[20190423]oradebug peek测试脚本.txt

--//工作测试需要写一个oradebug peek测试脚本,不断看某个区域内存地址的值。

1.环境:

SCOTT@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

2.建立测试例子：
create table t as select rownum id,'test' name from dual ;
create unique index pk_t on t(id);
alter table t modify id  not null ;
--//分析表和索引略.

SCOTT@book> select rowid,t.* from t;
ROWID                      ID NAME
------------------ ---------- ----------------------------------------
AAAknXAAEAAAAILAAA          1 test

SCOTT@book> @ rowid AAAknXAAEAAAAILAAA
    OBJECT       FILE      BLOCK        ROW ROWID_DBA            DBA                  TEXT
---------- ---------- ---------- ---------- -------------------- -------------------- ----------------------------------------
    149975          4        523          0  0x100020B           4,523                alter system dump datafile 4 block 523 ;

SCOTT@book> select header_file,header_block from dba_segments where owner=user and segment_name='PK_T';
HEADER_FILE HEADER_BLOCK
----------- ------------
          4          554
--//从以上信息可以知道dba=4,523数据块,dba=4,522表T段头.dba=4,555(554+1)是索引的root块(因为索引很小也是叶子和分支块)

SYS@book> @ bh 4 522
HLADDR              DBARFIL     DBABLK      CLASS CLASS_TYPE         STATE             TCH CR_SCN_BAS CR_SCN_WRP CR_UBA_FIL CR_UBA_BLK CR_UBA_SEQ BA               OBJECT_NAME
---------------- ---------- ---------- ---------- ------------------ ---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------------- --------------------
0000000084DA2730          4        522          4 segment header     xcur                1          0          0          0          0          0 0000000071F46000 T

SYS@book> @ bh 4 523
HLADDR              DBARFIL     DBABLK      CLASS CLASS_TYPE         STATE             TCH CR_SCN_BAS CR_SCN_WRP CR_UBA_FIL CR_UBA_BLK CR_UBA_SEQ BA               OBJECT_NAME
---------------- ---------- ---------- ---------- ------------------ ---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------------- --------------------
0000000084C92150          4        523          1 data block         xcur                2          0          0          0          0          0 0000000072208000 T

SYS@book> @ bh 4 555
HLADDR              DBARFIL     DBABLK      CLASS CLASS_TYPE         STATE             TCH CR_SCN_BAS CR_SCN_WRP CR_UBA_FIL CR_UBA_BLK CR_UBA_SEQ BA               OBJECT_NAME
---------------- ---------- ---------- ---------- ------------------ ---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------------- --------------------
000000008577A438          4        555          1 data block         xcur                1          0          0          0          0          0 00000000721E2000 PK_T
--//获得这些块的cbc latch地址。
--//0000000084da2730,0000000084c92150,000000008577a438

3.编写脚本:
$ cat peek_laddr.sh
#! /bin/bash
# argument : laddr_list(delimiter using ,)  Monitor_count  peek_length sleep_duration
vdate=$(date '+%Y%m%d%H%M%S')
echo $vdate
laddr_list=$(echo $1| tr ',' '\n')
# p=$(echo "$laddr_list"|wc -l)
# echo $p
>| /tmp/pp_${vdate}.txt
for a in $laddr_list
do
        sqlplus -s -l / as sysdba <<EOF | timestamp.pl  >> /tmp/pp_${vdate}_${a}.txt &
oradebug setmypid
$(seq $2 | xargs -I{} echo  -e "oradebug peek 0x$a  $3\nhost sleep $4" )
quit
EOF
done

--//写的很丑陋,不过能用^_^.

$ cat z1.txt
set verify off
host sleep $(echo &&3/50 | bc -l )
variable vmethod varchar2(20);
exec :vmethod := '&&2';
insert into job_times values ( sys_context ('userenv', 'sid') ,dbms_utility.get_time ,:vmethod) ;
commit ;
declare
v_id number;
v_d date;
v_name varchar2(4) ;
begin
    for i in 1 .. &&1 loop
        --//select /*+ index(t) &&3 */ count (*) into v_id from t ;
        select /*+ full(t) &&3 */ count (*) into v_id from t ;
    end loop;
end ;
/
update job_times set time_ela = dbms_utility.get_time - time_ela where sid=sys_context ('userenv', 'sid') and method=:vmethod;
commit;
quit

4.测试脚本
-//编写脚本如下：
$ cat aaa.sh
#! /bin/bash
#vdate=$(date '+%Y%m%d%H%M%S')
#echo $vdate
laddr_list="$1"
echo $laddr_list
source peek_laddr.sh ${laddr_list} 200 8 0.1 &
seq 50 | xargs -I{} -P 50 sqlplus -s -l scott/book @z1.txt 1e5 id=index1 {} >/dev/null &

--//说明测试最好避开awr报表生成时间点.没有其它事务执行sql语句.
$ . aaa.sh 000000084da2730,0000000084c92150,000000008577a438
000000084da2730,0000000084c92150,000000008577a438
20190430104022

[1]-  Done                    source peek_laddr.sh ${laddr_list} 100 8 0.1
[2]+  Done                    seq 50 | xargs -I{} -P 50 sqlplus -s -l scott/book @z1.txt 1e5 id=full50 {} > /dev/null

--//等10秒看看.
SCOTT@book> Select method,count(*),round(avg(TIME_ELA),0),sum(TIME_ELA) from job_times group by method order by 3 ;
METHOD                 COUNT(*) ROUND(AVG(TIME_ELA),0) SUM(TIME_ELA)
-------------------- ---------- ---------------------- -------------
id=full50                    50                    977         48864

$ echo 000000084da2730,0000000084c92150,000000008577a438 | tr ',' '\n' | xargs -I{} grep  -v '^.*: $' /tmp/pp_20190430104022_{}.txt |cut -c10- |sort| uniq -c
     71  [084C92150, 084C92158) = 00000000 00000000
     21  [084C92150, 084C92158) = 00000001 00000000
      7  [084C92150, 084C92158) = 00000002 00000000
      1  [084C92150, 084C92158) = 00000004 00000000

60  [084DA2730, 084DA2738) = 00000000 00000000
     27  [084DA2730, 084DA2738) = 00000001 00000000
     12  [084DA2730, 084DA2738) = 00000002 00000000
      1  [084DA2730, 084DA2738) = 00000003 00000000

100  [08577A438, 08577A440) = 00000000 00000000
      3  Statement processed.
--//没有访问索引root节点块.
--//你可以发现大部分都是00000000 00000000.
--//可以发现没有任何阻塞,shared latch,我在链接的测试总结如下:http://blog.itpub.net/267265/viewspace-2641414/
--//总结:
--//A. S mode 下: peek记录的前4位持有S mode的数量.后4位是0x0. (这里针对的64位的系统)
--//B. S mode 下,如果出现X mode,peek记录的前4位持有S mode的数量.后4位是0x40000000.
--//一旦X mode持有变成 前4位持有会话PID号,后4位0x20000000.
--//你可以发现全表扫描的情况下这些块都是S模式获取.

--//我加大数据量(2e5)以及监测时间(200次)
$ cat aaa.sh
#! /bin/bash
#vdate=$(date '+%Y%m%d%H%M%S')
#echo $vdate
laddr_list="$1"
echo $laddr_list
source peek_laddr.sh ${laddr_list} 200 8 0.1 &
seq 50 | xargs -I{} -P 50 sqlplus -s -l scott/book @z1.txt 2e5 id=full50 {} >/dev/null &

$ echo 000000084da2730,0000000084c92150,000000008577a438 | tr ',' '\n' | xargs -I{} grep  -v '^.*: $' /tmp/pp_20190430104409_{}.txt |cut -c10- |sort| uniq -c
    150  [084C92150, 084C92158) = 00000000 00000000
     23  [084C92150, 084C92158) = 00000001 00000000
     22  [084C92150, 084C92158) = 00000002 00000000
      5  [084C92150, 084C92158) = 00000003 00000000
    124  [084DA2730, 084DA2738) = 00000000 00000000
     35  [084DA2730, 084DA2738) = 00000001 00000000
     25  [084DA2730, 084DA2738) = 00000002 00000000
     11  [084DA2730, 084DA2738) = 00000003 00000000
      4  [084DA2730, 084DA2738) = 00000004 00000000
      1  [084DA2730, 084DA2738) = 00000005 00000000
    200  [08577A438, 08577A440) = 00000000 00000000
      3  Statement processed.
--//后4位全部是00000000.

5.如果改用索引呢?
--//修改z1.txt如下:
$ cat z1.txt
set verify off
host sleep $(echo &&3/50 | bc -l )
variable vmethod varchar2(20);
exec :vmethod := '&&2';
insert into job_times values ( sys_context ('userenv', 'sid') ,dbms_utility.get_time ,:vmethod) ;
commit ;
declare
v_id number;
v_d date;
v_name varchar2(4) ;
begin
    for i in 1 .. &&1 loop
        select /*+ index(t) &&3 */ count (*) into v_id from t ;
        --//select /*+ full(t) &&3 */ count (*) into v_id from t ;
    end loop;
end ;
/
update job_times set time_ela = dbms_utility.get_time - time_ela where sid=sys_context ('userenv', 'sid') and method=:vmethod;
commit;
quit

$ cat aaa.sh
#! /bin/bash
#vdate=$(date '+%Y%m%d%H%M%S')
#echo $vdate
laddr_list="$1"
echo $laddr_list
source peek_laddr.sh ${laddr_list} 240 8 0.1 &
seq 50 | xargs -I{} -P 50 sqlplus -s -l scott/book @z1.txt 1e5 id=index50 {} >/dev/null &
--//前面我的测试需要24秒之内完成.

$ . aaa.sh 000000084da2730,0000000084c92150,000000008577a438
000000084da2730,0000000084c92150,000000008577a438
$ 20190430104822
[1]-  Done                    source peek_laddr.sh ${laddr_list} 240 8 0.1
[2]+  Done                    seq 50 | xargs -I{} -P 50 sqlplus -s -l scott/book @z1.txt 1e5 id=index50 {} > /dev/null

SCOTT@book> Select method,count(*),round(avg(TIME_ELA),0),sum(TIME_ELA) from job_times group by method order by 3 ;
METHOD                 COUNT(*) ROUND(AVG(TIME_ELA),0) SUM(TIME_ELA)
-------------------- ---------- ---------------------- -------------
id=index50                   50                   2268        113411

$ echo 000000084da2730,0000000084c92150,000000008577a438 | tr ',' '\n' | xargs -I{} grep  -v '^.*: $' /tmp/pp_20190430104822_{}.txt |cut -c10- |sort| uniq -c
    240  [084C92150, 084C92158) = 00000000 00000000
    240  [084DA2730, 084DA2738) = 00000000 00000000
--//注:没有访问数据块.dba=4,522 4,523.
     29  [08577A438, 08577A440) = 00000000 00000000
     18  [08577A438, 08577A440) = 00000000 20000000
     19  [08577A438, 08577A440) = 00000001 00000000
     12  [08577A438, 08577A440) = 00000001 20000000
      3  [08577A438, 08577A440) = 00000001 40000000
     17  [08577A438, 08577A440) = 00000002 00000000
      2  [08577A438, 08577A440) = 00000002 20000000
      1  [08577A438, 08577A440) = 00000002 40000000
     11  [08577A438, 08577A440) = 00000003 00000000
      1  [08577A438, 08577A440) = 00000003 40000000
      7  [08577A438, 08577A440) = 00000004 00000000
      1  [08577A438, 08577A440) = 00000005 00000000
      1  [08577A438, 08577A440) = 00000006 00000000
      1  [08577A438, 08577A440) = 00000007 00000000
      2  [08577A438, 08577A440) = 0000001B 20000000
      3  [08577A438, 08577A440) = 0000001D 20000000
      2  [08577A438, 08577A440) = 0000001E 20000000
      1  [08577A438, 08577A440) = 0000001F 00000000
      3  [08577A438, 08577A440) = 0000001F 20000000
      2  [08577A438, 08577A440) = 00000020 00000000
      3  [08577A438, 08577A440) = 00000020 20000000
      2  [08577A438, 08577A440) = 00000021 00000000
      1  [08577A438, 08577A440) = 00000021 20000000
      3  [08577A438, 08577A440) = 00000022 20000000
      1  [08577A438, 08577A440) = 00000023 20000000
      1  [08577A438, 08577A440) = 00000024 00000000
      1  [08577A438, 08577A440) = 00000024 20000000
      1  [08577A438, 08577A440) = 00000025 20000000
      1  [08577A438, 08577A440) = 00000026 20000000
      1  [08577A438, 08577A440) = 00000027 00000000
      3  [08577A438, 08577A440) = 00000027 20000000
      1  [08577A438, 08577A440) = 00000028 00000000
      2  [08577A438, 08577A440) = 00000028 20000000
      2  [08577A438, 08577A440) = 00000029 20000000
      2  [08577A438, 08577A440) = 0000002A 00000000
      2  [08577A438, 08577A440) = 0000002A 20000000
      2  [08577A438, 08577A440) = 0000002B 20000000
      3  [08577A438, 08577A440) = 0000002C 20000000
      1  [08577A438, 08577A440) = 0000002E 20000000
      3  [08577A438, 08577A440) = 0000002F 00000000
      2  [08577A438, 08577A440) = 0000002F 20000000
      2  [08577A438, 08577A440) = 00000030 20000000
      2  [08577A438, 08577A440) = 00000031 20000000
      1  [08577A438, 08577A440) = 00000032 20000000
      2  [08577A438, 08577A440) = 00000033 00000000
      1  [08577A438, 08577A440) = 00000034 00000000
      6  [08577A438, 08577A440) = 00000034 20000000
      3  [08577A438, 08577A440) = 00000035 20000000
      1  [08577A438, 08577A440) = 00000036 00000000
      1  [08577A438, 08577A440) = 00000037 00000000
      1  [08577A438, 08577A440) = 00000038 00000000
      3  [08577A438, 08577A440) = 00000038 20000000
      1  [08577A438, 08577A440) = 00000039 20000000
      3  [08577A438, 08577A440) = 0000003A 20000000
      2  [08577A438, 08577A440) = 0000003B 20000000
      1  [08577A438, 08577A440) = 0000003C 00000000
      2  [08577A438, 08577A440) = 0000003C 20000000
      1  [08577A438, 08577A440) = 0000003D 00000000
      1  [08577A438, 08577A440) = 0000003D 20000000
      2  [08577A438, 08577A440) = 0000003E 20000000
      2  [08577A438, 08577A440) = 0000003F 00000000
      1  [08577A438, 08577A440) = 0000003F 20000000
      1  [08577A438, 08577A440) = 00000040 20000000
      2  [08577A438, 08577A440) = 00000041 20000000
      2  [08577A438, 08577A440) = 00000042 20000000
      1  [08577A438, 08577A440) = 00000044 20000000
      2  [08577A438, 08577A440) = 00000045 20000000
      1  [08577A438, 08577A440) = 00000046 00000000
      3  [08577A438, 08577A440) = 00000046 20000000
      3  [08577A438, 08577A440) = 00000047 20000000
      2  [08577A438, 08577A440) = 00000048 20000000
      1  [08577A438, 08577A440) = 00000049 20000000
      1  [08577A438, 08577A440) = 0000004A 20000000
      1  [08577A438, 08577A440) = 0000004B 00000000
      2  [08577A438, 08577A440) = 0000004B 20000000
      5  [08577A438, 08577A440) = 0000004C 20000000
      3  Statement processed.

$ echo 000000084da2730,0000000084c92150,000000008577a438 | tr ',' '\n' | xargs -I{} grep  -v '^.*: $' /tmp/pp_20190430104822_{}.txt |cut -c10- |sort| uniq -c | grep 08577A438| sort -nr | head
     29  [08577A438, 08577A440) = 00000000 00000000
     19  [08577A438, 08577A440) = 00000001 00000000
     18  [08577A438, 08577A440) = 00000000 20000000
     17  [08577A438, 08577A440) = 00000002 00000000
     12  [08577A438, 08577A440) = 00000001 20000000
     11  [08577A438, 08577A440) = 00000003 00000000
      7  [08577A438, 08577A440) = 00000004 00000000
      6  [08577A438, 08577A440) = 00000034 20000000
      5  [08577A438, 08577A440) = 0000004C 20000000
      3  [08577A438, 08577A440) = 00000047 20000000

$ grep " 40000000$" /tmp/pp_20190430104822_000000008577a438.txt
10:48:23: [08577A438, 08577A440) = 00000001 40000000
10:48:35: [08577A438, 08577A440) = 00000001 40000000
10:48:37: [08577A438, 08577A440) = 00000003 40000000
10:48:44: [08577A438, 08577A440) = 00000001 40000000
10:48:45: [08577A438, 08577A440) = 00000002 40000000

--//我估计全部扫描全部使用是S mode获取cbc latch.而 INDEX FULL SCAN我估计有部分操作采用X mode获取cbc latch.
--//这也许就是为什么11g全表扫描快于INDEX FULL SCAN的原因吗?
--//实际上这个给oracle一些优化提供一些信息,比如一个表dept,经常查询deptno,DNAME两个字段,如果通过建立符合索引减少逻辑读.
--//如果程序出现大量频繁访问,反而全表扫描会更快.因为可能遇到cbc latch更少.
--//优化一定要考虑这些细节.有许多还是不是很清楚,先放一下...实际上仅仅11.2.0.4才会出现这样的情况.全表扫描快于INDEX FULL SCAN.

[20190423]oradebug peek测试脚本.txt的更多相关文章

  1. [20190416]exclusive latch测试脚本.txt

    [20190416]exclusive latch测试脚本.txt --//昨天做了shared latch的测试脚本,今天完善exclusive latch测试脚本,上个星期的测试我是手工执行的.- ...

  2. [20190416]完善shared latch测试脚本2.txt

    [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-264 ...

  3. [20180619]oradebug peek.txt

    [20180619]oradebug peek.txt --//我以前一直以为oradebug peek查看某个地址开始的内容,后面的长度有限制的.--//在linux下,我的测试是60.实际上ora ...

  4. [20180625]oradebug peek 2.txt

    [20180625]oradebug peek 2.txt --//上个星期演示了oradebug peek查看内存数据块的情况,oradebug peek {address} length 1,最后 ...

  5. 锤子便签的 monkeyrunner 测试脚本（转）

    https://testerhome.com/topics/878 MonkeyRunner可能大家已经听过无数次了,大家在网上也看过了各种关于的它的资料了,我这里就不再过多的啰嗦它的用途了,它可以对 ...

  6. 【NO.11】Jmeter - 构建1个可供Linux使用的Jmeter测试脚本 - 共3个步骤

    在Linux使用Jmeter做性能测试需要4个前提条件,这4个前提条件已经在之前的文档里提到了,重复一下加深印象: (1) 在本地已安装xshell 参考<SecureCRT-转换密钥-Xshe ...

  7. [20180814]慎用查看表压缩率脚本.txt

    [20180814]慎用查看表压缩率脚本.txt --//最近看exadata方面书籍,书中提供1个脚本,查看某些表采用那些压缩模式压缩比能达到多少.--//通过调用DBMS_COMPRESSION. ...

  8. 记录项目代码迁移后，UI测试框架的搭建（配置文件的修改、测试脚本试运行）

    前文:记录一次项目代码迁移过程 上文代码迁移的目的就是为了新增vue脚手架自带的UI测试框架,工具有了,就需要实践运行在项目中了(修改配置文件.编写测试脚本等). 一.单元测试 测试框架 karma ...

  9. [20191013]oracle number类型存储转化脚本.txt

    [20191013]oracle number类型存储转化脚本.txt --//测试看看是否可以利用bc obase=100的输出解决问题.另外以前脚本忘记考虑尾数的四舍五入问题.--//也许编程就是 ...

随机推荐

  1. [译]作为一个web开发人员，哪些技术细节是在发布站点前你需要考虑到的

    前日在cnblogs上看到一遍文章<每个程序员都必读的12篇文章>,其中大多数是E文的. 先译其中一篇web相关的”每个程序员必知之WEB开发”. 原文: http://programme ...

  2. 温故而知新-面向对象的PHP

    1 类的多态 不同的类对同一操作可以有不同的行为. 比如自行车和汽车都有移动这个成员函数行为, 那么自行车类可以移动,行为和汽车的移动行为肯定不同. 2 析构函数不能有参数 3 __set和__get ...

  3. 关闭 iTunes 自动同步

    关闭 iTunes 自动同步 iTunes>Edit>perererset>device>Prevent iPods,iPhone, and iPads from syncin ...

  4. 一劳永逸解决VLC播放中文字幕乱码问题

    VLC对于Mac/Ubuntu用户来说算得上是必备软件.其相当于PC机上的“暴风影音”,但Mac/Ubuntu的新手使用VLC播放avi时都会碰 到字幕乱码的问题.avi字幕的格式有多种,这里假设你使 ...

  5. 2 数据库开发--MySQL下载（windows）

    Windows:(mysql) 操作: 0.下载安装mysql www.mysql.org downloads->进入社区community community 5.7.21 下载5.6 Mic ...

  6. 为什么需要maven，maven能解决什么问题 ，maven是什么

  7. Matlab中插值函数汇总(下)

    Matlab中插值函数汇总分上下两个部分,主要整合自matlabsky论坛dynamic发表于2009-2-21 21:53:26 的主题帖,以及豆丁网rickoon上传的教材第8章<插值,拟合 ...

  8. C++ 基类指针，子类指针，多态

    基类指针和子类指针之间相互赋值(1)将子类指针赋值给基类指针时,不需要进行强制类型转换,C++编译器将自动进行类型转换.因为子类对象也是一个基类对象. (2)将基类指针赋值给子类指针时,需要进行强制类 ...

  9. freetype 编译

    https://blog.csdn.net/yapingxin/article/details/51841039

  10. Maven的pom.xml介绍

    6.1 简介 pom.xml文件是Maven进行工作的主要配置文件.在这个文件中我们可以配置Maven项目的groupId.artifactId和version等Maven项目必须的元素:可以配置Ma ...