[20190416]完善shared latch测试脚本2.txt

来源：这里教程网时间：2026-03-03 13:17:45 作者：

[20190416]完善shared latch测试脚本2.txt --//昨天测试shared latch,链接:http://blog.itpub.net/267265/viewspace-2641414/,感觉有点开窍了.^_^. http://andreynikolaev.wordpress.com/2010/11/17/shared-latch-behaves-like-enqueue/ For the shared latches Oracle 10g uses kslgetsl(laddr, wait, why, where, mode) function. Oracle 11g has kslgetsl_w() function with the same interface, but internally uses ksl_get_shared_latch(). Like in my previous post, I guess the meaning of kslgetsl() arguments as: --//对于共享锁存，Oracle 10g使用kslgetsl(laddr，wait，why，where，mode)函数。Oracle 11g具有相同接口的kslgetsl_w()函数，但 --//在内部使用ksl_get_share_latch()。与上一篇文章一样，我认为kslgetsl()参数的含义是： --//注:我以前一直以为还是kslgetsl,原来11g已经改为kslgetsl_w,不过内部使用还是ksl_get_shared_latch(). laddress -- address of latch in SGA wait -- flag. If not 0, then willing-to-wait latch get why -- context why the latch is acquired at this where. where -- location from where the latch is acquired (x$ksllw.indx) And the last one is: mode – Exclusive or shared mode the mode argument took only two values: 8 -- "SHARED" 16 -- "EXCLUSIVE" --//我当时的测试针对'gcs partitioned table hash' latch,完善修改测试脚本,增加一些通用性. 1.环境: SYS@book> @ ver1 PORT_STRING VERSION BANNER ------------------------------ -------------- -------------------------------------------------------------------------------- x86_64/Linux 2.4.xx 11.2.0.4.0 Oracle Database 11g Enterprise Edition Release 11.2.0.4.0 - 64bit Production $ cat peek.sh #! /bib/bash # 参数如下:latch_name Monitoring_duration sqlplus -s -l / as sysdba <<EOF col laddr new_value laddr SELECT sysdate,addr laddr FROM v\$latch_parent WHERE NAME='$1'; oradebug setmypid $(seq $2|xargs -I{} echo -e 'oradebug peek 0x&laddr 8\nhost sleep 1' ) EOF $ cat shared_latch.txt /* 参数如下: @ latch.txt latch_name willing why where mode sleep_num */ connect / as sysdba col laddr new_value laddr col vmode new_value vmode select decode(lower('&&5'),'s',8,'x',16,'8',8,'16',16) vmode from dual ; SELECT addr laddr FROM v$latch_parent WHERE NAME='&&1'; oradebug setmypid oradebug call kslgetsl_w 0x&laddr &&2 &&3 &&4 &vmode host sleep &&6 oradebug call kslfre 0x&laddr exit $ cat latch_free.sql /* This file is part of demos for "Contemporary Latch Internals" seminar v.18.09.2010 Andrey S. Nikolaev (Andrey.Nikolaev@rdtex.ru) http://AndreyNikolaev.wordpress.com This query shows trees of processes currently holding and waiting for latches Tree output enumerates these processes and latches as following: Process <PID1> <latch1 holding by PID1> <processes waiting for latch1> ... <latch2 holding by PID1> <processes waiting for latch2> ... Process <PID2> ... */ set head off set feedback off set linesize 120 select sysdate from dual; select LPAD(' ', (LEVEL - 1) ) ||case when latch_holding is null then 'Process '||pid else 'holding: '||latch_holding||' "'||name||'" lvl='||level#||' whr='||whr||' why='||why ||', SID='||sid end || case when latch_waiting is not null then ', waiting for: '||latch_waiting||' whr='||whr||' why='||why end latchtree from ( /* Latch holders */ select ksuprpid pid,ksuprlat latch_holding, null latch_waiting, to_char(ksuprpid) parent_id, rawtohex(ksuprlat) id, ksuprsid sid,ksuprllv level#,ksuprlnm name,ksuprlmd mode_,ksulawhy why,ksulawhr whr from x$ksuprlat union all /* Latch waiters */ select indx pid,null latch_holding, ksllawat latch_waiting,rawtohex(ksllawat) parent_id,to_char(indx) id, null,null,null,null,ksllawhy why,ksllawer whr from x$ksupr where ksllawat !='00' union all /* The roots of latch trees: processes holding latch but not waiting for latch */ select pid, null, null, null, to_char(pid),null,null,null,null,null,null from ( select distinct ksuprpid pid from x$ksuprlat minus select indx pid from x$ksupr where ksllawat !='00') ) latch_op connect by prior id=parent_id start with parent_id is null; $ cat /usr/local/bin/timestamp.pl #!/usr/bin/perl while (<>) { ($sec,$min,$hour,$mday,$mon,$year,$wday,$yday,$isdst) = localtime(); printf("%02d:%02d:%02d", $hour, $min, $sec); print ": $_"; #print localtime() . ": $_"; } --//使用timestamp.pl在开始标注时间.这样更加清晰. 2.测试: --//补充测试 X mode,S mode ,X 模式的情况. $ cat f1.sh #! /bin/bash source peek.sh 'gcs partitioned table hash' 30 | timestamp.pl >| /tmp/peeks.txt & seq 30 | xargs -I{} echo -e 'sqlplus -s -l / as sysdba <<< @latch_free\nsleep 1' | bash >| /tmp/latch_free.txt & # 参数如下: @ latch.txt latch_name willing why where mode sleep_num sqlplus /nolog @ shared_latch.txt 'gcs partitioned table hash' 1 4 5 x 6 > /dev/null & sleep 2 sqlplus /nolog @ shared_latch.txt 'gcs partitioned table hash' 1 4 5 s 6 > /dev/null & sqlplus /nolog @ shared_latch.txt 'gcs partitioned table hash' 1 4 5 s 6 > /dev/null & sleep 0.1 sqlplus /nolog @ shared_latch.txt 'gcs partitioned table hash' 1 4 5 x 6 > /dev/null & wait $ grep -v '^.*: $' /tmp/peeks.txt 09:28:38: SYSDATE LADDR 09:28:38: ------------------- ---------------- 09:28:38: 2019-04-16 09:28:38 0000000060018A18 09:28:38: Statement processed. 09:28:38: [060018A18, 060018A20) = 0000001C 20000000 09:28:39: [060018A18, 060018A20) = 0000001C 20000000 09:28:40: [060018A18, 060018A20) = 0000001C 20000000 09:28:41: [060018A18, 060018A20) = 0000001C 20000000 09:28:42: [060018A18, 060018A20) = 0000001C 20000000 09:28:43: [060018A18, 060018A20) = 0000001C 20000000 09:28:44: [060018A18, 060018A20) = 00000001 00000000 09:28:45: [060018A18, 060018A20) = 00000001 00000000 09:28:46: [060018A18, 060018A20) = 00000001 00000000 09:28:47: [060018A18, 060018A20) = 00000001 00000000 09:28:48: [060018A18, 060018A20) = 00000001 00000000 09:28:49: [060018A18, 060018A20) = 00000001 00000000 09:28:50: [060018A18, 060018A20) = 00000001 00000000 09:28:51: [060018A18, 060018A20) = 00000001 00000000 09:28:52: [060018A18, 060018A20) = 00000001 00000000 09:28:53: [060018A18, 060018A20) = 00000001 00000000 09:28:54: [060018A18, 060018A20) = 00000001 00000000 09:28:55: [060018A18, 060018A20) = 00000001 00000000 09:28:56: [060018A18, 060018A20) = 0000001F 20000000 09:28:57: [060018A18, 060018A20) = 0000001F 20000000 09:28:58: [060018A18, 060018A20) = 0000001F 20000000 09:28:59: [060018A18, 060018A20) = 0000001F 20000000 09:29:00: [060018A18, 060018A20) = 0000001F 20000000 09:29:01: [060018A18, 060018A20) = 0000001F 20000000 09:29:02: [060018A18, 060018A20) = 00000000 00000000 09:29:03: [060018A18, 060018A20) = 00000000 00000000 09:29:04: [060018A18, 060018A20) = 00000000 00000000 09:29:05: [060018A18, 060018A20) = 00000000 00000000 09:29:06: [060018A18, 060018A20) = 00000000 00000000 09:29:07: [060018A18, 060018A20) = 00000000 00000000 $ grep -v '^.*: $' /tmp/peeks.txt | cut -c10- | uniq -c 1 SYSDATE LADDR 1 ------------------- ---------------- 1 2019-04-16 09:28:38 0000000060018A18 1 Statement processed. 6 [060018A18, 060018A20) = 0000001C 20000000 12 [060018A18, 060018A20) = 00000001 00000000 6 [060018A18, 060018A20) = 0000001F 20000000 6 [060018A18, 060018A20) = 00000000 00000000 --//仅仅注意一个细节,peek值的后4位并没有出现0x40000000的情况. $ cat /tmp/latch_free.txt 2019-04-16 09:28:38 2019-04-16 09:28:39 Process 28 holding: 0000000060018A18 "gcs partitioned table hash" lvl=6 whr=5 why=4, SID=58 2019-04-16 09:28:40 Process 28 holding: 0000000060018A18 "gcs partitioned table hash" lvl=6 whr=5 why=4, SID=58 Process 29, waiting for: 0000000060018A18 whr=5 why=4 Process 30, waiting for: 0000000060018A18 whr=5 why=4 Process 31, waiting for: 0000000060018A18 whr=5 why=4 2019-04-16 09:28:41 Process 28 holding: 0000000060018A18 "gcs partitioned table hash" lvl=6 whr=5 why=4, SID=58 Process 29, waiting for: 0000000060018A18 whr=5 why=4 Process 30, waiting for: 0000000060018A18 whr=5 why=4 Process 31, waiting for: 0000000060018A18 whr=5 why=4 2019-04-16 09:28:42 Process 28 holding: 0000000060018A18 "gcs partitioned table hash" lvl=6 whr=5 why=4, SID=58 Process 29, waiting for: 0000000060018A18 whr=5 why=4 Process 30, waiting for: 0000000060018A18 whr=5 why=4 Process 31, waiting for: 0000000060018A18 whr=5 why=4 2019-04-16 09:28:44 Process 28 holding: 0000000060018A18 "gcs partitioned table hash" lvl=6 whr=5 why=4, SID=58 Process 29, waiting for: 0000000060018A18 whr=5 why=4 Process 30, waiting for: 0000000060018A18 whr=5 why=4 Process 31, waiting for: 0000000060018A18 whr=5 why=4 2019-04-16 09:28:45 Process 29 holding: 0000000060018A18 "gcs partitioned table hash" lvl=6 whr=5 why=4, SID=72 Process 30, waiting for: 0000000060018A18 whr=5 why=4 Process 31, waiting for: 0000000060018A18 whr=5 why=4 2019-04-16 09:28:46 Process 29 holding: 0000000060018A18 "gcs partitioned table hash" lvl=6 whr=5 why=4, SID=72 Process 30, waiting for: 0000000060018A18 whr=5 why=4 Process 31, waiting for: 0000000060018A18 whr=5 why=4 2019-04-16 09:28:47 Process 29 holding: 0000000060018A18 "gcs partitioned table hash" lvl=6 whr=5 why=4, SID=72 Process 30, waiting for: 0000000060018A18 whr=5 why=4 Process 31, waiting for: 0000000060018A18 whr=5 why=4 2019-04-16 09:28:48 Process 29 holding: 0000000060018A18 "gcs partitioned table hash" lvl=6 whr=5 why=4, SID=72 Process 30, waiting for: 0000000060018A18 whr=5 why=4 Process 31, waiting for: 0000000060018A18 whr=5 why=4 2019-04-16 09:28:49 Process 29 holding: 0000000060018A18 "gcs partitioned table hash" lvl=6 whr=5 why=4, SID=72 Process 30, waiting for: 0000000060018A18 whr=5 why=4 Process 31, waiting for: 0000000060018A18 whr=5 why=4 2019-04-16 09:28:50 Process 29 holding: 0000000060018A18 "gcs partitioned table hash" lvl=6 whr=5 why=4, SID=72 Process 30, waiting for: 0000000060018A18 whr=5 why=4 Process 31, waiting for: 0000000060018A18 whr=5 why=4 2019-04-16 09:28:51 Process 30 holding: 0000000060018A18 "gcs partitioned table hash" lvl=6 whr=5 why=4, SID=86 Process 31, waiting for: 0000000060018A18 whr=5 why=4 2019-04-16 09:28:52 Process 30 holding: 0000000060018A18 "gcs partitioned table hash" lvl=6 whr=5 why=4, SID=86 Process 31, waiting for: 0000000060018A18 whr=5 why=4 2019-04-16 09:28:53 Process 30 holding: 0000000060018A18 "gcs partitioned table hash" lvl=6 whr=5 why=4, SID=86 Process 31, waiting for: 0000000060018A18 whr=5 why=4 2019-04-16 09:28:54 Process 30 holding: 0000000060018A18 "gcs partitioned table hash" lvl=6 whr=5 why=4, SID=86 Process 31, waiting for: 0000000060018A18 whr=5 why=4 2019-04-16 09:28:56 Process 30 holding: 0000000060018A18 "gcs partitioned table hash" lvl=6 whr=5 why=4, SID=86 Process 31, waiting for: 0000000060018A18 whr=5 why=4 2019-04-16 09:28:57 Process 31 holding: 0000000060018A18 "gcs partitioned table hash" lvl=6 whr=5 why=4, SID=101 2019-04-16 09:28:58 Process 31 holding: 0000000060018A18 "gcs partitioned table hash" lvl=6 whr=5 why=4, SID=101 2019-04-16 09:28:59 Process 31 holding: 0000000060018A18 "gcs partitioned table hash" lvl=6 whr=5 why=4, SID=101 2019-04-16 09:29:00 Process 31 holding: 0000000060018A18 "gcs partitioned table hash" lvl=6 whr=5 why=4, SID=101 2019-04-16 09:29:01 Process 31 holding: 0000000060018A18 "gcs partitioned table hash" lvl=6 whr=5 why=4, SID=101 2019-04-16 09:29:02 Process 31 holding: 0000000060018A18 "gcs partitioned table hash" lvl=6 whr=5 why=4, SID=101 2019-04-16 09:29:03 2019-04-16 09:29:04 --//结果我就不再讲解了. --//不过有点奇怪的是,如果修改f1.sh如下: $ cat f1.sh #! /bin/bash source peek.sh 'gcs partitioned table hash' 30 | timestamp.pl >| /tmp/peeks.txt & seq 30 | xargs -I{} echo -e 'sqlplus -s -l / as sysdba <<< @latch_free\nsleep 1' | bash >| /tmp/latch_free.txt & # 参数如下: @ latch.txt latch_name willing why where mode sleep_num sqlplus /nolog @ shared_latch.txt 'gcs partitioned table hash' 1 4 5 x 6 > /dev/null & sleep 2 sqlplus /nolog @ shared_latch.txt 'gcs partitioned table hash' 1 7 8 s 6 > /dev/null & sqlplus /nolog @ shared_latch.txt 'gcs partitioned table hash' 1 9 10 s 6 > /dev/null & ##sleep 0.1 sqlplus /nolog @ shared_latch.txt 'gcs partitioned table hash' 1 11 12 x 6 > /dev/null & wait --//注解sleep 0.01秒.结果如下: $ grep -v '^.*: $' /tmp/peeks.txt | cut -c10- | uniq -c 1 SYSDATE LADDR 1 ------------------- ---------------- 1 2019-04-16 09:33:37 0000000060018A18 1 Statement processed. 6 [060018A18, 060018A20) = 0000001D 20000000 6 [060018A18, 060018A20) = 0000001F 20000000 12 [060018A18, 060018A20) = 00000001 00000000 6 [060018A18, 060018A20) = 00000000 00000000 --//给人的感觉优先处理X mode 锁,然后才是S mode.我测试多次结果都一样.如果修改如下: $ cat f1.sh #! /bin/bash source peek.sh 'gcs partitioned table hash' 30 | timestamp.pl >| /tmp/peeks.txt & seq 30 | xargs -I{} echo -e 'sqlplus -s -l / as sysdba <<< @latch_free\nsleep 1' | bash >| /tmp/latch_free.txt & # 参数如下: @ latch.txt latch_name willing why where mode sleep_num sqlplus /nolog @ shared_latch.txt 'gcs partitioned table hash' 1 4 5 x 6 > /dev/null & sleep 2 sqlplus /nolog @ shared_latch.txt 'gcs partitioned table hash' 1 7 8 s 6 > /dev/null & sqlplus /nolog @ shared_latch.txt 'gcs partitioned table hash' 1 11 12 x 6 > /dev/null & sqlplus /nolog @ shared_latch.txt 'gcs partitioned table hash' 1 9 10 s 6 > /dev/null & ##sleep 0.1 wait $ grep -v '^.*: $' /tmp/peeks.txt | cut -c10- | uniq -c 1 SYSDATE LADDR 1 ------------------- ---------------- 1 2019-04-16 09:37:56 0000000060018A18 1 Statement processed. 6 [060018A18, 060018A20) = 0000001B 20000000 6 [060018A18, 060018A20) = 00000001 00000000 6 [060018A18, 060018A20) = 0000001E 20000000 6 [060018A18, 060018A20) = 00000001 00000000 6 [060018A18, 060018A20) = 00000000 00000000 --//仅仅注意一个细节,peek值的后4位并没有出现0x40000000的情况. --//latch_free.txt 2019-04-16 09:37:56 2019-04-16 09:37:57 Process 27 holding: 0000000060018A18 "gcs partitioned table hash" lvl=6 whr=5 why=4, SID=44 2019-04-16 09:37:58 Process 27 holding: 0000000060018A18 "gcs partitioned table hash" lvl=6 whr=5 why=4, SID=44 Process 29, waiting for: 0000000060018A18 whr=10 why=9 Process 30, waiting for: 0000000060018A18 whr=12 why=11 Process 31, waiting for: 0000000060018A18 whr=8 why=7 --//同时执行的sql语句,总是最后1个先启动执行. 2019-04-16 09:37:59 Process 27 holding: 0000000060018A18 "gcs partitioned table hash" lvl=6 whr=5 why=4, SID=44 Process 29, waiting for: 0000000060018A18 whr=10 why=9 Process 30, waiting for: 0000000060018A18 whr=12 why=11 Process 31, waiting for: 0000000060018A18 whr=8 why=7 2019-04-16 09:38:01 Process 27 holding: 0000000060018A18 "gcs partitioned table hash" lvl=6 whr=5 why=4, SID=44 Process 29, waiting for: 0000000060018A18 whr=10 why=9 Process 30, waiting for: 0000000060018A18 whr=12 why=11 Process 31, waiting for: 0000000060018A18 whr=8 why=7 2019-04-16 09:38:02 Process 27 holding: 0000000060018A18 "gcs partitioned table hash" lvl=6 whr=5 why=4, SID=44 Process 29, waiting for: 0000000060018A18 whr=10 why=9 Process 30, waiting for: 0000000060018A18 whr=12 why=11 Process 31, waiting for: 0000000060018A18 whr=8 why=7 2019-04-16 09:38:03 Process 29 holding: 0000000060018A18 "gcs partitioned table hash" lvl=6 whr=10 why=9, SID=72 Process 30, waiting for: 0000000060018A18 whr=12 why=11 Process 31, waiting for: 0000000060018A18 whr=8 why=7 --//总之,有了这些脚本大家可以自行组合测试.我仅仅测试 --//SSS XSS SXS XXX --//这里算是XSX,是否后4位出现的规律与第1次持有的mode是shared还是EXCLUSIVE有关. --//视乎peek看到的值与入队时当前持有的状态shared,exclusive有关. $ cat g1.sh #! /bin/bash source peek.sh 'gcs partitioned table hash' 30 | timestamp.pl >| /tmp/peeks.txt & seq 30 | xargs -I{} echo -e 'sqlplus -s -l / as sysdba <<< @latch_free\nsleep 1' | bash >| /tmp/latch_free.txt & # 参数如下: @ latch.txt latch_name willing why where mode sleep_num sqlplus /nolog @ shared_latch.txt 'gcs partitioned table hash' 1 4 5 s 6 > /dev/null & sleep 2 sqlplus /nolog @ shared_latch.txt 'gcs partitioned table hash' 1 7 8 x 6 > /dev/null & sleep 2 sqlplus /nolog @ shared_latch.txt 'gcs partitioned table hash' 1 9 10 x 6 > /dev/null & wait $ grep -v '^.*: $' /tmp/peeks.txt | cut -c10- | uniq -c 1 SYSDATE LADDR 1 ------------------- ---------------- 1 2019-04-16 09:48:14 0000000060018A18 1 Statement processed. 2 [060018A18, 060018A20) = 00000001 00000000 4 [060018A18, 060018A20) = 00000001 40000000 6 [060018A18, 060018A20) = 0000001D 20000000 6 [060018A18, 060018A20) = 0000001E 20000000 12 [060018A18, 060018A20) = 00000000 00000000 --//这样出现后4位是0x40000000好像仅仅一种可能,就是当前持有S mode,入队X mode时才会出现这样的情况. --//再做一个例子: $ cat g1.sh #! /bin/bash source peek.sh 'gcs partitioned table hash' 30 | timestamp.pl >| /tmp/peeks.txt & seq 30 | xargs -I{} echo -e 'sqlplus -s -l / as sysdba <<< @latch_free\nsleep 1' | bash >| /tmp/latch_free.txt & # 参数如下: @ latch.txt latch_name willing why where mode sleep_num sqlplus /nolog @ shared_latch.txt 'gcs partitioned table hash' 1 4 5 s 6 > /dev/null & sleep 2 sqlplus /nolog @ shared_latch.txt 'gcs partitioned table hash' 1 7 8 s 6 > /dev/null & sleep 2 sqlplus /nolog @ shared_latch.txt 'gcs partitioned table hash' 1 9 10 x 6 > /dev/null & sleep 1 sqlplus /nolog @ shared_latch.txt 'gcs partitioned table hash' 1 11 12 x 6 > /dev/null & wait $ grep -v '^.*: $' /tmp/peeks.txt | cut -c10- | uniq -c 1 SYSDATE LADDR 1 ------------------- ---------------- 1 2019-04-16 09:58:45 0000000060018A18 1 Statement processed. 2 [060018A18, 060018A20) = 00000001 00000000 2 [060018A18, 060018A20) = 00000002 00000000 2 [060018A18, 060018A20) = 00000002 40000000 2 [060018A18, 060018A20) = 00000001 40000000 6 [060018A18, 060018A20) = 0000001E 20000000 6 [060018A18, 060018A20) = 0000001F 20000000 10 [060018A18, 060018A20) = 00000000 00000000 --//出现2次后4位是0x40000000的情况.可以理解这样模式持有S mode的情况下,有X mode入队,才会出现这样的情况. $ cat h1.sh #! /bin/bash source peek.sh 'gcs partitioned table hash' 30 | timestamp.pl >| /tmp/peeks.txt & seq 30 | xargs -I{} echo -e 'sqlplus -s -l / as sysdba <<< @latch_free\nsleep 1' | bash >| /tmp/latch_free.txt & # 参数如下: @ latch.txt latch_name willing why where mode sleep_num sqlplus /nolog @ shared_latch.txt 'gcs partitioned table hash' 1 4 5 x 6 > /dev/null & sleep 2 sqlplus /nolog @ shared_latch.txt 'gcs partitioned table hash' 1 6 7 s 6 > /dev/null & sleep 4.1 sqlplus /nolog @ shared_latch.txt 'gcs partitioned table hash' 1 9 10 x 6 > /dev/null & wait $ grep -v '^.*: $' /tmp/peeks.txt | cut -c10- | uniq -c 1 SYSDATE LADDR 1 ------------------- ---------------- 1 2019-04-16 10:11:26 0000000060018A18 1 Statement processed. 6 [060018A18, 060018A20) = 0000001C 20000000 1 [060018A18, 060018A20) = 00000001 00000000 5 [060018A18, 060018A20) = 00000001 40000000 6 [060018A18, 060018A20) = 0000001C 20000000 12 [060018A18, 060018A20) = 00000000 00000000 --//出现1次.最后我感觉脚本写的还是不好,每次都覆盖前面的测试结果.加入时间变量,修改如下: $ cat g1.sh #! /bin/bash zdate=$(date '+%H%M%S') echo $zdate source peek.sh 'gcs partitioned table hash' 30 | timestamp.pl >| /tmp/peeks_${zdate}.txt & seq 30 | xargs -I{} echo -e 'sqlplus -s -l / as sysdba <<< @latch_free\nsleep 1' | bash >| /tmp/latch_free_${zdate}.txt & # 参数如下: @ latch.txt latch_name willing why where mode sleep_num sqlplus /nolog @ shared_latch.txt 'gcs partitioned table hash' 1 4 5 s 6 > /dev/null & sleep 2 sqlplus /nolog @ shared_latch.txt 'gcs partitioned table hash' 1 7 8 x 6 > /dev/null & sleep 2 sqlplus /nolog @ shared_latch.txt 'gcs partitioned table hash' 1 9 10 s 6 > /dev/null & sleep 2.1 sqlplus /nolog @ shared_latch.txt 'gcs partitioned table hash' 1 11 12 x 6 > /dev/null & wait --//大家可以自行组合,peek看到的值也许不重要,只要知道请求都是S mode下不会阻塞. --//X模式下,请求的S模式都会导致串行化.同时S mode也会阻塞X mode就足够了.最后测试一种情况看看: $ cat i1.sh #! /bin/bash zdate=$(date '+%H%M%S') echo $zdate source peek.sh 'gcs partitioned table hash' 36 | timestamp.pl >| /tmp/peeks_${zdate}.txt & seq 36 | xargs -I{} echo -e 'sqlplus -s -l / as sysdba <<< @latch_free\nsleep 1' | bash >| /tmp/latch_free_${zdate}.txt & # 参数如下: @ latch.txt latch_name willing why where mode sleep_num sqlplus /nolog @ shared_latch.txt 'gcs partitioned table hash' 1 4 5 x 6 > /dev/null & sleep 2 sqlplus /nolog @ shared_latch.txt 'gcs partitioned table hash' 1 7 8 s 6 > /dev/null & sleep 2 sqlplus /nolog @ shared_latch.txt 'gcs partitioned table hash' 1 9 10 s 6 > /dev/null & sleep 2.1 sqlplus /nolog @ shared_latch.txt 'gcs partitioned table hash' 1 11 12 s 6 > /dev/null & wait $ grep -v '^.*: $' /tmp/peeks_102719.txt | cut -c10- | uniq -c 1 SYSDATE LADDR 1 ------------------- ---------------- 1 2019-04-16 10:27:19 0000000060018A18 1 Statement processed. 6 [060018A18, 060018A20) = 0000001C 20000000 1 [060018A18, 060018A20) = 00000001 00000000 6 [060018A18, 060018A20) = 00000002 00000000 5 [060018A18, 060018A20) = 00000001 00000000 18 [060018A18, 060018A20) = 00000000 00000000 --//可以最后请求S mode 没有阻塞,需要18秒完成.如果修改如下: $ cat i1.sh #! /bin/bash zdate=$(date '+%H%M%S') echo $zdate source peek.sh 'gcs partitioned table hash' 36 | timestamp.pl >| /tmp/peeks_${zdate}.txt & seq 36 | xargs -I{} echo -e 'sqlplus -s -l / as sysdba <<< @latch_free\nsleep 1' | bash >| /tmp/latch_free_${zdate}.txt & # 参数如下: @ latch.txt latch_name willing why where mode sleep_num sqlplus /nolog @ shared_latch.txt 'gcs partitioned table hash' 1 4 5 x 6 > /dev/null & sleep 2 sqlplus /nolog @ shared_latch.txt 'gcs partitioned table hash' 1 7 8 s 6 > /dev/null & sleep 2 sqlplus /nolog @ shared_latch.txt 'gcs partitioned table hash' 1 9 10 s 6 > /dev/null & sleep 1.9 sqlplus /nolog @ shared_latch.txt 'gcs partitioned table hash' 1 11 12 s 6 > /dev/null & wait $ grep -v '^.*: $' /tmp/peeks_103201.txt | cut -c10- | uniq -c 1 SYSDATE LADDR 1 ------------------- ---------------- 1 2019-04-16 10:32:01 0000000060018A18 1 Statement processed. 6 [060018A18, 060018A20) = 0000001C 20000000 18 [060018A18, 060018A20) = 00000001 00000000 12 [060018A18, 060018A20) = 00000000 00000000 --//可以发现我仅仅修改sleep 1.9秒,就导致后面3个S mode串行化.需要24秒完成.一旦串行化就很慢. --//有点想作者说的那样shared latch like enquence.