[teacher Zhao Yuqiang] use Oracle's tracking file

One 、 What is a tracking file ？

         The trace file contains a lot of detailed diagnostic and debugging information . Through the interpretation and analysis of tracking files , We can position the problem 、 Analyze and solve problems . From the source of the trace file , Tracking files can be divided into two categories ： One is generated intentionally by database operators ; The other is due to an abnormal error , Automatically generated by the database . For the latter category , Only right Oracle The in-house technical support staff are useful , But for us , Most of them don't understand . The former category , That's what we use all the time , Help us analyze 、 Adjust and optimize application performance , Deal with and solve problems .

        So where can I find the trace file ？ By querying the data dictionary v$diag_info You can determine the storage path of the trace file , As shown below .

select * from v$diag_info;

        Entry directory /u01/app/oracle/diag/rdbms/orcl/orcl/trace You can see that trc Trace file for suffix , As shown in the figure below .

Two 、 Naming rules for trace files

         The name of a trace file generally consists of the following parts ：

• ORACLE_SID
• Fixed character
• Process of server ID Number
• File suffix .trc
• The parts are connected by the following lines .

for example ：orcl_mmon_12210.trc, among ：“orcl" It is the database in this environment SID,"12210" The server process used to generate the trace file session ID Number . How to know my ORACLE_SID And the server process used by the session ID Well ？

3、 ... and 、 How to determine the trace file ？

        For the convenience of demonstration , We give it to an ordinary user scott grant dba Role .

1、 Log in with administration , And conferred scott grant dba Role

[[email protected] ~]$ sqlplus / as sysdba

SQL> grant dba to scott;

Grant succeeded.

SQL> 

2、 determine Oracle SID, As shown below . there SID Namely ：orcl

SQL> select instance_name from V$instance;

INSTANCE_NAME
----------------
orcl

SQL> 

3、 Switch to scott user , And make sure the conversation ID

SQL> conn scott/tiger
Connected.
SQL> select sid from v$mystat where rownum=1;

       SID
----------
	70

SQL> 

4、 According to the conversation ID, Determine the address information of the session

SQL> select paddr from v$session where sid=70;

PADDR
----------------
000000006DAB6588

SQL> 

5、 According to the address information of the session , Determine the process number of the operating system

SQL> select spid from v$process where addr='000000006DAB6588';

SPID
------------------------
54685

SQL> 

         Entry directory /u01/app/oracle/diag/rdbms/orcl/orcl/trace Will find , There is no such thing as 54685 Tracking files for , The reason is that to use the trace file, you need to manually turn on the trace of the session .

6、 Turn on session tracking

SQL> alter session set sql_trace=true;

Session altered.

SQL> 

7、 Perform a simple SQL sentence , And check /u01/app/oracle/diag/rdbms/orcl/orcl/trace Catalog , At this point, you can see the generated trace file .

[[email protected] trace]$ pwd
/u01/app/oracle/diag/rdbms/orcl/orcl/trace
[[email protected] trace]$ ls *54685.trc
orcl_ora_54685.trc
[[email protected] trace]$ 

Four 、 Use trace files to diagnose SQL

According to tracking, for diagnosis SQL Sentences are very useful , Here is a simple example to illustrate .

1、 Do the following SQL sentence

select * from scott.emp where deptno=10;

select * from scott.emp where deptno=20;

select * from scott.emp where deptno=30;

        These three SQL Check separately 10、20 and 30 Employees of Department . Through observation, we found that , These three SQL except where The parameter values of the condition are different , The other parts are the same . In this way SQL Statements are called “ Repetitive SQL”. If there is a lot of duplication in the database SQL, Will make every time in the execution will be carried out SQL Parsing , And then generate the execution plan . This will affect the performance of the database .

        The following is to verify the above conclusion by tracking the file .

2、 Because the tracking of the session was turned on earlier , If tracking is no longer needed , You need to turn it off manually .

SQL> alter session set sql_trace=false;

Session altered.

SQL> 

3、 Use tkprof Tools to format trace files

[[email protected] trace]$ tkprof orcl_ora_54685.trc /home/oracle/a.txt sys=no sort=fchela

TKPROF: Release 12.2.0.1.0 - Development on Mon Jun 28 10:37:48 2021

Copyright (c) 1982, 2017, Oracle and/or its affiliates.  All rights reserved.


[[email protected] trace]$ 

4、 View the generated a.txt file , As shown below ：

SQL ID: 1mvxd868z75nf Plan Hash: 3956160932
select *
from scott.emp where deptno=30

SQL ID: 2nbac4n9hnzth Plan Hash: 3956160932
select *
from scott.emp where deptno=20

SQL ID: 062r5atccuyv4 Plan Hash: 3956160932
select *
from scott.emp where deptno=10

        It can be seen that although there are three SQL Corresponding SQL ID Dissimilarity , But generated Plan Hash It's the same . This shows the three points SQL The implementation plan is the same . In that case , We can rewrite these three by binding variables SQL. Let these three SQL When a statement is executed , You don't have to generate an execution plan every time . Just reuse the execution plan generated for the first time . To improve performance .