在11G，sql_trace增加只对某个sql语句生成10046。在一些特殊的环境中我们可能使用到，如：v$sql中sql_fulltext为空，就可以通过配置sql_trace来dump只定的sql语句，而不需要生成更多的其它的trace信息。关于sql_trace对sid,pname,orapid的使用方法，见mos文件。
欢迎大家加入QQ群:
讨论
1，DB与OS版本
|
www.htz.pw > select * from v$version;
BANNER ——————————————————————————– Oracle Database 11g Enterprise Edition Release 11.2.0.3.0 – 64bit Production PL/SQL Release 11.2.0.3.0 – Production CORE 11.2.0.3.0 Production TNS for Linux: Version 11.2.0.3.0 – Production NLSRTL Version 11.2.0.3.0 – Production
[oracle@www.htz.pw admin]$lsb_release -a LSB Version: :core-3.0-amd64:core-3.0-ia32:core-3.0-noarch:graphics-3.0-amd64:graphics-3.0-ia32:graphics-3.0-noarch Distributor ID: RedHatEnterpriseAS Description: Red Hat Enterprise Linux AS release 4 (Nahant Update 8) Release: 4 Codename: NahantUpdate8 |
2，对指定SQL生成10046信息
这里使用了trace_10046_sqlid.sql脚本，脚本内容如下：
|
www.htz.pw > !cat trace_10046_sqlid.sql set echo off set lines 2000 pages 5000 verify off heading on undefine system_or_session; undefine level; undefine sqllist; #如果是多条sql，在sql之间使用,来隔开 alter &&system_or_session set events ‘sql_trace[SQL: &&sqllist ] level &level’; oradebug setmypid oradebug eventdump &&system_or_session;
pause Press return to stop tracing
alter &&system_or_session set events ‘sql_trace[SQL: &&sqllist ] off’; oradebug eventdump &&system_or_session; undefine system_or_session; undefine level; undefine sqllist; |
|
www.htz.pw > conn scott/oracle Connected. www.htz.pw > select /*+ htz*/count(*) from scott.emp;
COUNT(*) ———- 14
www.htz.pw > @find_sql.sql Enter value for sql_text: htz Enter value for sql_id:
SQL_ID CHILD HASH_VALUE PLAN_HASH EXECS ETIME AVG_ETIME USERNAME ————- —— ———- ———- ———- ————- ————- ————- SQLTEXT —————————————————————————————– 1twzndftjst38 0 3004982376 2937609675 1 .00 .00 SCOTT select /*+ htz*/count(*) from scott.emp
www.htz.pw > @trace_10046_sqlid.sql Enter value for system_or_session: system Enter value for sqllist: 1twzndftjst38 Enter value for level: 12
System altered.
Statement processed. sql_trace[SQL: 1twzndftjst38 ] level 12 10949 trace name context forever 28401 trace name context forever, level 1 Press return to stop tracing |
sql_trace配置后，只会对新连接的会话有效，原来已经连接的会话不生效的。
|
打开session2,session3连接的会话执行 www.htz.pw > select /*+ htz*/count(*) from scott.emp;
COUNT(*) ———- 14 可以看到生成下面的日志 PARSING IN CURSOR #182936901240 len=39 dep=0 uid=83 oct=3 lid=83 tim=1409812013408343 hv=3004982376 ad=’883540b0′ sqlid=’1twzndftjst38′ select /*+ htz*/count(*) from scott.emp END OF STMT EXEC #182936901240:c=0,e=33,p=0,cr=0,cu=0,mis=0,r=0,dep=0,og=1,plh=2937609675,tim=1409812013408342 WAIT #182936901240: nam=’SQL*Net message to client’ ela= 2 driver id=1650815232 #bytes=1 p3=0 obj#=-1 tim=1409812013408407 FETCH #182936901240:c=0,e=40,p=0,cr=1,cu=0,mis=0,r=1,dep=0,og=1,plh=2937609675,tim=1409812013408466 STAT #182936901240 id=1 cnt=1 pid=0 pos=1 obj=0 op=’SORT AGGREGATE (cr=1 pr=0 pw=0 time=35 us)’ STAT #182936901240 id=2 cnt=14 pid=1 pos=1 obj=79714 op=’INDEX FULL SCAN PK_EMP (cr=1 pr=0 pw=0 time=34 us cost=1 size=0 card=14)’ FETCH #182936901240:c=0,e=2,p=0,cr=0,cu=0,mis=0,r=0,dep=0,og=0,plh=2937609675,tim=1409812013408672 WAIT #182936901240: nam=’SQL*Net message to client’ ela= 1 driver id=1650815232 #bytes=1 p3=0 obj#=-1 tim=1409812013408689 已经trace成功 生成了exec,wait,stat的信息，因为SQL没有绑定变量，所以未见绑定变量信息 |
3 关闭trace
|
回来session1 www.htz.pw > @trace_10046_sqlid.sql Enter value for system_or_session: system Enter value for sqllist: 1twzndftjst38 Enter value for level: 12
System altered.
Statement processed. sql_trace[SQL: 1twzndftjst38 ] level 12 10949 trace name context forever 28401 trace name context forever, level 1 Press return to stop tracing 这是之前的显示的内容 按回车，取消trace。
System altered.
10949 trace name context forever 28401 trace name context forever, level 1 |
4，验证是否关闭trace
回到session 2窗口中
|
www.htz.pw > /
COUNT(*) ———- 14 仍然生成了trace信息 PARSING IN CURSOR #182936887640 len=39 dep=0 uid=83 oct=3 lid=83 tim=1409812149842911 hv=3004982376 ad=’883540b0′ sqlid=’1twzndftjst38′ select /*+ htz*/count(*) from scott.emp END OF STMT EXEC #182936887640:c=0,e=48,p=0,cr=0,cu=0,mis=0,r=0,dep=0,og=1,plh=2937609675,tim=1409812149842909 WAIT #182936887640: nam=’SQL*Net message to client’ ela= 9 driver id=1650815232 #bytes=1 p3=0 obj#=-1 tim=1409812149842978 FETCH #182936887640:c=999,e=81,p=0,cr=1,cu=0,mis=0,r=1,dep=0,og=1,plh=2937609675,tim=1409812149843079 STAT #182936887640 id=1 cnt=1 pid=0 pos=1 obj=0 op=’SORT AGGREGATE (cr=1 pr=0 pw=0 time=83 us)’ STAT #182936887640 id=2 cnt=14 pid=1 pos=1 obj=79714 op=’INDEX FULL SCAN PK_EMP (cr=1 pr=0 pw=0 time=67 us cost=1 size=0 card=14)’ FETCH #182936887640:c=0,e=2,p=0,cr=0,cu=0,mis=0,r=0,dep=0,og=0,plh=2937609675,tim=1409812149843273 WAIT #182936887640: nam=’SQL*Net message to client’ ela= 1 driver id=1650815232 #bytes=1 p3=0 obj#=-1 tim=1409812149843301
执行一条其它的SQL后再回来执行原来的SQL www.htz.pw > select count(*) from scott.emp;
COUNT(*) ———- 14 再次执行trace的sql语句 www.htz.pw > select /*+ htz*/count(*) from scott.emp;
COUNT(*) ———- 14 在trace中没有看到任何的TRACE了内容了 |
回到session 3窗口中
|
www.htz.pw > /
COUNT(*) ———- 14
PARSING IN CURSOR #182936910800 len=39 dep=0 uid=83 oct=3 lid=83 tim=1409812294019897 hv=3004982376 ad=’883540b0′ sqlid=’1twzndftjst38′ select /*+ htz*/count(*) from scott.emp END OF STMT EXEC #182936910800:c=0,e=56,p=0,cr=0,cu=0,mis=0,r=0,dep=0,og=1,plh=2937609675,tim=1409812294019895 WAIT #182936910800: nam=’SQL*Net message to client’ ela= 2 driver id=1650815232 #bytes=1 p3=0 obj#=-1 tim=1409812294019964 FETCH #182936910800:c=0,e=35,p=0,cr=1,cu=0,mis=0,r=1,dep=0,og=1,plh=2937609675,tim=1409812294020018 STAT #182936910800 id=1 cnt=1 pid=0 pos=1 obj=0 op=’SORT AGGREGATE (cr=1 pr=0 pw=0 time=42 us)’ STAT #182936910800 id=2 cnt=14 pid=1 pos=1 obj=79714 op=’INDEX FULL SCAN PK_EMP (cr=1 pr=0 pw=0 time=40 us cost=1 size=0 card=14)’ FETCH #182936910800:c=0,e=2,p=0,cr=0,cu=0,mis=0,r=0,dep=0,og=0,plh=2937609675,tim=1409812294020203 WAIT #182936910800: nam=’SQL*Net message to client’ ela= 1 driver id=1650815232 #bytes=1 p3=0 obj#=-1 tim=1409812294020231
还在成trace，所以排除窗口2是由于时间到而停止trace sql内容 再次执行其它SQL www.htz.pw > select count(*) from SEQ;
COUNT(*) ———- 225
www.htz.pw > select /*+ htz*/count(*) from scott.emp;
COUNT(*) ———- 14 同时trace文件中再也没有生成sqlid的trace内容
*** 2014-09-04 14:33:02.697 WAIT #182936910800: nam=’SQL*Net message from client’ ela= 88676734 driver id=1650815232 #bytes=1 p3=0 obj#=-1 tim=1409812382696974 CLOSE #182936910800:c=0,e=12,dep=0,type=1,tim=1409812382697082 |
通过session 2,session 3操作，可以猜测，取消trace后，对会话来说，需要执行一条非trace的SQL语句后才会生效，如果一直执行trace sql，那么会一直生成10046的trace信息的。
sql_trace开启对sql语法的trace：等您坐沙发呢！