- 一个OGG实例最多支持300个并发extract和replicat进程，每个进程需要25-55M内存。
- GGSCI，view report，找到PROCESS VM AVAIL FROM OS，大约就是系统所需要的swap space大小
3． 磁盘空间每个实例50-150M，外加trail等等，至少给1G。trail所需空间的预估方法是：[log volume in one hour] x [number of hours downtime] * 0.4
4．/tmp 空间，默认是dirtmp目录，可以通过CACHEMGR参数的CACHEDIRECTORY选项指定
5． 安装OGG的用户必须属于ORACLE数据库实例的组，通常就是dba。
6．要想支持RAC里的RMAN log retention特性, 得在add extract之前安装patch BUGFIX 11879974
- OGG的几种架构
如果只有一个数据库实例，那么就设置在profile里，如果有多个实例，就在extract 和 replicat的参数文件里设置，例如：
EXTRACT ora9a
SETENV (ORACLE_HOME = “/home/oracle/ora9/product”)
SETENV (ORACLE_SID = “ora9a”)
SETENV (NLS_LANG = “AMERICAN_AMERICA.AL32UTF8”)
USERID ggsa, PASSWORD ggsa
RMTHOST sysb
RMTTRAIL /home/ggs/dirdat/rt
TABLE hr.emp;
也可以在每次启动的时候，将
export ORACLE_HOME
export ORACLE_SID
放到特定的启动脚本里,然后启动进程
另外需要设置的
export PATH=<installation directory>:$PATH
export LIBPATH=<absolute path of installation directory>:$ LIBPATH
解压安装包
ggsci
create subdirs
exit
解压安装包
ggsci
create subdirs
exit
可以将mgr配置成一个服务，随着系统自动启动，可以配置一个客户化的服务名字
ggsci
edit params ./GLOBALS
MGRSERVNAME <name>
Install addevent addservice [autostart | manualstart]
create user oggadm
identified by oggadm
default tablespace users
temporary tablespace temp;
extract:
grant CREATE SESSION, ALTER SESSION to oggadm;
grant RESOURCE to oggadm;
grant CONNECT to oggadm;
grant SELECT ANY DICTIONARY to oggadm;
grant FLASHBACK ANY TABLE to oggadm;
grant SELECT ANY TABLE to oggadm;
grant SELECT on dba_clusters to oggadm;
grant EXECUTE on DBMS_FLASHBACK to oggadm;
sys.dbms_internal_clkm 复制Oracle Transparent Data Encryption (TDE)时需要.
SELECT ANY TRANSACTION 新的ASM特性时需要
10gR2之后，rman会协同extract进行归档日志的保留策略，需要如下权限
10g2
exec dbms_streams_auth.grant_admin_privilege('oggadm');
grant insert on system.logmnr_restart_ckpt$ to oggadm;
grant update on sys.streams$_capture_process to oggadm;
grant become user to oggadm;
11.1 and 11.2.0.1
exec dbms_streams_auth.grant_admin_privilege('oggadm');
grant become user to oggadm;
11.2.0.2 and later
exec dbms_goldengate_auth.grant_admin_privilege('oggadm');
replicat:
grant CREATE SESSION, ALTER SESSION to oggadm;
grant RESOURCE to oggadm;
grant CONNECT to oggadm;
grant SELECT ANY DICTIONARY to oggadm;
grant SELECT ANY TABLE to oggadm;
grant CREATE TABLE to oggadm;
grant LOCK ANY TABLE to oggadm;
INSERT, UPDATE, DELETE ON <target tables>
在Source和 target都得执行，
grant dba to oggadm;
ggsci
EDIT PARAMS ./GLOBALS
GGSCHEMA oggadm
SQL> @sequence.sql
在source 系统里
GRANT EXECUTE on oggadm.updateSequence TO oggadm;
在target 系统里
GRANT EXECUTE on oggadm.replicateSequence TO oggadm;
DDL支持配置完毕，就可以支持sequences (CREATE, ALTER, DROP, RENAME) 操作的复制的，但如果只是单纯的想复制sequence的话，可以不配置DDL支持，只需要sequence参数就行。
GGSCI> dblogin userid oggadm@source,password oggadm
FLUSH SEQUENCE <owner.sequence> 或者wildcard
在initial synchronization之后，必须执行的就是在target上对没一个sequence至少取一次nextval
配置DDL环境，需要在数据库里创建如下一系列的表：
GRANT EXECUTE ON utl_file TO oggadm;
ggsci
EDIT PARAMS ./GLOBALS
GGSCHEMA oggadm
Sqlplus “/as sysdba”
@marker_setup.sql 必须到这个脚本所在目录进入sqlplus
@ddl_setup.sql
@role_setup.sql
Grant GGS_GGSUSER_ROLE to oggadm;
@ddl_enable.sql
@$ORACLE_HOME\rdbms\admin\dbmspool.sql
@ddl_pin oggadm;
开启和关闭DDL trigger
ddl_disable
ddl_enable
维护DDL marker表
Manager 参数文件的PURGEMARKERHISTORY提供设置最大和最小保留时间（根据最后一个修改动作的时间计算）
维护DDL history表
Manager 参数文件的PURGEDDLHISTORY参数设置最大和最小保留时间
清空DDL trace文件
运行ddl_cleartrace脚本删除trace文件，默认名字是ggs_ddl_trace.log在oracle 的 USER_DUMP_DEST目录里
DDL复制只支持两个系统之间的负责，active-passive 或者 active-active，而且必须是两个object必须是identical的，必须使用ASSUMETARGETDEFS参数。
如果使用数据泵的话，可以将PASSTHRU放在所有的要进行DDL复制的TABLE语句前，你可以把NOPASSTHRU放在任何不进行DDL复制而需要进行data转换的TABLE之前。数据泵不能使用任何DDL配置参数，比如DDL, DDLOPTIONS,
DDLSUBST, PURGEDDLHISTORY, PURGEMARKERHISTORY, DDLERROR
DDL &
INCLUDE UNMAPPED &
OPTYPE alter &
OBJTYPE ‘table’ &
OBJNAME “users.tab*” &
INCLUDE MAPPED OBJNAME “*” &
EXCLUDE MAPPED OBJNAME "temporary.tab*"
1. Run GGSCI.
2. STOP EXTRACT <group>
3. STOP REPLICAT <group>
4. Run SQL*Plus and log in as a user that has SYSDBA privileges.
5. Run the ddl_disable script.
6. Run the ddl_remove script,
7. Run the marker_remove script.
8. Run the marker_setup script.
9. Run the ddl_setup script.
10. Run the role_setup script.
11. Grant the role to all Oracle GoldenGate users
12. Run the ddl_enable.sql script.
1. Run GGSCI.
2. STOP EXTRACT <group>
3. STOP REPLICAT <group>
4. Run SQL*Plus and log in as a user that has SYSDBA privileges.
5. Run the ddl_disable script .
6. Run the ddl_remove script
7. Run the marker_remove script.
alter system set recyclebin=off;
sqlplus “/as sysdba”
ALTER DATABASE ADD SUPPLEMENTAL LOG DATA;
ALTER SYSTEM SWITCH LOGFILE;
SELECT SUPPLEMENTAL_LOG_DATA_MIN FROM V$DATABASE;
ADD SCHEMATRANDATA:当配置DDL复制的时候应该使用，DDL不使用的时候，可以代替ADD TRANDATA，表没有主键和唯一键的时候，使用ADD TRANDATA的cols和table map的keycols来配合使用ADD SCHEMATRANDATA最好。
ADD TRANDATA：不适用DDL复制特性的时候使用，在下保证列情况下，在DDL复制环境下也可以使用ADD TRANDATA:在对表进行DDL时，一定不能有任何DML，可以有对其他不相关表的DML
1. (ADD SCHEMATRANDATA only) Apply Oracle Patch 10423000 to the source Oracle database. Oracle数据库11g以后才支持ADD/INFO/DELETE SCHEMATRANDATA命令
先sys执行exec dbms_streams_auth.grant_admin_privilege('oggadm');
2. On the source system, run GGSCI.
3. dblogin userid oggadm@source,password oggadm
4. Issue the ADD TRANDATA or ADD SCHEMATRANDATA command.
ADD TRANDATA <table> [, COLS <columns>][,NOKEY]
其中的COLS是配合extract中的TABLE参数和replicat中的MAP参数里的KEYCOLS子句来使用的，如果指定NOKEY，那就必须指定COLS，也必须指定KEYCOLS字句
ADD SCHEMATRANDATA <schema>
查看效果
INFO TRANDATA <owner/schema>.TCUST*
10.2.0.5.0以后，Replicat parameter DBOPTIONS with the SUPPRESSTRIGGERS option
9.2.0.7.0 以后，Replicat parameter DBOPTIONS with the DEFERREFCONST option
再之前的数据库版本就得手工停止trigger，或者手工忽略Replicat database user
Source上是DEFERRABLE的约束，在target里必须也是DEFERRABLE的，在Replicat parameter file文件里，添加一行SQLEXEC (“alter session set constraint deferred”)
9.2.0.7.0 以后，Replicat parameter DBOPTIONS with DEFERREFCONST
如果target上的表没有主键或唯一健，那么就会有重复记录存在，那么应用过来的操作就会针对多条记录进行操作，造成source和target不一致，需要设置一个选项DBOPTIONS LIMITROWS in the Replicat parameter file.
Oracle GoldenGate supports the capture and replication of TIMESTAMP WITH TIME ZONE as a UTC offset (TIMESTAMP '2011-01-01 8:00:00 -8:00') but abends on TIMESTAMP WITH TIME ZONE as TZR (TIMESTAMP '2011-01-01 8:00:00 US/Pacific') by default。
要想支持TIMESTAMP WITH TIME ZONE as TZR, 得在Extract parameter 文件里使用选项TRANLOGOPTIONS 如下两个选项：
INCLUDEREGIONID 从oracle到同版本oracle或更高版本
INCLUDEREGIONIDWITHOFFSET 从oracle到低版本或者非oracle
Ggsci
EDIT PARAMS MGR
port 7809
AUTOSTART ER *
AUTORESTART EXTRACT *,RETRIES 3,WAITMINUTES 5,RESETMINUTES 60
DYNAMICPORTLIST 7940-8100,7930
SOURCEDB gzcrm20,USERID gguser,PASSWD gguser
PURGEOLDEXTRACTS /ggs/dirdat/aa*,USECHECKPOINTS,MINKEEPDAYS 2
LAGREPORTHOURS 1 检查lag的间隔
LAGINFOMINUTES 3 报info
LAGCRITICALMINUTES 5 报error
INFO MGR
创建checkpoint table有两种方法，
一种是使用chkpt_ora_create.sql脚本创建，
另一种方式是使用ggsci里的命令如下：
dblogin userid oggadm@source,password oggadm
add checkpointtable oggadm.ggs_checkpoint
通过EDIT PARAMS ./GLOBALS加一行CHECKPOINTTABLE oggadm.ggs_checkpoint来指定checkpoint table的默认位置
如果没有在GLOBALS中添加的话，需要添加replicat的时候指定
ADD REPLICAT emp_rep, EXTTRAIL C:\OGG10G\dirdat\et, CHECKPOINTTABLE oggadm.ggs_checkpoint
如果不想使用checkpoint，只是使用默认的trail文件中记录的checkpoint接受数据不一致的话，可以使用NODBCHECKPOINT
ADD EXTRACT ext_src, TRANLOG, BEGIN NOW
ADD EXTTRAIL G:\OGG\V26068-01-source\dirdat\et, EXTRACT ext_src
extract ext1
userid ggs,password ggs
exttrail g:\ogg\v26068-01-source\dirdat\et
rmthost 192.168.234.112, mgrport 7809
rmttrail c:\ggs\dirdat\r1
dynamicresolution
gettruncates
TRANLOGOPTIONS INCLUDEREGIONID,EXCLUDEUSER oggadm
DDL include all
table ggs.*;
SEQUENCE hr.employees_seq;
DDLERROR [RESTARTSKIP <1-10万>] [SKIPTRIGGERERROR <num errors>]
捕获DDL errors，Extract最多可以跳过十万个DDL 操作而不终止
DDLOPTIONS GETAPPLOPS, IGNOREREPLICATES（默认配置）
IGNOREREPLICATES 和 GETREPLICATES 控制extract是否忽略由replicat产生的ddl操作
GETAPPLOPS 和 IGNOREAPPLOPS 控制extract是否忽略由应用等其他操作产生的ddl操作
要想避免数据的looping复制，要做到两点：
第一点是合理配置IGNOREREPLICATES 和 GETREPLICATES，GETAPPLOPS 和 IGNOREAPPLOPS这两对参数；
第二点是identify local Replicat transactions，这个使用TRANLOGOPTIONS EXCLUDEUSER oggadm来实现
下面是针对DML的配置，直接写在extract文件里即可
GETAPPLOPS
IGNOREREPLICATES
下面是DDLOPTIONS里的配置
bi-directional模式里，需要配置成
Extract (primary and secondary)
DDLOPTIONS GETREPLICATES, GETAPPLOPS
Replicat (primary and secondary)
DDLOPTIONS UPDATEMETADATA
Cascading模式
DDLOPTIONS GETREPLICATES, IGNOREAPPLOPS
一个实验过的例子：
edit params ext_src
extract ext_src
userid oggadm@source,password oggadm
EXTTRAIL G:\OGG\V26068-01-source\dirdat\et
dynamicresolution
gettruncates
TRANLOGOPTIONS INCLUDEREGIONID,EXCLUDEUSER oggadm
GETAPPLOPS
IGNOREREPLICATES
DDL include all
DDLOPTIONS GETAPPLOPS, GETREPLICATES
table source.*;
SEND EXTRACT <name> SHOWTRANS 查看extract需要的最早的log信息
ADD EXTRACT pump_src, EXTTRAILSOURCE G:\OGG\V26068-01-source\dirdat\et, BEGIN now
ADD RMTTRAIL G:\OGG\V26068-01-target\dirdat\rt, EXTRACT pump_src
Edit params pump_src
extract pump_src
userid oggadm@source,password oggadm
rmthost pc2011022014ptf, mgrport 7810
rmttrail g:\ogg\v26068-01-target\dirdat\rt
passthru
table source.*;
ADD REPLICAT rep_tag, EXTTRAIL G:\OGG\V26068-01-target\dirdat\rt, BEGIN now
replicat rep1
userid oggadm@target,password oggadm
assumetargetdefs
reperror (default,discard,abend)
DBOPTIONS LIMITROWS SUPPRESSTRIGGERS
DDL include all
HANDLECOLLISIONS 此参数在source active状态时做initial load才需要，initial load 之后SEND REPLICAT <Replicat group name>, NOHANDLECOLLISIONS
discardfile c:\oradata\repsz.dsc,append,megabytes 100
map ggs.*, target ggs.*;
MAP hr.employees_seq, TARGET payroll.employees_seq;
DDLOPTIONS UPDATEMETADATA
DDLERROR 捕获DDL errors
{<error> | DEFAULT} {<response>}
{INCLUDE <inclusion clause> | EXCLUDE <exclusion clause>}
[IGNOREMISSINGOBJECTS | ABENDONMISSINGOBJECTS]
例如：DDLERROR <error> IGNORE RETRYOP MAXRETRIES 3 RETRYDELAY 10 &
INCLUDE ALL OBJTYPE TABLE OBJNAME “tab*” EXCLUDE OBJNAME “tab1*”
一个实验用过的例子
Edit params rep_tag
replicat rep_src
userid oggadm@source,password oggadm
assumetargetdefs
reperror (default,discard,abend)
DBOPTIONS LIMITROWS DEFERREFCONST
DDL include all
DDLOPTIONS UPDATEMETADATA
map target.*, target source.*;
START REPLICAT
Syntax START REPLICAT <group name> ATCSN <csn> | AFTERCSN <csn>]
ATCSN <csn> 小于这个csn值的操作被跳过.
AFTERCSN <csn> 大于这个csn值的操作被跳过。
ADD EXTRACT ext_tag, TRANLOG, BEGIN NOW
ADD EXTTRAIL G:\OGG\V26068-01-target\dirdat\et, EXTRACT ext_tag
一个实验用过的例子
Edit params ext_tag
extract ext_tag
userid oggadm@target,password oggadm
EXTTRAIL G:\OGG\V26068-01-target\dirdat\et
dynamicresolution
gettruncates
TRANLOGOPTIONS INCLUDEREGIONID
DDL include all
DDLOPTIONS GETREREPLICATES, GETAPPLOPS
table target.*;
ADD EXTRACT pump_tag, EXTTRAILSOURCE G:\OGG\V26068-01-target\dirdat\et, BEGIN now
ADD RMTTRAIL G:\OGG\V26068-01-source\dirdat\rt, EXTRACT pump_tag
Edit params pump_tag
extract pump_tag
userid oggadm@target,password oggadm
rmthost pc2011022014ptf, mgrport 7809
rmttrail g:\ogg\v26068-01-source\dirdat\rt
passthru
table target.*;
ADD REPLICAT rep_src, EXTTRAIL G:\OGG\V26068-01-source\dirdat\rt, BEGIN now
Edit params rep_src
replicat rep_src
userid oggadm@source,password oggadm
assumetargetdefs
reperror (default,discard,abend)
DBOPTIONS LIMITROWS DEFERREFCONST
DDL include all
map target.*, target source.*;
DDLOPTIONS UPDATEMETADATA
在extract中使用filter和where选行
TABLE <table spec>, FILTER ([, ON INSERT | ON UPDATE| ON DELETE][, IGNORE INSERT | IGNORE UPDATE | IGNORE DELETE], <filter clause>);
MAP <table spec>, TARGET <table spec>,
, FILTER ([, ON INSERT | ON UPDATE| ON DELETE][, IGNORE INSERT | IGNORE UPDATE | IGNORE DELETE][, RAISEERROR <error_num>], <filter clause>);
使用cols和colsexcept选列
column mapping分为表级别的和global级别的，在TABLE和MAP参数里 ，使用COLMAP选项，如果source和target的表是identical的，那么就只使用COLMAP选项就行，但是如果不是identical的，就要创建data-definitions file了
表级别的column mapping，如果使用了COLMAP，那么在TABLE中就得像MAP语句一样，增加一个TARGET选项，例如：
MAP sales.acctbl,TARGET sales.accttab,
COLMAP (
USEDEFAULTS,
name = cust_name,
transaction_date =@DATE (“YYYY-MM-DD”, “YY”,YEAR, “MM”, MONTH, “DD”, DAY),
area_code =@STREXT (phone_no, 1, 3),
phone_prefix =@STREXT (phone_no, 4, 6),
phone_number =@STREXT (phone_no, 7, 10));
GLOBAL级别的column mapping，如下：
COLMATCH NAMES customer_code = cust_code
COLMATCH NAMES customer_name = cust_name
COLMATCH NAMES customer_address = cust_addr
COLMATCH PREFIX S_
COLMATCH SUFFIX _tab
MAP sales.acct, TARGET sales.account, COLMAP (USEDEFAULTS);
MAP sales.ord, TARGET sales.order, COLMAP (USEDEFAULTS);
COLMATCH RESET
MAP sales.reg, TARGET sales.reg;
MAP sales.price, TARGET sales.price;
Tokens,传递数据用
TABLE ora.oratest,TOKENS(TK-HOST=@GETENV("GGENVIRONMENT","HOSTNAME"));
MAP
ora.oratest,TARGET ra.rpt,COLMAP(USEDEFAULTS,host=@token("TK-HOST"));
source 和 target 的表有不同的data definitions（比如oracle到sql server，列名不同，数据类型不同，列长度不一样，表中列的顺序不一致）的时候，就要使用data-definitions file，在source上创建source-definitions file, 在target上创建target-definitions file,步骤一样：
ggsic
EDIT PARAMS DEFGEN
DEFSFILE C:\ggs\dirdef\record.def
USERID oddadm, PASSWORD oggadm
TABLE acct.cust100, DEF custdef; 此处定义了一个模板
TABLE ord.*;
TABLE hr.*;
保存退出后，运行defgen工具来生成出相应的data-definitions file
defgen paramfile dirprm/defgen.prm [reportfile dirrpt/defgen.rpt]
不指定reportfile的话，报告信息就只是输出到屏幕上
引用相应的data-definitions file的方法，是在extract参数文件里添加TARGETDEFS参数，在replicat参数文件里添加SOURCEDEFS参数
上面的例子对应的replicat的参数文件如下：
REPLICAT acctrep
SOURCEDEFS c:\ggs\dirdef\record.def
USERID ggs, PASSWORD ggs
MAP acct.cust*, TARGET acct.cust*, DEF custdef;
MAP ord.prod, TARGET ord.prod;
MAP ord.parts, TARGET ord.parts;
MAP hr.emp, TARGET hr.emp;
MAP hr.salary, TARGET hr.salary;
● “Loading data with a database utility”
● “Loading data from file to Replicat”
Extract writes records to an extract file and Replicat applies them to the target tables. This is the slowest initial-load method.
SOURCEISTABLE
SPECIALRUN
END RUNTIME
● “Loading data from file to database utility”
Extract writes records to extract files in external ASCII format. The files are used as data files for input into target tables by a bulk load utility. Replicat creates the run and control files.
SOURCEISTABLE
FORMATASCII, {BCP | SQLLOADER}
GENLOADFILES <template file>
● “Loading data with an Oracle GoldenGate direct load”
Extract communicates with Replicat directly across TCP/IP without using a Collector process or files. Replicat applies the data through the database engine.
ADD EXTRACT <initial-load Extract name>, SOURCEISTABLE
EXTRACT <initial-load Extract name>
ADD REPLICAT <initial-load Replicat name>, SPECIALRUN
REPLICAT <initial-load Replicat name>
- 关闭回收站和配置附加日志
- initiate the logging of key values
- 防止由于triggers 和cascade constraints带来的数据重复修改
- Deferring constraint checking
- Limiting row changes in tables that do not have a key
- 复制TIMESTAMP with TIME ZONE
- “Loading data with a direct bulk load to SQL*Loader” on page 232. Extract extracts records in external ASCII format and delivers them directly to Replicat, which delivers them to Oracle’s SQL*Loader bulk-load utility. This is the fastest method of loading Oracle data with Oracle GoldenGate.