Database Replay将生产系统上的负荷进行采集后还原到测试系统上进行重放，这个特性有助于我们在数据库升级、主机参数修改、数据库参数修改等重大变更实施前在测试系统上完全仿照生产系统的负荷进行全面的测试，量化评估出变更实施后对现有的性能的影响程度。
Database Replay实施的流程是workload_capture->workload preprocess->replay client prepare->replay->generate replay report，对于同一个capture可以进行多次replay
实施过程中需要的服务器如下：
生产数据库服务器：用于workload_capture，即负载捕获
测试数据库服务器：用于workload_ preprocess和replay，即负载预处理和负载重放
Replay client：用于发起workload的客户端进程
如果纯粹是用于测试那么生产数据库、测试数据库、Replay client可以指向同一台机器。为了较为清晰的展示Database replay的整个过程，接下来的演示中我们会用到三台主机：
Prod db Server：生产数据库服务器
Replay Client：用于发起workload的客户端进程wrc
Replay db Server：测试数据库服务器
//1、 workload_capture阶段，这些操作均在prod Server上执行
###创建capture dir，确保目录下无任何文件
mkdir -p /oradata06/repdir
create or replace directory repdir as '/oradata06/repdir';
###创建过滤器，过滤器中的内容是被包含还是排除，取决于capture是include还是exclude属性，如果是include，那么下面过滤器的内容将会被排除掉，此演示中将会采用include的capture所以我们过滤器中指定的是我们不需要的内容：排除掉OMS和emagent这两个program；如果是exclusion属性，那么过滤器中的才是我们需要捕捉的，除此之外都会被排除；
exec dbms_workload_capture.add_filter(fname=>'filter_prog1',fattribute=>'PROGRAM',fvalue=>'%OMS%');
exec dbms_workload_capture.add_filter(fname=>'filter_prog2',fattribute=>'PROGRAM',fvalue=>'%emagent%');
col set_name format a20
col type format a10
col id format a20
col name format a15
col attribute format a15
col value format a18
set linesize 130
select * from dba_workload_filters;
TYPE ID STATUS SET_NAME NAME ATTRIBUTE VALUE
---------- -------------------- ------ -------------------- --------------- --------------- ------------------
CAPTURE NEW FILTER_PROG2 PROGRAM %emagent%
CAPTURE NEW FILTER_PROG1 PROGRAM %OMS%
###开启workload capture之前，重启数据库到restricted模式->此步骤可选，为的是能一个不落的capture到所有session
shutdown immediate
startup restrict
###创建两张测试表
create table tabfix1 (id1 number,id1_mod number);
declare
begin
for i in 1..100000 loop
insert into tabfix1 values (i,i+dbms_random.value(-2,2));
end loop;
commit;
end;
/
create table tabvar1 (vid1 number,vid1_mod number);
declare
begin
for i in 1..100000 loop
insert into tabvar1 values (i,i+dbms_random.value(-2,2));
end loop;
commit;
end;
/
###准备模拟负载运行脚本s_cap1.sh、s_cap2.sh
---s_cap1.sh内容
while [ true ]
do
sqlplus system/xxxxxx@tstdb1 << EOF
update tabfix1 set id1_mod=id1_mod+dbms_random.value(-2,2) where id1<=50000;
commit;
select count(*) from tabfix1 where id1<=50000;
EOF
sleep 1
done
---s_cap2.sh内容
while [ true ]
do
sqlplus system/xxxxxx@tstdb1 << EOF
update tabvar1 set vid1_mod=vid1_mod+dbms_random.value(-2,2) where vid1_mod<=50000;
commit;
select count(*) from tabfix1 where id1<=50000;
EOF
sleep 1
done
###准备Replay db Server用于replay，将Replay db恢复到prod db Server开启capture前的状态，oracle推荐使用duplicate database、expdp & impdp、snapshot standby
三种方式，本演示中仅将上面两张测试表导入到replay db server
expdp system/xxxxxx tables=tabfix1,tabvar1 directory=hisdmp logfile=tabf.log dumpfile=tabf.dmp
scp /oradata01/hisdmp/monthly/tabf.dmp oracle@192.168.0.207:/oradata01/hisdmp/monthly/
impdp system/xxxxxx directory=hisdmp logfile=tabf.log dumpfile=tabf.dmp
###正式开启workload capture，capture包括了filter_prog1、filter_prog2 两个filter以外的所有内容，capture动作开始时自动解除restrict模式，以10分钟为间隔同时捕捉sql tuning sets，STS可以在Replay报告比较阶段生产更细化的性能数据
exec dbms_workload_capture.start_capture(name=>'v1026c1',dir=>'REPDIR',duration=>600,default_action=>'INCLUDE',auto_unrestrict=>TRUE,capture_sts=>TRUE,sts_cap_interval=>300);
###运行模拟负载脚本
./s_cap1.sh
./s_cap2.sh
###capture期间可以查看dba_workload_captures了解capture状态
col name format a10
col status format a20
col sqlset_name format a25
set linesize 120
select id,name,status,duration_secs,awr_begin_snap,awr_end_snap,sqlset_name from dba_workload_captures;
ID NAME STATUS DURATION_SECS AWR_BEGIN_SNAP AWR_END_SNAP SQLSET_NAME
---------- ---------- -------------------- ------------- -------------- ------------ -------------------------
91 v1026c1 COMPLETED 597 131 132 v1026c1_c_9196364
###dba_sqlsets标示着此sqlset来自于id=91的capture
col description format a40
col owner format a10
col name format a20
set linesize 140
select * from dba_sqlset where name='v1026c1_c_9196364';
ID NAME OWNER DESCRIPTION CREATED LAST_MODIFIE STATEMENT_COUNT
---------- -------------------- ---------- ---------------------------------------- ------------ ------------ ---------------
8 v1026c1_c_9196364 SYSTEM STS capture for capture with ID=91 27-OCT-14 27-OCT-14 378
###停止workload capture->此步骤可选，如果start capture时没有指定持续时间可以人工停止
exec dbms_workload_capture.finish_capture(timeout=>0);
###导出capture期间的AWR数据->此步骤可选，如果之后要生成compare period report，即两个replay的结果的比较报告或者replay与capture之间的比较报告，就要用到export_awr
select id from dba_workload_captures; --得到Capture_id
exec dbms_workload_capture.export_awr(capture_id=>91)
###生成capture阶段报告到v1026c1.html文件
set serveroutput on
spool /home/tstdb1/v1026c1.html
set long 200000
set pagesize 20000
DECLARE
v_cap_id NUMBER;
v_cap_rpt CLOB;
BEGIN
v_cap_id:=DBMS_WORKLOAD_CAPTURE.GET_CAPTURE_INFO(dir => 'REPDIR');
v_cap_rpt:=DBMS_WORKLOAD_CAPTURE.REPORT(capture_id =>v_cap_id,format => DBMS_WORKLOAD_CAPTURE.TYPE_HTML);
dbms_output.put_line(v_cap_rpt);
END;
/
spool off
###capture报告内容截取如下：
主要包含了Capture Profile、Captured Workload Statistics、Top Events Captured、Top SQL Captured等内容
注：在/oradata06/repdir/cap目录下其实已经存在wcr_cr.html、wcr_cr.text两个文件分别对应了html和txt版本的capture报告，拿来用即可
///
// 2、preprocess 阶段
///
###先将capture file从prod db server复制到replay db server
scp -r /oradata06/repdir/* oracle@192.168.0.207:/oradata01/hisdmp/monthly/
***copy完成后replay db server主机上的目录结构如下
ls -rlt /oradata01/hisdmp/monthly/
drwxr-xr-x 3 oracle oinstall 96 Oct 29 14:59 capfiles
drwxr-xr-x 2 oracle oinstall 8192 Oct 29 14:59 cap
-rw-r--r-- 1 oracle oinstall 0 Oct 29 16:00 wcr_cap_0002x.start
-rw-r--r-- 1 oracle oinstall 0 Oct 29 16:00 wcr_cap_0002w.start
###在replay db server上进行preprocess
exec dbms_workload_replay.process_capture(capture_dir=>'HISDMP');
###preprocess完成后在replay db server的/oradata01/hisdmp/monthly/目录下生成了pp11.2.0.3.0子目录
oracle@jq570322a:/oradata01/hisdmp/monthly>ls -rlt pp11.2.0.3.0
total 432
-rw-r----- 1 oracle oinstall 28672 Oct 29 16:03 wcr_seq_data.extb
-rw-r----- 1 oracle oinstall 28672 Oct 29 16:03 wcr_scn_order.extb
-rw-r--r-- 1 oracle oinstall 25787 Oct 29 16:03 wcr_login.pp
-rw-r----- 1 oracle oinstall 12288 Oct 29 16:03 wcr_conn_data.extb
-rw-r----- 1 oracle oinstall 12288 Oct 29 16:03 wcr_data.extb
-rw-r----- 1 oracle oinstall 16384 Oct 29 16:03 wcr_references.extb
-rw-r--r-- 1 oracle oinstall 35 Oct 29 16:03 wcr_process.wmd
-rw-r----- 1 oracle oinstall 36864 Oct 29 16:03 wcr_dep_graph.extb
-rw-r----- 1 oracle oinstall 12288 Oct 29 16:03 wcr_commits.extb
-rw-r--r-- 1 oracle oinstall 3510 Oct 29 16:03 wcr_calibrate.xml
这些文件主要是对capfiles目录下.rec文件进行汇总，在之后的replay阶段起到索引的作用。其中wcr_calibrate.xml是对replay时发起replay client数量的预估结果，内容如下：
v1026c1
2028908302
TSTDB1
11.2.0.3.0
NO
HISDMP
/oradata01/hisdmp/monthly
TRUE
COMPLETED
27-10-14 20:03:35
27-10-14 20:13:32
9 minutes 57 seconds
3022402
3088331
INCLUDE
2
1096881
534902718
.9
605738506
7939
9345
1222
286
469
312
370
2
NOT POSSIBLE
0
11.2.0.3.0
11.2.0.3.0
4
50
4
313
1
1
15
///
//3、 replay client准备阶段
///
###准备replay clients环境，replay clients主要作用是模拟客户端连接到test database发起压力测试
首先，运行replay clients的主机至少要安装oracle client；
replay client主机上的sqlnet.ora必须包含有DIAG_ADR_ENABLED=ON，否则会收到ORA-15555 "workload replay client encountered unexpected error: %s错误
###在replay db server上执行：将capture及preprocess的结果copy到运行replay clients，需要指出的是这里不仅要将preprocess的结果copy过去，还要将cap，capfiles两个目录一起copy到replay clients主机
scp -r /oradata01/hisdmp/monthly/pp11.2.0.3.0 oracle@192.168.0.221:/oradata01/hisdmp/monthly
scp -r /oradata01/hisdmp/monthly/cap oracle@192.168.0.221:/oradata01/hisdmp/monthly
scp -r /oradata01/hisdmp/monthly/capfiles oracle@192.168.0.221:/oradata01/hisdmp/monthly
###解释一下为何cap、capfiles目录也要一起copy过去
原因很简单在preprocess生成一堆文件相当于给负载编制好的一套目录，其中有个wcr_login.pp文件记录了capture阶段生成的所有.rec文件，一个session登陆后就生成一个.rec文件，用strings命令可以清楚的看到:
strings /oradata01/hisdmp/monthly/pp11.2.0.3.0/wcr_login.pp | grep .rec | head -n 10
wcr_4wd4rh0000000.rec
wcr_4wd4sh0000003.rec
wcr_4wd4th0000004.rec
wcr_4wd4uh0000006.rec
wcr_4wd4uh0000007.rec
wcr_4wd4vh0000009.rec
以上每一个文件都能在/oradata01/hisdmp/monthly/capfiles/inst1/目录下找到，负载重放时会到此目录下读取对应的文件，如果仅将pp11.2.0.3.0目录copy过去在replay Client运行wrc的时候会出现：ORA-15559: workload replay client cannot open workload capture file
###每个客户端可以模拟出多个session，calibrate 模式下使用wrc可以根据捕捉到的workload信息估算出需要发起多少个Replay client，在replay client上执行：
wrc mode=calibrate replaydir=/oradata01/hisdmp/monthly
Workload Replay Client: Release 11.2.0.3.0 - Production on Wed Oct 29 15:07:23 2014
Copyright (c) 1982, 2011, Oracle and/or its affiliates. All rights reserved.
Report for Workload in: /oradata01/hisdmp/monthly
-----------------------
Recommendation:
Consider using at least 1 clients divided among 1 CPU(s)
You will need at least 15 MB of memory per client process.
If your machine(s) cannot match that number, consider using more clients.
Workload Characteristics:
- max concurrency: 4 sessions
- total number of sessions: 313
Assumptions:
- 1 client process per 50 concurrent sessions
- 4 client process per CPU
- 256 KB of memory cache per concurrent session
- think time scale = 100
- connect time scale = 100
- synchronization = TRUE
/
// 4、replay 阶段，在replay db server上执行
/
###initialize replay，生成新的replay信息，同时会将将replay_dir目录下的capture信息、之前已经存在的replay信息一并导入到数据库，可以通过dba_workload_replays、dba_workload_connection_map、dba_workload_filters等视图查看
exec dbms_workload_replay.initialize_replay(replay_name=>'v1028r1',replay_dir=>'HISDMP'); --replay_dir指定的是包含有preprocess data及capture data的目录
***replay状态为initialized
col name format a30
set linesize 120
select id,name,status from dba_workload_replays;
ID NAME STATUS
---------- ------------------------------ ----------------------------------------
24 v1028r1 INITIALIZED
###设置Replay时的filter->此为可选步骤
利用capture到的负载进行replay时如果要过滤掉部分sql，可以按照如下步骤，创建filter，过滤掉程序名包含emagent、perl的session
exec dbms_workload_replay.add_filter(fname=>'v1028f1',fattribute=>'PROGRAM',fvalue=>'%emagent%');
exec dbms_workload_replay.add_filter(fname=>'v1028f1_1',fattribute=>'PROGRAM',fvalue=>'%perl%');
***dba_workload_filter看到TYPE=REPLAY类型的filter，过滤掉program=perl或者emagent的session
col set_name format a20
col type format a10
col id format a20
col name format a15
col attribute format a15
col value format a18
set linesize 130
select * from dba_workload_filters;
TYPE ID STATUS SET_NAME NAME ATTRIBUTE VALUE
---------- -------------------- ------ -------------------- --------------- --------------- ------------------
CAPTURE 34 USED FILTER_PROG2 PROGRAM %emagent%
CAPTURE 34 USED FILTER_PROG1 PROGRAM %OMS%
REPLAY NEW V1028F1 CONNECTION_STRI %(PROGRAM=%emagent
NG %)%
REPLAY NEW V1028F1_1 CONNECTION_STRI %(PROGRAM=%perl%)%
NG
***将前面创建的filter加入到Filter_set，入参里没有指定filter_name，默认是把上一次create_filter_set执行后使用add_filter添加的所有filter加入到该filter_set内
exec dbms_workload_replay.create_filter_set(replay_dir=>'HISDMP',filter_set=>'v1028fs1',default_action=>'INCLUDE');
*** dba_workload_filters里type=REPLAY的filter状态从NEW变为了IN SET，表示已经加入到filter Set里了,ID列为空说明还未被任何replay使用
select * from dba_workload_filters;
TYPE ID STATUS SET_NAME NAME ATTRIBUTE VALUE
---------- -------------------- ---------- -------------------- --------------- --------------- ------------------
CAPTURE 34 USED FILTER_PROG2 PROGRAM %emagent%
CAPTURE 34 USED FILTER_PROG1 PROGRAM %OMS%
REPLAY IN SET v1028fs1 V1028F1_1 CONNECTION_STRI %(PROGRAM=%perl%)%
NG
REPLAY IN SET v1028fs1 V1028F1 CONNECTION_STRI %(PROGRAM=%emagent
NG %)%
---使用名为v1028fs1的filter set过滤接下来的replay，type=REPLAY行的ID=24，status=IN USE说明这个filter_set正在被replay所使用
exec dbms_workload_replay.use_filter_set(filter_set=>'v1028fs1');
col set_name format a20
col type format a10
col id format a20
col name format a15
col attribute format a15
col value format a18
set linesize 130
select * from dba_workload_filters;
TYPE ID STATUS SET_NAME NAME ATTRIBUTE VALUE
---------- -------------------- ---------- -------------------- --------------- --------------- ------------------
CAPTURE 34 USED FILTER_PROG2 PROGRAM %emagent%
CAPTURE 34 USED FILTER_PROG1 PROGRAM %OMS%
REPLAY 24 IN USE v1028fs1 V1028F1 CONNECTION_STRI %(PROGRAM=%emagent
NG %)%
REPLAY 24 IN USE v1028fs1 V1028F1_1 CONNECTION_STRI %(PROGRAM=%perl%)%
NG
###replay之前检查capture里包含的连接串信息
***capture内容里的连接串还是指向production database，在replay之前需要重定向到test database
col capture_conn format a50
set linesize 120 pagesize 120
select * from dba_workload_connection_map;
REPLAY_ID CONN_ID CAPTURE_CONN
---------- ---------- --------------------------------------------------
REPLAY_CONN
------------------------------------------------------------------------------------------------------------------------
24 1 (DESCRIPTION=(ADDRESS=(PROTOCOL=TCP)(HOST=10.10.14
1.209)(PORT=1521))(CONNECT_DATA=(SID=tstdb1)(CID=(
PROGRAM=E:\Program?Files??x86?\PLSQL?Developer\pls
qldev.exe)(HOST=CHHHHC)(USER=CHH))))
24 2 (DESCRIPTION=(ADDRESS=(PROTOCOL=TCP)(HOST=10.10.14
1.209)(PORT=1521))(CONNECT_DATA=(SID=tstdb1)(SERVE
R=DEDICATED)(CID=(PROGRAM=sqlplus)(HOST=jq570322b)
(USER=tstdb1))))
24 3 (DESCRIPTION=(CONNECT_DATA=(SID=tstdb1)(CID=(PROGR
AM=perl@jq570322b)(HOST=jq570322b)(USER=oracle)))(
ADDRESS=(PROTOCOL=TCP)(HOST=192.168.0.209)(Port=15
21)))
###可选步骤：对于capture阶段捕捉到的连接串信息可以在Replay阶段将这些连接信息重新映射到其它数据库，例如下面的例子中将connect_id=1、2、3的连接串分别映射到test1、test2、test3对应的数据库
exec DBMS_WORKLOAD_REPLAY.REMAP_CONNECTION(connection_id=>1,replay_connection=>'test1');
exec DBMS_WORKLOAD_REPLAY.REMAP_CONNECTION(connection_id=>2,replay_connection=>'test2');
exec DBMS_WORKLOAD_REPLAY.REMAP_CONNECTION(connection_id=>3,replay_connection=>'test3');
***重新映射连接串后，replay_conn显示replay client时使用的连接串
col replay_conn format a15
select replay_id,conn_id,capture_conn,REPLAY_CONN from dba_workload_connection_map;
REPLAY_ID CONN_ID CAPTURE_CONN REPLAY_CONN
---------- ---------- -------------------------------------------------- ---------------
24 1 (DESCRIPTION=(ADDRESS=(PROTOCOL=TCP)(HOST=10.10.14 test1
1.209)(PORT=1521))(CONNECT_DATA=(SID=tstdb1)(CID=(
PROGRAM=E:\Program?Files??x86?\PLSQL?Developer\pls
qldev.exe)(HOST=CHHHHC)(USER=CHH))))
24 2 (DESCRIPTION=(ADDRESS=(PROTOCOL=TCP)(HOST=10.10.14 test2
1.209)(PORT=1521))(CONNECT_DATA=(SID=tstdb1)(SERVE
R=DEDICATED)(CID=(PROGRAM=sqlplus)(HOST=jq570322b)
(USER=tstdb1))))
24 3 (DESCRIPTION=(CONNECT_DATA=(SID=tstdb1)(CID=(PROGR test3
AM=perl@jq570322b)(HOST=jq570322b)(USER=oracle)))(
ADDRESS=(PROTOCOL=TCP)(HOST=192.168.0.209)(Port=15
21)))
注意test1、test2、test3必须在replay client端能够tnsping通这些别名，或者干脆使用"(DESCRIPTION=(ADDRESS=(PROTOCOL=TCP)(HOST=192.168.0.209)(PORT=1521))(CONNECT_DATA=(SERVICE_NAME=shzw)))"这种显示指定的方法来实现重映射
但在实际使用时发现connection_remap特性无法正常用起来，在Replay client侧发起客户端连接进程的时候报ORA-15561错误
oracle@jq570314a:/oradata01/hisdmp/monthly>wrc system/XXXXXX@test mode=replay replaydir=/oradata01/hisdmp/monthly
Workload Replay Client: Release 11.2.0.3.0 - Production on Wed Oct 29 09:36:47 2014
Copyright (c) 1982, 2011, Oracle and/or its affiliates. All rights reserved.
Errors in file :
ORA-15561: workload replay client cannot connect to the remapped connection with conn_id : 1
要规避这个错误按照MOS 1135960.1的做法是在wrc命令里加上CONNECTION_OVERRIDE=TRUE参数，使用wrc命令里自带的连接串指向test database，但这个做法的问题在于只能将多个capture连接串重定义到同一个replay连接串，无法为每个capture连接串指定各自不同replay连接串
###prepare replay，控制replay的参数设定,synchronization=>'OBJECT_ID'针对于DML操作，在维持同一个对象前后事务依赖性的前提下，必要时会略微调整DML语句的执行顺序(调整不会影响语句结果)，特别是在多个事务并发操作不同的对象时，相比synchronization=>'SCN'更能提高replay期间DML的执行效率，capture_sts=>TRUE会在replay时同时生成STS
exec dbms_workload_replay.prepare_replay(synchronization=>'OBJECT_ID',capture_sts=>TRUE,sts_cap_interval=>300);
***prepare后在replay_dir下生成了一个repXXX开头的目录，其中XXX是oracle为每个replay分配的唯一标识，可以在dba_workload_replays视图的replay_dir_number字段中查到
drwxr-xr-x 3 oracle oinstall 96 Oct 29 14:59 capfiles
-rw-r--r-- 1 oracle oinstall 0 Oct 29 16:00 wcr_cap_0002x.start
-rw-r--r-- 1 oracle oinstall 0 Oct 29 16:00 wcr_cap_0002w.start
drwxr-xr-x 2 oracle oinstall 8192 Oct 29 16:03 pp11.2.0.3.0
drwxr-xr-x 2 oracle oinstall 8192 Oct 30 09:48 cap
drwxr-xr-x 2 oracle oinstall 96 Oct 30 10:17 filter11.2.0.3.0
drwxr-xr-x 2 oracle oinstall 96 Oct 30 10:20 rep423075657
set linesize 140
select name,replay_dir_number from dba_workload_replays
NAME REPLAY_DIR_NUMBER
---------------------------------------------------------------------------------------------------- -----------------
v1028r1 423075657
***replay任务的状态变为了prepare
col name format a30
set linesize 120
select id,name,status from dba_workload_replays;
ID NAME STATUS
---------- ------------------------------ ----------------------------------------
24 v1028r1 PREPARE
###在replay client所在主机发起replay client，这时客户端提示等待start_replay发起，使用CONNECTION_OVERRIDE覆盖掉dba_workload_connection_map设置，规避掉ORA-15561错误
wrc system/xxxxxx@test mode=replay CONNECTION_OVERRIDE=TRUE REPLAYDIR=/oradata01/hisdmp/monthly
Wait for the replay to start (10:24:23)
###在replay db server执行：开启replay
exec DBMS_WORKLOAD_REPLAY.START_REPLAY;
###replay client主机立刻显示replay开始
Wait for the replay to start (10:24:23)
Replay started (10:24:43)
###replay期间主要通过v$workload_replay_thread视图观察进展情况，视图里显示了每个session当前正在处理哪一个capture file以及当前所耗费的dbtime等信息
col event format a40
col file_name format a30
set linesize 130
select sid,session_type,event,file_name,dbtime from v$workload_replay_thread where session_type='REPLAY';
SID SESSION_TYPE EVENT FILE_NAME DBTIME
---------- ------------- ---------------------------------------- ------------------------------ ----------
1912 REPLAY Disk file operations I/O wcr_4wdmch000008b.rec 74599
3303 REPLAY SQL*Net message from client wcr_4wdhyh000006k.rec 69859
3775 REPLAY SQL*Net message from client wcr_4wdhnh0000069.rec 87932
5658 REPLAY Disk file operations I/O wcr_4wdmch000008c.rec 73601
7080 REPLAY null event wcr_4wd59h000000t.rec 0
###至此一个完整的replay完成，为了辅助讲解后面的compare period report，我们又进行了一次replay，产生一个id=26的replay
select id,name,status,start_time,end_time from dba_workload_replays;
ID NAME STATUS START_TIME END_TIME
---------- ---------- ---------- ----------------- -----------------
24 v1028r1 COMPLETED 20141030 10:23:37 20141030 10:33:36
26 v1028r2 COMPLETED 20141030 12:16:00 20141030 12:26:00
//5、生成replay报告
Replay报告有几种，本文主要介绍replay report、compare period report两种。
replay reports将replay执行时的性能统计信息列出来，并与capture作简单的比较
compare period report侧重于在两个不同的replay之间，或者replay和capture之间从数据库参数配置、主机硬件环境、Top SQL耗用资源等方面形成全方位的诊断报告
###生成replay report
set serveroutput on
spool /home/oracle/v1028r1.html
set pagesize 20000
set long 200000
declare
v_offset number;
v_length number;
v_nowlength number;
v_char1 varchar2(32767);
v_reprpt clob;
begin
v_reprpt:=dbms_workload_replay.report(replay_id=>24,format=>dbms_workload_replay.TYPE_HTML);
v_nowlength:=1;
v_length:=dbms_lob.getlength(lob_loc=>v_reprpt);
v_offset:=1;
while ( v_offset < v_length ) loop
v_char1:=dbms_lob.substr(lob_loc=>v_reprpt,offset=>v_offset);
dbms_output.put_line(v_char1);
v_offset:=v_offset+32767;
end loop;
end;
/
spool off
###生成的replay报告主要包括replay information、options、statistics、divergence等信息，也会包含细化到SQL语句的负载分析
截图如下：
上图中Replay Divergence Summary用来反应replay和capture阶段执行SQL时遇到的错误次数，这些错误是否仅在replay还是capture阶段出现；DML或SELECT语句返回记录数是否不一致的情况等
上图对于SQL语句按照负荷从高到低进行排序，并对这些负荷的分布进行breakdown
###生成compare period report
select id,name,status,start_time,end_time from dba_workload_replays;
ID NAME STATUS START_TIME END_TIME
---------- ---------- ---------- ----------------- -----------------
24 v1028r1 COMPLETED 20141030 10:23:37 20141030 10:33:36
26 v1028r2 COMPLETED 20141030 12:16:00 20141030 12:26:00
###将之前生成的id=24、id=26两个replay形成比较报告之前需要从prod db server上将capture阶段的AWR负载export出来，import到replay db server中，以下在product database上实施export_awr操作
exec DBMS_WORKLOAD_CAPTURE.EXPORT_AWR(capture_id=>91);
###在prod db server上执行：将export_awr导出的文件传输到replay db server，以下标红的文件都要复制到replay db server
oracle@jq570322b:/oradata06/repdir/cap>ls -rlt
total 28336
-rw-r--r-- 1 oracle oinstall 162 Oct 27 20:03 wcr_scapture.wmd
-rw-r----- 1 oracle oinstall 268 Oct 27 20:14 wcr_fcapture.wmd
-rw-r----- 1 oracle oinstall 46708 Oct 27 20:14 wcr_cr.html
-rw-r----- 1 oracle oinstall 19196 Oct 27 20:14 wcr_cr.text
-rw-r----- 1 oracle oinstall 29141 Oct 30 09:36 wcr_ca.log
-rw-r----- 1 oracle oinstall 12222464 Oct 30 09:36 wcr_ca.dmp
-rw-r----- 1 oracle oinstall 12288 Oct 30 09:36 wcr_cap_uc_graph.extb
-rw-r----- 1 oracle oinstall 2146304 Oct 30 09:36 wcr_ca_sts.dmp
scp /oradata06/repdir/cap/wcr_ca.log oracle@192.168.0.207:/oradata01/hisdmp/monthly/cap/
scp /oradata06/repdir/cap/wcr_ca.dmp oracle@192.168.0.207:/oradata01/hisdmp/monthly/cap/
scp /oradata06/repdir/cap/wcr_cap_uc_graph.extb oracle@192.168.0.207:/oradata01/hisdmp/monthly/cap/
scp /oradata06/repdir/cap/wcr_ca_sts.dmp oracle@192.168.0.207:/oradata01/hisdmp/monthly/cap/
***replay db server上操作：将上述export_awr的结果通过函数DBMS_WORKLOAD_CAPTURE.IMPORT_AWR导入到replay db server，DBMS_WORKLOAD_CAPTURE.IMPORT_AWR的返回值是随机生成的DBID，可以在dba_workload_captures.awr_dbid里找到
var v_randbid number;
exec :v_randbid:=DBMS_WORKLOAD_CAPTURE.IMPORT_AWR(capture_id=>34,staging_schema=>'SYSTEM',force_cleanup=>FALSE);
print :v_randbid;
V_RANDBID
----------
383379053
select id,name,status,start_time,end_time,dir_path,awr_dbid,awr_begin_snap,awr_end_snap from dba_workload_captures;
ID NAME STATUS START_TIME END_TIME DIR_PATH AWR_DBID AWR_BEGIN_SNAP AWR_END_SNAP
---------- --------------- ---------- ----------------- ----------------- -------------------- ---------- -------------- ------------
34 v1026c1 COMPLETED 20141027 20:03:35 20141027 20:13:32 /oradata01/hisdmp/mo 383379053 131 132
nthly
***dba_hist_snapshot里可以看到AWR负载已经导入
col begin_interval_time format a30
col end_interval_time format a30
set linesize 120
select snap_id,dbid,begin_interval_time,end_interval_time from dba_hist_snapshot where dbid=383379053;
SNAP_ID DBID BEGIN_INTERVAL_TIME END_INTERVAL_TIME
---------- ---------- ------------------------------ ------------------------------
131 383379053 27-OCT-14 08.00.35.987 PM 27-OCT-14 08.03.32.072 PM
132 383379053 27-OCT-14 08.03.32.072 PM 27-OCT-14 08.14.02.139 PM
###生成两个replay间的比较报告输出到v1028r1r2.html文件
set serveroutput on
spool /home/oracle/v1028r1r2.html
declare
v_rlt clob;
v_replay_id1 number:=24;
v_replay_id2 number:=26;
v_snum number:=1;
v_length number;
v_char varchar2(32767);
begin
dbms_workload_replay.compare_period_report(replay_id1=>v_replay_id1,replay_id2=>v_replay_id2,format=>'HTML',result=>v_rlt);
v_length:=dbms_lob.GETLENGTH(v_rlt);
while ( v_snum < v_length ) loop
v_char:=dbms_lob.substr(lob_loc=>v_rlt,amount=>32767,offset=>v_snum);
v_snum:=v_snum+32767;
dbms_output.put_line(v_char);
end loop;
end;
/
spool off
###生成的compare period report从hardware、top sql、Divergence等维度比对两次replay的结果
###如果要进行更细化的针对每条sql语句的分析，可以使用dbms_workload_replay.compare_sqlset_report函数，其实质是利用SPA对于两次replay期间生成的STS进行分析，函数dbms_workload_replay.compare_sqlset_report返回值是task_name，在dba_advisor_tasks中能够查到
set serveroutput on
spool /home/oracle/v1028r1r2_spa.html
declare
v_rlt clob;
v_replay_id1 number:=26;
v_replay_id2 number:=24;
v_snum number:=1;
v_length number;
v_char varchar2(32767);
v_ret varchar2(32767);
begin
v_ret:=dbms_workload_replay.compare_sqlset_report(replay_id1=>v_replay_id1,replay_id2=>v_replay_id2,format=>'HTML',result=>v_rlt);
v_length:=dbms_lob.GETLENGTH(v_rlt);
while ( v_snum < v_length ) loop
v_char:=dbms_lob.substr(lob_loc=>v_rlt,amount=>32767,offset=>v_snum);
v_snum:=v_snum+32767;
dbms_output.put_line(v_char);
end loop;
dbms_output.put_line(v_ret);
end;
/
spool off
###SPA报告截图如下
