2673 mem# 0: /app/oracle/oradata/CNYDB/redo01.log

Wed Jun 03 15:54:04 2015

Archived Log entry 1205 added for thread 1 sequence 2672 ID 0x6779dfc4 dest 1:

Thread 1 advanced to log sequence 2674 (LGWR switch)

Current log# 6 seq# 2674 mem# 0: /app/oracle/flash_recovery_area/CNYDB/onlinelog/o1_mf_6bpxdcjx2.log

Wed Jun 03 15:54:05 2015

Archived Log entry 1206 added for thread 1 sequence 2673 ID 0x6779dfc4 dest 1:

由告警日志可以看出redo切换非常迅速，归档来不及，所以还是需要在空闲的时候来做实验。

create_chunks_by_number_col过程

SELECT * FROM DBA_PARALLEL_EXECUTE_CHUNKS;

SELECT * FROM DBA_PARALLEL_EXECUTE_TASKS;

select status, count(*) from dba_parallel_execute_chunks group by status;

select sid, serial#, status, PROGRAM, SQL_ID, event from v$session where action like 'TASK$%';

select D.owner,D.job_name,D.JOB_STYLE,D.JOB_TYPE,D.JOB_ACTION from dba_scheduler_jobs d where d.owner='LHR';

实验总结

由实验可以看出，采用dbms_parallel_execute.create_chunks_by_rowid方法，4千万的数据量大约4G大小的表更新完大约4分钟，这个速度还是可以的，另外2种方式更新下来速度太慢就没有测试了，具体可以参考这里：http://blog.itpub.net/26736162/viewspace-1683912/，http://blog.itpub.net/26736162/viewspace-1683913/。

实验脚本

create_chunks_by_rowid方式

create_chunks_by_number_col

create_chunks_by_SQL