首页 > 数据库DBA > oracle知识 > oracle调优 >

Oracle PL/SQL 性能分析工具 profiler 说明

作者: 初见博客 分类: oracle调优 发布时间: 2021-07-08 15:44

一.说明

Oracle提供了profiler工具包,利用该工具包可以查看PL/SQL执行过程中各模块的性能。可以从MOS上下载该脚本:

Implementing andUsing the PL/SQL Profiler [ID 243755.1]   脚本下载:profiler.zip

When there is asignificant gap between user elapsed time and SQL processing elapsed time, andthere is PL/SQL code involved, the PL/SQL Profiler becomes a very useful tool.It helps to identify the lines of PL/SQL code which are taking longer toprocess.

–当user elapsed time 和 SQL processing elapsedtime 有很大差别,且涉及到PL/SQL 代码时,就可以使用PL/SQL Profiler 工具,其可以指明行级PL/SQL 的时间。

 

For example, ifa transaction which uses PL/SQL Libraries (packages, procedures, functions ortriggers) executes in one hour of user elapsed time, and reviewing the resultsof the Trace Analyzer or TKPROF only 10 minutes of the elapsed time can beexplained with SQL commands being executed, then, by using the PL/SQL Profiler,a line-by-line of the executed PL/SQL application code is reported, includingthe total execution time for each line of code, and how many times each ofthese lines was executed.

–比如执行了一个PL/SQL 库(包,过程,函数或者触发器),userelapsed time 用了1个小时,但是通过trace 显示仅用了10分钟,这时就可以用PL/SQL Profiler,其可以报告每行代码执行的时间。

 

The actualPL/SQL Profiler is provided with the core RDBMS code, and it is well documentedon the Supplied PL/SQL Packages and Types Reference manual, under the packagename DBMS_PROFILER. This Note is about implementing and using the PL/SQLProfiler on any 9i or higher database, in order to debug the performance of anyPL/SQL application Library. The main script provided in this Note (profiler.sql)generates a comprehensive HTML report on the performance data extracted by theDBMS_PROFILER package.

–实际上,PL/SQL Profiler 仅提供了核心的RDBMS 代码,其在DBMS_PROFILER包里也说明。

 

PL/SQL Profiler 包含如下3个脚本:

profiler.sql – Reporting PL/SQL Profilerdata generated by DBMS_PROFILER (main script)

profgsrc.sql – Get source code for PL/SQLLibrary (package, procedure, function or trigger)

proftab.sql – Create tables for the PL/SQLprofiler

 

 

二.使用前的准备工作

 

2.1 If used for the first time, determineif DBMS_PROFILER is installed by doing a describe on that package

–如果是第一次使用,需要检查DBMS_PROFILER 包是否安装。

#sqlplus APPS/<pwd> –注意用户
SQL> DESC DBMS_PROFILER;

 

2.2 If DBMS_PROFILER is not installed,connect as SYS into SQL*Plus on database server, and execute command below tocreate the missing package:

–如果DBMS_PROFILER没有安装,使用SYS 用来执行如下命令来安装该包。

#sqlplus SYS/<pwd> –注意用户
SQL> START ?/rdbms/admin/profload.sql;

 

2.3  If used for the first time, andonce DBMS_PROFILER is installed, connect as application user into SQL*Plus, andcreate the repository tables PLSQL_PROFILER_RUNS, PLSQL_PROFILER_UNITS andPLSQL_PROFILER_DATA (proftab.sql is provided within PROF.zip, and is alsoavailable under $ORACLE_HOME/rdbms/admin)

–如果是第一次使用,且DBMS_PROFILER 已经安装,那么执行proftab.sql 脚本,创建存放统计数据的表:PLSQL_PROFILER_RUNS,PLSQL_PROFILER_UNITS and PLSQL_PROFILER_DATA。 该脚本在PROF 包里有,$ORACLE_HOME/rdbms/admin目录下也有。

#sqlplus APPS/<pwd>
SQL> START proftab.sql;

 

或:

SQL> start ?/rdbms/admin/proftab.sql

drop table plsql_profiler_data cascadeconstraints

*

ERROR at line 1:

ORA-00942: table or view does not exist

 

drop table plsql_profiler_units cascadeconstraints

*

ERROR at line 1:

ORA-00942: table or view does not exist

 

drop table plsql_profiler_runs cascadeconstraints

*

ERROR at line 1:

ORA-00942: table or view does not exist

 

drop sequence plsql_profiler_runnumber

*

ERROR at line 1:

ORA-02289: sequence does not exist

 

Table created.

Comment created.

 

Table created.

Comment created.

 

Table created.

Comment created.

 

Sequence created.

 

SQL>

 

 

2.4 Since main script on this Note(profiler.sql) reports on data generated by package DBMS_PROFILER, be sure toprofile your PL/SQL Library prior to try reporting the results. To profile aPL/SQL Library (package, procedure, function or trigger), include in its bodythe two calls to actually start and complete the profiling.

–因为主脚本(profiler.sql)显示的数据是用DBMS_PROFILER包生成的,所以显示结果之前要先profile PL/SQL 代码。

 

Use the example below on any PL/SQL Libraryto profile.

 

BEGIN
DBMS_PROFILER.START_PROFILER(‘any comment to identify this execution’);

DBMS_PROFILER.STOP_PROFILER;
EXCEPTION — this line may exist in your code

END;
/

 

2.5 In order to modify your PL/SQL Library,find first the script that creates it, make a backup, and insert manually theSTART and STOP calls for the profiler. If unable to find the script thatcreates your package, procedure, function or trigger, use the provided scriptprofgsrc.sql executing with PL/SQL Library name as inline parameter:

–为了修改PL/SQL库,首先需要找到脚本,备份后,然后手工的添加START 和 STOP,如果不能找到原代码,可以使用profgsrc.sql 脚本来获取原代码:

 

#sqlplus APPS/<pwd>
SQL> START profgsrc.sql <PL/SQL Library name>;

 

SQL> start ?/rdbms/admin/profgsrc.sql  logon_audit;

Usage:

sqlplus apps/<pwd>

SQL> START profgsrc.sql <PL/SQLLibrary name>

 

NAME                           TYPE

—————————— ————

LOGON_AUDIT                    TRIGGER

 

Generating spool file for TRIGGERLOGON_AUDIT

 

我这里将脚本copy到了$ORACLE_HOME/rdbms/admin下,执行之后其会在sqlplus 的当前目录下,生成一个同名的sql 脚本,脚本就是我们对象的的SQL 代码。

 

 

2.6 Script profgsrc.sql extracts fromUSER_SOURCE the actual source code for the requested PL/SQL Library. Itgenerates a text spool file as a SQL script to regenerate the PL/SQL Library.Make a backup of the spool file before modifying it. Compile your modifiedPL/SQL Library by executing it from SQL*Plus and connecting as your applicationuser.

–profgsrc.sql 从USER_SOURCE中抽取对象的代码。

 

2.7 Once your compiled PL/SQL Librarycontains the START and STOP profiler procedure calls, execute your Library fromyour application. Every execution generates a new run_id which can then bereported on, by using the profiler.sql script.

–当我们在PL/SQL对象中添加完START 和 STOP profiler 调用后,然后启动应用,每次执行都会生成一个run_id, 然后使用profiler.sql 工具,结合run_id 就可以获取对应的报告。

 

生成报告命令

SQL> START profiler.sql <run_id>

 

Where run_id isthe execution id returned by the DBMS_PROFILER (which must be installed first).

If run_id isunknown, execute without any parameter and the script will display a list tochoose from.

–如果没有提供任何参数,那么会自动列出所有的run_id.

 

 

三.使用示例

测试的存储过程如下:

CREATE OR REPLACE PROCEDURE proc_test

AS

BEGIN

   insert intotest_pro  select * from all_objects;

      commit;

   FOR x IN (select * fromall_users where user_id>90)

   LOOP

      DBMS_OUTPUT.put_line(x.username);

   END LOOP;

END proc_test;

 

 

3.1 示例1

 

SQL> set serveroutput on

SQL> begin

2 DBMS_PROFILER.START_PROFILER(‘Dave Test PL/SQL Profiler’);

3  proc_test;

4  DBMS_PROFILER.STOP_PROFILER;

5  end;

6  /

XS$NULL

DAVE3

DAVE2

DAVE1

XEZF

MGMT_VIEW

SYSMAN

ANQING

DVD

 

PL/SQL procedure successfully completed.

 

–调用PL/SQLprofiler.sql

SQL> @?/rdbms/admin/profiler.sql

 

RUNID RUN_DATE             RUN_COMMENT

———- ——————————————————

1 08-FEB-12 14:55:38    Dave TestPL/SQL Profiler

2 08-FEB-12 15:03:11    Dave TestPL/SQL Profiler

 

 

Usage:

sqlplus apps/<pwd>

SQL> START profiler.sql <runid>

 

Enter value for 1: 2

SQL>

–其会自动列出所有RUNID,并启动,我们这里选择上面对应的RUNID,输入2,回车即可。

 

执行完毕后会在SQLPLUS的当前目录下生成一个profiler_2.html的报告,这里的2是RUNID.

 

 

3.2 示例2

 

CREATE OR REPLACE PROCEDURE proc_test2

AS

BEGIN

DBMS_PROFILER.START_PROFILER(‘Dave Test PL/SQL Profiler’);

   insert intotest_pro  select * from all_objects;

   commit;

   FOR x IN (select * fromall_users where user_id>90)

   LOOP

      DBMS_OUTPUT.put_line(x.username);

   END LOOP;

DBMS_PROFILER.STOP_PROFILER;

END proc_test2;

 

 

SQL> exec proc_test2

PL/SQL procedure successfully completed.

 

SQL> @?/rdbms/admin/profiler.sql

RUNID RUN_DATE             RUN_COMMENT

———- —————————————————–

1 08-FEB-12 14:55:38    Dave TestPL/SQL Profiler

2 08-FEB-12 15:03:11    Dave TestPL/SQL Profiler

3 08-FEB-12 15:07:26    Dave TestPL/SQL Profiler

4 08-FEB-12 15:09:40    Dave TestPL/SQL Profiler

–这里这里的RUN_COMMENT内容,其在DBMS_PROFILER.START_PROFILER 中指定,主要用来辨别我们的RUNID.

 

Usage:

sqlplus apps/<pwd>

SQL> START profiler.sql <runid>

 

Enter value for 1: 4

SQL>

 

 

3.3 注意事项:

1.     proftab.sql 脚本需要在PL/SQL对象所在的用户下执行,否则就会报如下错误:

 

SQL> exec proc_test2

BEGIN proc_test2; END;

 

*

ERROR at line 1:

ORA-06528: Error executing PL/SQL profiler

ORA-06512: at”SYS.DBMS_PROFILER”, line 123

ORA-06512: at”SYS.DBMS_PROFILER”, line 132

ORA-06512: at “DAVE.PROC_TEST2”,line 4

ORA-06512: at line 1

 

2.     起初在win7+ oracle 11gR2 下测试,执行没有报错,但最终生成的报告没有数据。后转移到linux + oracle 10gR2 下测试,正常。

后经过测试,发现问题,之前的环境是64位的win7+32位的Oracle 11gR2,后把Oracle 换成64位 11gR2 后正常。这个原因可能与版本有关系,因为也不能排除重新安装解决的某些未知因素。

 

测试结果如下:

采集失败,请手动处理

http://hi.csdn.net/attachment/201202/8/0_132869168524ue.gif

 

3. 两个示例的区别在于示例1没有修改PL/SQL 对象,而是在PL/SQL 代码块中start 和stop了DBMS_PROFILER。 而示例2中是修改了PL/SQL 对象,将DBMS_PROFILER 添加到PL/SQL 对象中。

官网的资料是先备份PL/SQL对象,然后修改,将DBMS_PROFILER 添加到对象里,这样的好处是每次调用时都会统计PROFILER的信息。当然麻烦的是需要修改对象,而在示例1中,在PL/SQL 代码块中启动和关闭,就免于修改PL/SQL 对象。

所以具体采用哪种方式要结合自己的需求,能达到目的即可。

发表回复

您的电子邮箱地址不会被公开。 必填项已用*标注