Advanced Replication同步复制实验(基于Trigger&基于Materialized View)
1. 高级复制和流复制介绍
1.1 高级复制(Advanced Replication)
高级复制也称为对称复制，分为多主体站点复制(Multiple Master Rplication)、物化视图站点复制(Materialized View Replication)和前两种的混合复制。高级复制主要是用在对称的、等同的数据库(仅限Oracle数据库之间)表之间的(单向或双向)的复制，以满足分布式应用的需求。多主体站点复制基于Oracle的触发器(Trigger)捕获DML和DDL操作，并将这些操作封装在Remote Procedure Calls(RPCs)里，并借助Deferred Transaction Queue来传递RPCs并在目标数据库利用Internal Trigger执行传递过来的RPCs从而实现了数据的同步，如果仅仅是高级复制环境中的只读物化视图环境，那么会直接利用解析物化视图日志的方式来实现数据的同步，而不用DTQ来传递RPCs，基于这一特点，高级复制对网络的稳定性和传输速度要求较高，配置也稍显麻烦。另外一点要特别注意，就是物化视图复制方式不支持DDL的变更同步。
1.2 流复制(Stream)
流复制是Oracle 10g新推出的同步技术，基于logminer（数据库必须运行在Archive Log模式），通过Oracle Advanced Queue来实现数据的同步复制，可以实现表，用户，数据库级别的同步，而高级复制支持的对象仅为表，索引，同义词，触发器，视图，过程，函数，程序包，程序包体等。Stream相对高级复制而言配置简单,通过oracle gateway还可用于异构平台和异构数据库的数据同步，这是高级复制做不到的。
2. 基于触发器的用户表复制
环境：11.2.0.3+oel 5.7
主体站：host:zlm 192.168.1.55 sid:zlm11g/global name:zlm11g/net service name:zlm11g
复制站：host:adrep 192.168.1.65 sid:strmtarget/global name:target/net service name:target
以下操作如不指定目标端，即表示在源端操作
2.1 建立高级复制管理用户表空间，用户，并赋予权限
--source
SQL> conn sys/oracle@zlm11g as sysdba
SQL> create tablespace reptbs datafile '/u01/app/oracle/oradata/zlm11g/rep_tbs.dbf' size 20m reuse autoextend on maxsize unlimited;
SQL> create user repadmin identified by oracle default tablespace reptbs temporary tablespace temp quota unlimited on reptbs;
SQL> exec dbms_repcat_admin.grant_admin_any_schema('repadmin');
SQL> exec dbms_defer_sys.register_propagator('repadmin')
--target
SQL> conn system/oracle@target
SQL> create tablespace reptbs datafile '/u01/app/oracle/oradata/strmtarget/rep_tbs.dbf' size 20m reuse autoextend on maxsize unlimited;
SQL> create user repadmin identified by oracle default tablespace reptbs temporary tablespace temp quota unlimited on reptbs;
SQL> exec dbms_repcat_admin.grant_admin_any_schema('repadmin');
SQL> exec dbms_defer_sys.register_propagator('repadmin')
2.2 创建复制SCHEMA及复制表
--source
SQL> conn system/oracle@zlm11g
SQL> create user repuser identified by repuser default tablespace reptbs temporary tablespace temp quota unlimited on reptbs;
SQL> grant resource,connect to repuser;
SQL> conn repuser/repuser@zlm11g;
SQL> create table rptest(id number primary key,name varchar2(20));
--target
SQL> conn system/oracle@target
SQL> create user repuser identified by repuser default tablespace reptbs temporary tablespace temp quota unlimited on reptbs;
SQL> grant resource,connect to repuser;
SQL> conn repuser/repuser@target
SQL> create table rptest(id number primary key,name varchar2(20));
2.3 创建dblink
--source
SQL> conn repadmin/oracle@zlm11g
SQL> alter system set job_queue_process=10 --11g默认是1000,也可以不用设置
SQL> alter system set global_names = true; --11g默认是false，需要设置
SQL> create database link target connect to repadmin identified by oracle using 'target';
SQL> col db_link for a15
SQL> select owner,db_link from dba_db_links;
OWNER DB_LINK
------------------------------ ---------------
STRMADMIN TARGET
REPADMIN TARGET
SQL> select name from v$database@target;
NAME
---------
STRMTARG --目标端的db_name
--target
SQL> conn repadmin/oracle@target
SQL> alter system set job_queue_process=10
SQL> alter system set global_names = true;
SQL> create database link zlm11g connect to repadmin identified by oracle using 'zlm11g';
SQL> col db_link for a15
SQL> select owner,db_link from dba_db_links;
OWNER DB_LINK
------------------------------ ---------------
STRMADMIN STRMTARGET
REPADMIN ZLM11G
2.4 创建复制组
SQL> conn repadmin/oracle@zlm11g
SQL> exec dbms_repcat.create_master_repgroup('rep_gp');
SQL> select gname,master,status from dba_repgroup where gname='REP_GP';
GNAME MASTER STATUS
--------------- --------------- ---------
REP_GP Y QUIESCED
2.5 添加复制对象(此处是表)
SQL> exec dbms_repcat.create_master_repobject(sname=>'repuser',oname=>'rptest',type=>'table',use_existing_object=>true,gname=>'rep_gp',copy_rows=>false);
SQL> select sname,oname,status,gname from dba_repobject where gname='REP_GP';
SNAME ONAME STATUS GNAME
------------------------------ ------------------------------ ---------- ------------------------------
REPUSER RPTEST VALID REP_GP
2.6 对复制对象启动复制支持
SQL> exec dbms_repcat.generate_replication_support('repuser','rptest','table');
SQL> select sname,oname,status,gname from dba_repobject where gname='REP_GP';
SNAME ONAME STATUS GNAME
------------------------------ ------------------------------ ---------- ------------------------------
REPUSER RPTEST VALID REP_GP
REPUSER RPTEST$RP VALID REP_GP
REPUSER RPTEST$RP VALID REP_GP
启用复制支持以后会多出ONAME=’%ONAME$RP’的两行内容，并且目标端也的dba_repobject也会生成相应内容，源端执行复制之前的内容是空的
2.7 添加复制主体节点
SQL> exec dbms_repcat.add_master_database(gname=>'rep_gp',master=>'target',
use_existing_objects=>true,copy_rows=>false,propagation_mode=>'synchronous');
SQL> col gname format a15
SQL> col dblink format a15
SQL> col masterdef for a15
SQL> col master format a15
SQL> select gname,dblink,masterdef,master from dba_repsites where gname='REP_GP';
GNAME DBLINK MASTERDEF MASTER
--------------- --------------- --------------- ---------------
REP_GP ZLM11G Y Y
REP_GP TARGET N Y
2.8 登录主体复制节点(目标端)查看复制站点信息和复制对象信息
SQL> conn repadmin/oracle@target
SQL> select gname,dblink,masterdef,master from dba_repsites where gname='REP_GP';
GNAME DBLINK MASTERDEF MASTER
--------------- --------------- --------------- ---------------
REP_GP TARGET N Y
REP_GP ZLM11G Y Y
SQL> select sname,oname,status,gname from dba_repobject where gname='REP_GP';
SNAME ONAME STATUS GNAME
------------------------------ ------------------------------ ---------- ---------------
REPUSER RPTEST VALID REP_GP
REPUSER RPTEST$RP VALID REP_GP
REPUSER RPTEST$RP VALID REP_GP
2.9 主体定义站启动复制
SQL> exec dbms_repcat.resume_master_activity('rep_gp');
2.10 测试复制同步(DML)
2.10.1 主体定义站测试用户修改测试表
SQL> conn repuser/repuser@zlm11g
Connected.
SQL> insert into rptest values(1,'aaron8219');
1 row created.
SQL> commit;
Commit complete.
2.10.2 主体站登录测试用户验证测试表同步
SQL> conn repuser/repuser@target
Connected.
SQL> select * from rptest;
ID NAME
---------- --------------------
1 aaron8219
可以看到，表数据可以已经同步
这里还碰到了一个小插曲，就是原先没有先在主体站(target端)先创建dblink，等创建完复制组，复制对象，启用复制支持，创建复制站后，发现不能同步，再去补创建dblink后，删除复制组，复制对象后依然不见效，复制定义主站对表进行操作时，皆会报如下错误：
SQL> insert into rptest values(3,'zhao');
insert into rptest values(3,'zhao')
*
ERROR at line 1:
ORA-24277: invalid database link "REPUSER"."RPTEST$RP"."REP_INSERT"
ORA-02063: preceding line from TARGET
SQL> delete from rptest;
delete from rptest
*
ERROR at line 1:
ORA-24277: invalid database link "REPUSER"."RPTEST$RP"."REP_DELETE"
ORA-02063: preceding line from TARGET
[oracle@zlm ~]$ oerr ora 24277
24277, 00000, "invalid database link %s "
// *Cause: The database link did not exist.
// *Action: Make sure the database link name is correct.
查看了一下，关于这个报错，是因为dblink名字不对，可能和之前没有立即在target端创建到主体定义站的dblink，之后才创建有关。删除dblink后重新再两端创建dblink，再把复制组，复制对象过程重新再走一遍，问题解决。
2.11 测试复制同步(DDL)
2.11.1 修改表结构
如果要做ddl同步，需要把复制组先挂起，做ddl后重新生成复制支持，然后把组resume
SQL> conn repadmin/oracle@zlm11g
SQL> exec dbms_repcat.suspend_master_activity('rep_gp');
SQL> exec dbms_repcat.alter_master_repobject(sname => 'repuser',oname => 'rptest',type => 'table',ddl_text =>'alter table repuser.rptest add(city varchar(10))');
SQL> exec dbms_repcat.generate_replication_support(sname => 'repuser',oname => 'rptest',type => 'table',min_communication => true,generate_80_compatible => false);
SQL> select gname,master,status from dba_repgroup where gname='REP_GP';
SQL> select sname,oname,status,gname from dba_repobject where gname='REP_GP';
2.11.2 验证结果
SQL> conn repuser/repuser@zlm11g
Connected.
SQL> set lin 30 pages 200
SQL> desc rptest
Name Null? Type
----------------- -------- ------------
ID NOT NULL NUMBER
NAME VARCHAR2(20)
CITY VARCHAR2(10)
SQL> conn repuser/repuser@target
Connected.
SQL> set lin 30 pages 200
SQL> desc rptest
Name Null? Type
----------------- -------- ------------
ID NOT NULL NUMBER
NAME VARCHAR2(20)
CITY VARCHAR2(10)
可以看到，修改表结构的DDL操作也已经同步过来了
SQL> show user
USER is "REPUSER"
SQL> insert into rptest values(2,'zlm','shanghai');
insert into rptest values(2,'zlm','shanghai')
*
ERROR at line 1:
ORA-23326: object group
"PUBLIC"."REP_GP" is quiesced
SQL> !
[oracle@zlm ~]$ oerr ora 23326
23326, 00000, "object group \"%s\".\"%s\" is quiesced"
// *Cause: Either suspend_master_activity has been called before the object
// group has resumed normal operation or a (deferred) rpc operation
// was attempted while the object group was quiesced.
// *Action: If suspend_master_activity has been called and a
// resume_master_activity request is pending, wait until it
// completes, and then reinvoke suspend_master_activity.
// Otherwise, resume database activity with the resume_master_activity
// call.
[oracle@zlm ~]$ exit
exit
SQL> conn repadmin/oracle@zlm11g
Connected.
SQL> exec dbms_repcat.resume_master_activity('rep_gp');
PL/SQL procedure successfully completed.
SQL> select gname,master,status from dba_repgroup where gname='REP_GP';
GNAME MASTER STATUS
--------------- --------------- ---------
REP_GP Y NORMAL
SQL> conn repuser/repuser@zlm11g
Connected.
SQL> insert into rptest values(2,'zlm','shanghai');
1 row created.
SQL> commit;
Commit complete.
SQL> conn repuser/repuser@target
SQL> set lin 120 pages 200
SQL> select * from rptest;
ID NAME CITY
---------- -------------------- ----------
2 zlm shanghai
1 aaron8219
2.12 清除操作
如果实验中遇到不能同步，可以删除以下对象后重新做一遍(通常只需删除复制组，对象会一并删除)
删除用户
SQL> drop user repadmin cascade;
SQL> drop user repuser cascade;
删除复制对象
SQL> exec dbms_repcat.drop_master_repobject(sname=>'repuser',oname =>'rptest',
type=>'table');
删除复制组
SQL> exec dbms_repcat.drop_master_repgroup(gname=>'rep_gp',all_sites=>true);
检查错误的管理请求
SQL> select * from dba_repcatlog where status ='ERROR';
3. 基于物化视图的复制
3.1 源端赋予复制用户权限
SQL> conn system/oracle@zlm11g
SQL> grant comment any table to repuser;
SQL> grant lock any table to repuser;
SQL> grant execute any procedure to repuser;
SQL> grant create any table to repuser;
SQL> grant create materialized view to repuser;
3.2 源端创建测试表
SQL> conn repuser/repuser@zlm11g
SQL> create table mvtest(id number(5) primary key);
SQL> begin
for i in 1..5 loop
insert into mvtest values(i);
end loop;
end;
/
3.3 源端创建物化视图日志
SQL> create materialized view log on mvtest with primary key including new values;
3.4 目标端赋予复制用户权限
SQL> conn system/oracle@target
SQL> grant create database link to repuser;
SQL> grant create materialized view to repuser;
SQL> grant execute any procedure to repuser;
3.5 目标端创建物化视图
3.5.1 创建DBlink
SQL> conn repuser/repuser@target
SQL> alter system set global_names = true;
SQL> create database link zlm11g connect to repuser identified by repuser using 'zlm11g';
SQL> select db_link,owner from all_db_links;
DB_LINK OWNER
--------------- ------------------------------
ZLM11G REPUSER
3.5.2 创建测试表
SQL> create table mvtest as select * from repuser.mvtest@zlm11g where 1=2;
SQL> desc mvtest;
Name Null? Type
----------------------- -------- ----------------
ID NOT NULL NUMBER(5)
SQL> select * from mvtest;
no rows selected
3.5.3 创建物化视图
SQL> create materialized view mvtest
on prebuilt table
refresh force with primary key
on demand
as select * from repuser.mvtest@zlm11g;
3.6 全量刷新
SQL> exec dbms_mview.refresh('mvtest','complete');
3.7 验证目标端数据
SQL> conn repuser/repuser@target
SQL> select * from mvtest;
ID
----------
1
2
3
4
5
SQL> conn repuser/repuser@zlm11g
SQL> insert into mvtest values(6);
1 row created.
SQL> commit;
Commit complete.
SQL> conn repuser/repuser@target
SQL> select * from mvtest;
ID
----------
1
2
3
4
5
源端新插入的一条记录并没有同步过来，目标端再执行一次全量刷新
SQL> conn repuser/repuser@target
SQL> exec dbms_mview.refresh('mvtest','complete');
SQL> select * from mvtest;
ID
----------
1
2
3
4
5
6
又一次同步了mvtest表的数据，以上是最简单的基于手动的刷新物化视图，另外还可以配置job，让物化视图定期刷新。注意，物化视图复制并不支持DDL的复制
查看源端复制用户基表中的物化视图日志
SQL> select * from repuser.mlog$_mvtest;
查看目标物化视图的刷新情况
SQL> select mview_name,owner from user_mviews;
SQL> select * from user_jobs;
SQL> select * from dba_jobs_running;
查看物化视图所在的基表最后被成功刷新的时间
SQL> select master,log_table,current_snapshots from dba_snapshot_logs;
4 总结
高级复制(Advanced Replication)可以说是Oracle最早的HA技术了，最早出现于Oracle 7e，之后的8i,9i继续沿用，一直到10g，出现了流技术(Stream)，高级复制技术才逐渐开始被取代，虽然一直到11g，还是可以使用AR技术，但其弊端是显而易见的，由于它的复制机制是基于Oracle的内部对象触发器(Trigger)，他会严格控制事务的一致性，对于开发人员而言，也许是不错的特性，但是对于DBA来说，可谓是麻烦重重，一旦事务之间互相请求的资源被别的事务所占用，往往会造成等待，更严重的就是产生死锁，造成数据库hang住。虽然如此，高级复制技术的物化视图，还是有一定的利用价值的，因为它并不通过内部触发器来控制对象的复制，这样就可以把容易消耗资源的查询分散到物化视图上，通过查询实时更新的物化视图来获取需要的报表数据，这对于提高数据库性能是由一定积极作用的。
-------------------------------------------------------------------------------------------------------
By aaron8219 Chinaunix Blog：http://blog.chinaunix.net/uid/24612962.html
原创内容，转载请注明链接，谢谢！
http://blog.csdn.net/aaron8219/article/details/11707137