Oracle数据库10gR2中stream多源复制方法

  在10gR2中,oracle简化了stream的配置的全过程,在9iR2及10gR1中,需要针对stream的捕获、传播、应用进程进行配置的步骤
  已经被dbms_streams_adm新提供的过程(pre_instantiation_setup/post_instantiation_setup)给封装起来啦,
  配置stream只需要调用两个存储过程就可以搞定啦,真是太激动人心啦,下面简单地介绍一下dbms_streams_adm的新过程的用法。
  1.在stream进行配置前,需要做些准备工作
  a 源库与目标库初始化参数的设置
  alter system set aq_tm_processes=4 scope=spfile;
  alter system set job_queue_processes=5 scope=spfile;
  alter system set global_names=true scope=spfile;
  alter system set streams_pool_size=51m scope=spfile;
  说明streams_pool_size在生产环境中最好>200m
  b 源库与目标库tnsnames.ora配置
  确保正确,可用tnsping通
  c 源库与目标库复制管理员的创建
  create user strmadmin identified by strmadminpw
  default tablespace &tbs_name quota unlimited on &tbs_name;
  grant connect, resource, dba to strmadmin;
  d 源库与目标库创建互连的数据链
  connect strmadmin/strmadminpw@test96;
  create database link test99.net connect to strmadmin
  identified by strmadminpw using 'test99';
  connect strmadmin/strmadminpw@test99;
  create database link test96.net connect to strmadmin
  identified by strmadminpw using 'test96';
  说明:必须确保双方的数据库链是可以连通.
  用pre_instantiation_setup/post_instantiation_setup过程时
  db link必须用db_name.domain的格式
  e 源库与目标库必须处于归档模式
  shutdown immediate;
  startup mount;
  alter database archivelog;
  alter database open;
  2.执行pre_instantiation_setup过程
  在调用dbms_streams_adm的pre_instantiation_setup/post_instantiation_setup过程时, 它们必须成对出现,pre_instantiation_setup过程中,maintain_mode参数可取GLOBAL与 TRANSPORTABLE TABLESPACES,如果取GLOBAL时,表示全库复制,否则可以定义需要复制的表空间; perform_actions参数为TRUE,进行配置产生的源脚本将记录在dba_recoverable_*字典表, 如果pre_instantiation_setup执行时遇到错误,可以通过执行dbms_steams_adm的
  recover_operation过程 在更正错误后继续执行复制配置; source_database/destination_database是我们已经创建好的dblink,必须确保正确啊, 否则pre_instantiation_setup过程将会失败,报ORA-23621错误,解决办法在后面会介绍; bi_directional参数设置为true时,表示是多源复制,即目标与源库双向同步, 否则即只从源库向目标库同步数据; exclude_schemas参数指需要排除在全库同步的表空间,多个表空间的话用逗号分开,
  *表示排除配置stream时库中已经存在的表空间;
  start_processes参数指配置完成后启动捕获及应用进程。
  SQL>connect strmadmin/strmadminpw@test96;
  declare
  empty_tbs dbms_streams_tablespace_adm.tablespace_set;
  begin
  dbms_streams_adm.pre_instantiation_setup(
  maintain_mode => 'GLOBAL',
  tablespace_names => empty_tbs,
  source_database => 'test96.net',
  destination_database => 'test99.net',
  perform_actions => true,
  bi_directional => true,
  include_ddl => true,
  start_processes => true,
  exclude_schemas => 'WMSYS,STRMADMIN,DBSNMP,TSMSYS,',
  exclude_flags => dbms_streams_adm.exclude_flags_unsupported +
  dbms_streams_adm.exclude_flags_dml + dbms_streams_adm.exclude_flags_ddl);
  end;
  /
  如果参与复制的源库与目标库的db link等配置的正确,该过程将成功结束,
  并且在strmadmin模式创建一系统用于复制的队列与字典表,如果db link配置出错,
  执行pre_instantiation_setup过程时,将会报如下错误:
  ORA-23621: Operation corresponding
  3.用rman复制源库到目标库
  a 对源库用rman进行备份,拷贝备份集与产生的归档到目标库,
  并将目标库down下来,启动nomount状态。
  rman nocatalog target /
  rman>backup database;
  rman>sql'alter system archive log current';
  b 求得源库的scn
  SQL>connect strmadmin/strmadminpw@test96;
  SQL>set serveroutput on size 1000000
  SQL>declare
  until_scn number;
  begin
  until_scn:= dbms_flashback.get_system_change_number;
  dbms_output.put_line('until scn: '||until_scn);
  end;
  /
  until scn: 429596
  c 用rman将源库复制到目标库
  rman nocatalog target /
  rman> connect auxiliary sys/sys@test99;
  rman> run
  {
  set until scn 429596;
  duplicate target database to 'TEST'
  nofilenamecheck
  open restricted;
  }
  d 重新命名目标库的global_name
  alter database rename global_name to test99.net;
  e 重新创建目标库的db link
  connect strmadmin/strmadminpw@test99;
  create database link test96.net connect to strmadmin
  identified by strmadminpw using 'test96';
  4.执行post_instantiation_setup过程
  post_instantiation_setup也在源库执行,需要注意的参数是instantiation_scn
  它的取值是我们从源库上获的scn的值-1=429595.
  SQL>connect strmadmin/strmadminpw@test96;
  SQL>declare
  empty_tbs dbms_streams_tablespace_adm.tablespace_set;
  begin
  dbms_streams_adm.post_instantiation_setup(
  maintain_mode => 'GLOBAL',
  tablespace_names => empty_tbs,
  source_database => 'test96.net',
  destination_database => 'test99.net',
  perform_actions => true,
  bi_directional => true,
  include_ddl => true,
  start_processes => true,
  instantiation_scn => 429595,
  exclude_schemas => '*',
  exclude_flags => dbms_streams_adm.exclude_flags_unsupported +
  dbms_streams_adm.exclude_flags_dml + dbms_streams_adm.exclude_flags_ddl);
  end;
  /
  在目标库禁止restricted session
  SQL>connect sys/sys@test99.net as sysdba
  SQL>alter system disable restricted session;
  5.测试stream的配置结果
  a 在test96上创建一个schema,并在该schema下创建一些对象,可以在test99上看到
  b 在test99上创建一个schema,并在该schema下创建一些对象,可以在test96上看到
  6 关于双向复制中avoid change cycling
  查看目标库apply进程的tag:
  COLUMN APPLY_NAME HEADING 'Apply Process Name' FORMAT A30
  COLUMN APPLY_TAG HEADING 'Tag Value' FORMAT A30
  SQL>connect sys/sys@test99 as sysdba;
  SELECT APPLY_NAME, APPLY_TAG FROM DBA_APPLY;
  Apply Process Name Tag Value
  ------------------------------ -----------
  APPLY$_TEST96_42 010781
  查看源库apply进程的tag:
  COLUMN APPLY_NAME HEADING 'Apply Process Name' FORMAT A30
  COLUMN APPLY_TAG HEADING 'Tag Value' FORMAT A30
  SQL>connect sys/sys@test99 as sysdba;
  SELECT APPLY_NAME, APPLY_TAG FROM DBA_APPLY;
  Apply Process Name Tag Value
  ------------------------------ ------------------------------
  APPLY$_TEST99_15 010498
  说明:消除多源复制中的递归问题,stream中已经有很好的消除机制,
  源端正常作业写入的redo entry的tag是NULL的,如果是由于源端的apply进程
  产生的redo entry,在redo entry中将带有tag标志,这样在源端捕获进程在
  捕获的redo entry中,过虑掉tag是NULL的,然后就可以消除change cycling.
  10gR2stream全库复制
  以上主要为大家介绍了pre_instantiation_setup/post_instantiation_setup过程在配置全库复制的方法,以下介绍dbms_streams_adm的maintain_global过程如何配置stream全库复制方法,适用于10gR2及以后版本。
  1l.在stream进行配置前,需要做些准备工作
  a 源库与目标库初始化参数的设置
  alter system set aq_tm_processes=4 scope=spfile;
  alter system set job_queue_processes=5 scope=spfile;
  alter system set global_names=true scope=spfile;
  alter system set streams_pool_size=51m scope=spfile;
  说明streams_pool_size在生产环境中最好>200m
  b 源库与目标库tnsnames.ora配置
  确保正确,可用tnsping通
  c 源库与目标库复制管理员的创建
  create user strmadmin identified by strmadminpw
  default tablespace &tbs_name quota unlimited on &tbs_name;
  grant connect, resource, dba to strmadmin;
  d 源库与目标库创建互连的数据链
  connect strmadmin/strmadminpw@test96;
  create database link test99.net connect to strmadmin
  identified by strmadminpw using 'test99';
  connect strmadmin/strmadminpw@test99;
  create database link test96.net connect to strmadmin
  identified by strmadminpw using 'test96';
  说明:必须确保双方的数据库链是可以连通.
  用pre_instantiation_setup/post_instantiation_setup过程时
  db link必须用db_name.domain的格式
  e 源库与目标库必须处于归档模式
  shutdown immediate;
  startup mount;
  alter database archivelog;
  alter database open;
  f 源库与目标库必须创建directory
  create directory dir_test96 as '/home/oracle/worksh';
  create directory dir_test99 as '/home/oracle/worksh';
  2.在源库执行MAINTAIN_GLOBAL过程
  SQL>connect strmadmin/strmadminpw@test96;
  begin
  dbms_streams_adm.maintain_global(
  source_directory_object => 'dir_test96',
  destination_directory_object => 'dir_test99',
  source_database => 'test96.net',
  destination_database => 'test99.net',
  perform_actions => true,
  include_ddl => true,
  instantiation => DBMS_STREAMS_ADM.INSTANTIATION_FULL_NETWORK);
  end;
  说明:在执行maintain_global时,源库与目标库必须创建directory,然后在源库执行, 目标库几乎什么都不用做,stream环境已经配置完毕。