共计 3942 个字符，预计需要花费 10 分钟才能阅读完成。
最近有几项业务下线，需要从一张表中删除 6.8 亿多条数据。想办法把数据删除掉了，但对应的 ogg 灾备端复制时有了的延迟，而且延迟的时间起来越长。
对于表太多造成的延迟可以把所有表分为多个组来做复制，于是想复制进程是否可以对单表复制开并行。上网查到了相关的资料，可以使用 @RANGE 函数对单表作表内的拆分，通过对表上主键列作 hash 算法将该表上发生的变更均分到多个 replicat 上来降低单个 replicat 组件的负载。
动手实验一下：
ogg 搭建过程不再重复，从网上就可以查到。
实验过程：模拟在源端对表 scott.emp1 做大量的 dml 操作，复制进程出现延迟，在目标端对复制表 scott.emp1 开并行 3 个进程。
源端插入数据：
SQL> insert into scott.emp1 select * from scott.emp; 14 rows created. SQL> commit; Commit complete. SQL> insert into scott.emp1 select * from scott.emp1; 14 rows created. SQL> / 28 rows created. SQL> / ....... SQL> / 1835008 rows created. SQL> commit; Commit complete. SQL> select count(*) from scott.emp1; COUNT(*) ---------- 3670016目标端有延迟
GGSCI (rhel5) 15> info all Program Status Group Lag Time Since Chkpt MANAGER RUNNING REPLICAT RUNNING REPTAB 00:09:08 00:00:04停掉复制进程
拆分复制进程，对表 scott.emp1 分三个进程复制
# 源复制进程
GGSCI (rhel5) 23> view params reptab replicat reptab SETENV (NLS_LANG="AMERICAN_AMERICA.ZHS16GBK") SETENV (Oracle_SID="orcl") userid ogg,password 123456 reperror default,discard assumetargetdefs discardfile /goldengate/dirrpt/reptab.dsc,append,megabytes 1024 gettruncates dynamicresolution map scott.emp1, target scott.emp1 ; map scott.emp, target scott.emp ;源进程修改为
map scott.emp1, target scott.emp1 ,FILTER(@RANGE(1,3));
多复制出两个参数文件：
GGSCI (rhel5) 1> view params reptab02 replicat reptab02 SETENV (NLS_LANG="AMERICAN_AMERICA.ZHS16GBK") SETENV (ORACLE_SID="orcl") userid ogg,password 123456 reperror default,discard assumetargetdefs discardfile /goldengate/dirrpt/reptab.dsc,append,megabytes 1024 gettruncates dynamicresolution map scott.emp1, target scott.emp1 ,FILTER (@RANGE(2,3)); GGSCI (rhel5) 2> view params reptab03 replicat reptab03 SETENV (NLS_LANG="AMERICAN_AMERICA.ZHS16GBK") SETENV (ORACLE_SID="orcl") userid ogg,password 123456 reperror default,discard assumetargetdefs discardfile /goldengate/dirrpt/reptab.dsc,append,megabytes 1024 gettruncates dynamicresolution map scott.emp1, target scott.emp1 ,FILTER (@RANGE(3,3));添加两个复制进程，extseqno 和 extrba 与源进程一致
GGSCI (rhel5) 9> info reptab REPLICAT REPTAB Last Started 2017-05-05 16:18 Status ABENDED Checkpoint Lag 00:09:08 (updated 00:09:38 ago) Log Read Checkpoint File ./dirdat/tl000003 2017-05-05 16:09:11.000187 RBA 194186157 GGSCI (rhel5) 10> add replicat reptab02, exttrail ./dirdat/tl,extseqno 3 extrba 194186157,checkpointtable ogg.checkpointREPLICAT added. GGSCI (rhel5) 11> add replicat reptab03, exttrail ./dirdat/tl,extseqno 3 extrba 194186157,checkpointtable ogg.checkpointREPLICAT added.启动复制进程
GGSCI (rhel5) 12> start reptab* Sending START request to MANAGER ... REPLICAT REPTAB starting Sending START request to MANAGER ... REPLICAT REPTAB02 starting Sending START request to MANAGER ... REPLICAT REPTAB03 starting查看数据库里 ogg 对应的会话
SQL> select module,sql_id from v$session where username='OGG'; MODULE SQL_ID ------------------------------------------------------------------------------------------------------------------------------------------------ --------------------------------------- OGG-REPTAB03-OPEN_DATA_SOURCE 1cxrusnmn01hz OGG-REPTAB-OPEN_DATA_SOURCE 1cxrusnmn01hz OGG-REPTAB02-OPEN_DATA_SOURCE 1cxrusnmn01hz SQL> select sql_text from v$sqlarea where sql_id='1cxrusnmn01hz'; SQL_TEXT -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- INSERT INTO "SCOTT"."EMP1" ("EMPNO","ENAME","JOB","MGR","HIREDATE","SAL","COMM","DEPTNO") VALUES (:a0,:a1,:a2,:a3,:a4,:a5,:a6,:a7)可以看到出现了三个会话，都是对应的对表 scott.emp1 的插入语句。也就是说实现了对 scott.emp1 表的并行复制。
MOS 上也有相关的文档介绍相应的功能，文档：1320133.1 和 1512633.1
更多 Oracle 相关信息见 Oracle 专题页面 http://www.linuxidc.com/topicnews.aspx?tid=12
本文永久更新链接地址 ：http://www.linuxidc.com/Linux/2017-05/143493.htm
