【数据泵迁移】- expdp/impdp
一、数据泵迁移优缺点
-
1.1 优点
- 对数据库版本和平台无限制
- 支持开启并行（参数parallel）
- 可以直接从一个数据库到另一个数据库进行数据传输，无需将数据导出到中间文件
- 可以选择导出或导入特定的表、用户等，也可以指定导出或导入数据的具体部分
- 支持压缩功能（参数compress），可以减少存储空间和传输时间
-
1.2 缺点
- 不同版本的 Oracle 数据库之间可能会存在兼容性问题，特别是在源数据库和目标数据库版本差异较大的情况下
- 导出数据量大时，可能会对源数据库性能产生影响，尤其是当并行处理设置不当时
- 迁移前需要做好详细的规划和准备，迁移后需要校验数据
二、生产库信息检查及确认
2.1 确认要迁移的用户
SQL> SELECT USERNAME,CREATED FROM DBA_USERS WHERE ACCOUNT_STATUS='OPEN' AND USERNAME NOT IN ('SYS','SYSTEM') AND LOCK_DATE IS NULL ORDER BY 2;
2.2 查看字符集，表空间
字符集：
SQL> COL PARAMETER FOR a30
SQL> COL VALUE FOR a30
SQL> SELECT * FROM V$NLS_PARAMETERS WHERE PARAMETER LIKE 'NLS_%CHARACTERSET';
表空间：
当前表空间大小：
SQL> SELECT TS.TABLESPACE_NAME,TS.BIGFILE,SUM(DF.BYTES)/1024/1024 "M" FROM DBA_TABLESPACES TS,DBA_DATA_FILES DF WHERE TS.TABLESPACE_NAME=DF.TABLESPACE_NAME GROUP BY TS.TABLESPACE_NAME,TS.BIGFILE ORDER BY 1;
迁移用户使用的表空间：
SQL> SELECT ''''||TABLESPACE_NAME||'''' FROM (
SELECT LISTAGG(TABLESPACE_NAME,''',''') AS TABLESPACE_NAME FROM (
SELECT DISTINCT TABLESPACE_NAME FROM DBA_TABLES WHERE OWNER IN ('USERNAME1','USERNAME2')
UNION ALL
SELECT DISTINCT TABLESPACE_NAME FROM DBA_TAB_PARTITIONS WHERE TABLE_OWNER IN ('USERNAME1','USERNAME2')
UNION ALL
SELECT DISTINCT TABLESPACE_NAME FROM DBA_TAB_SUBPARTITIONS WHERE TABLE_OWNER IN ('USERNAME1','USERNAME2')
UNION ALL
SELECT DISTINCT TABLESPACE_NAME FROM DBA_INDEXES WHERE OWNER IN ('USERNAME1','USERNAME2')
UNION ALL
SELECT DISTINCT TABLESPACE_NAME FROM DBA_IND_PARTITIONS WHERE INDEX_OWNER IN ('USERNAME1','USERNAME2')
UNION ALL
SELECT DISTINCT TABLESPACE_NAME FROM DBA_IND_SUBPARTITIONS WHERE INDEX_OWNER IN ('USERNAME1','USERNAME2')));
迁移用户使用默认表空间：
SQL> SELECT DEFAULT_TABLESPACE FROM DBA_USERS WHERE USERNAME IN ('USERNAME1','USERNAME2');
迁移用户使用非默认临时表空间：
SQL> SELECT DISTINCT TEMPORARY_TABLESPACE FROM DBA_USERS WHERE USERNAME IN ('USERNAME1','USERNAME2') AND TEMPORARY_TABLESPACE!='TEMP';
生产端生产创建表空间SQL，目标端执行
(1)先创建表空间，小于30G，按照实际大小，大于30G按照30G创建，下面继续添加
SQL> SELECT 'CREATE TABLESPACE '|| T.TABLESPACE_NAME || ' DATAFILE SIZE '|| (CASE WHEN SUM(F.BYTES)/1024/1024 < 30720 THEN SUM(F.BYTES)/1024/1024 ELSE 30720 END)||'M AUTOEXTEND ON MAXSIZE 30G;' FROM DBA_TABLESPACES T , DBA_DATA_FILES F
WHERE T.CONTENTS <> 'UNDO' AND T.CONTENTS <> 'TEMPORARY' AND T.TABLESPACE_NAME NOT IN ('SYSTEM','SYSAUX','TEMP','USERS')
AND T.TABLESPACE_NAME = F.TABLESPACE_NAME
GROUP BY T.TABLESPACE_NAME;
(2)根据生产表空间大小，部分大于30G表空间继续添加数据文件至与生产相当大小
SQL> SELECT 'ALTER TABLESPACE '|| S.TABLESPACE_NAME || ' ADD DATAFILE SIZE 30G;' FROM (SELECT T.TABLESPACE_NAME,SUM(F.BYTES)/1024/1024 DBFSIZE FROM DBA_TABLESPACES T,DBA_DATA_FILES F WHERE T.TABLESPACE_NAME = F.TABLESPACE_NAME AND T.CONTENTS <> 'UNDO' AND T.CONTENTS <> 'TEMPORARY' AND T.TABLESPACE_NAME NOT IN ('SYSTEM','SYSAUX','TEMP') GROUP BY T.TABLESPACE_NAME) S
WHERE S.DBFSIZE > 32670;
(3)创建临时表空间
SQL> SELECT 'CREATE TEMPORARY TABLESPACE '|| T.TABLESPACE_NAME || ' TEMPFILE SIZE 30G;' FROM DBA_TEMP_FILES T WHERE T.TABLESPACE_NAME != 'TEMP';
2.3 查看profile，role，回收站
查看profile：
SET LINES 200 PAGES 2000
COL PROFILE FOR a20
COL RESOURCE_NAME FOR a30
COL LIMIT FOR a20
SELECT * FROM DBA_PROFILES
WHERE PROFILE IN (SELECT PROFILE FROM DBA_USERS
WHERE USERNAME IN ('USERNAME1','USERNAME2')) ORDER BY 1,2,3;
目标端创建对应的PROFILE语句
SET LINES 200 PAGES 2000
SELECT DBMS_METADATA.GET_DDL('PROFILE','PROFILE_NAME') FROM DUAL;
确认role：
SET LINES 200 PAGES 2000
COL GRANTEE FOR a20
COL GRANTED_ROLE FOR a20
SELECT * FROM DBA_ROLE_PRIVS WHERE GRANTEE IN ('USERNAME1','USERNAME2');
目标端创建role SQL：
SET LINES 200 PAGES 2000
SELECT DBMS_METADATA.GET_DDL('ROLE',ROLE_NAME) FROM DUAL;
--授权语句
SELECT * FROM ROLE_SYS_PRIVS WHERE ROLE='ROLE1,ROLE2'
UNION ALL
SELECT * FROM ROLE_ROLE_PRIVS WHERE ROLE='ROLE1,ROLE2'
UNION ALL
SELECT * FROM ROLE_ROLE_PRIVS WHERE ROLE='ROLE1,ROLE2';
查看回收站：
SET LINES 200 PAGES 2000
SELECT OBJECT_NAME FROM DBA_RECYCLEBIN WHERE OWNER IN ('USERNAME1','USERNAME2');
清理回收站：
PURGE RECYCLEBIN;
2.4 对象/INVALID对象数量
SET LINES 200 PAGES 2000
COL OWNER FOR a20
COL OBJECT_TYPE FOR a30
SELECT OWNER,OBJECT_TYPE,COUNT(*) TOTAL,SUM(CASE WHEN STATUS<>'VALID' THEN 1 ELSE 0 END) INVALID FROM DBA_OBJECTS WHERE OWNER IN ('USERNAME1','USERNAME2') GROUP BY OWNER,OBJECT_TYPE ORDER BY 1,2;
2.5 检查JOB，DBLINK
确认迁移开始前，JOB是否会执行，记录当前JOB状态
SET LINES 200
COL JOB FOR A20
COL WHAT FOR A50
COL NEXT_DATE FOR A20
COL NEXT_SEC FOR A20
SELECT JOB,WHAT,NEXT_DATE,NEXT_SEC,INTERVAL,BROKEN FROM DBA_JOBS WHERE LOG_USER IN ('USERNAME1','USERNAME2');
--SCHEDULER
SET LINES 200
COL JOB_NAME FOR A30
COL PROGRAM_OWNER FOR A10
COL PROGRAM_NAME FOR A30
SELECT JOB_NAME,PROGRAM_OWNER,PROGRAM_NAME,NEXT_RUN_DATE FROM DBA_SCHEDULER_JOBS WHERE OWNER IN ('USERNAME1','USERNAME2');
--CRONTAB(ROOT,GRID,ORACLE)
# CRONTAB -L
检查DBLINK
--业务用户
SET LINES 200 PAGES 2000
SELECT DB_LINK,USERNAME,HOST FROM DBA_DB_LINKS WHERE OWNER IN ('USERNAME1','USERNAME2');
--PUBLIC
SET LINES 200 PAGES 2000
COL DB_LINK FOR a10
COL USERNAME, FOR a10
COL HOST FOR a50
SELECT DB_LINK,USERNAME,HOST FROM DBA_DB_LINKS WHERE OWNER='PUBLIC';
生产端生成PUBLIC DBLINK语句，目标端执行
SET LINES 200 PAGES 2000
SELECT DBMS_METADATA.GET_DDL('DB_LINK','DBLINK_NAME,'PUBLIC') FROM DUAL;
2.6 检查外部表，SQL Profile
检查外部表：
SET LINES 200 PAGES 2000
SELECT DISTINCT DEFAULT_DIRECTORY_NAME FROM DBA_EXTERNAL_TABLES WHERE OWNER IN ('USERNAME1','USERNAME2');
检查SQL PROFILE：
SET LINES 200 PAGES 2000
SELECT NAME,CATEGORY,SQL_TEXT,STATUS FROM DBA_SQL_PROFILES ORDER BY 1;
2.7 查看目录
SET LINES 200 PAGES 2000
COL DIRECTORY_NAME FOR a30
COL DIRECTORY_PATH FOR a100
SELECT DIRECTORY_NAME,DIRECTORY_PATH FROM DBA_DIRECTORIES;
三、配置目标端数据库
3.1 创建步骤二用户，表空间，对象等
3.2 根据实际情况配置nfs
四、正式迁移
4.1 生产端停业务，数据库关闭监听，kill所有连接会话
停监听：
- 单实例：lsnrctl stop listener_name
- RAC:srvctl stop listener -l listener_name
srvctl stop scan_listener
kill所有连接会话：
ps -ef|grep LOCAL=NO|grep -v grep|grep DBinstance_sid|awk '{print $2}'|xargs kill -9
4.2 生产端导出数据
expdp \'/ as sysdba\' DIRECTORY=expdp SCHEMAS=USERNAME1,USERNAME2 DUMPFILE=expdp_%U.DMP LOGFILE=EXPDP.LOG PARALLEL=8 CLUSTER=N EXCLUDE=STATISTICS
4.3 目标端导入数据
impdp \'/ as sysdba\' DIRECTORY=expdp DUMPFILE=impdp_%U.DMP LOGFILE=IMPDP.LOG PARALLEL=8 CLUSTER=N
4.4 两端数据验证核对
核对profile：
SET LINES200 PAGES 2000
COL PROFILE FOR A40
COL RESOURCE_NAME FOR A30
COL LIMIT FOR A20
SELECT * FROM DBA_PROFILES
WHERE PROFILE IN (SELECT PROFILE FROM DBA_USERS
WHERE USERNAME IN ('USERNAME1','USERNAME2')) ORDER BY 1,2,3;
核对role：
SELECT * FROM DBA_ROLE_PRIVS WHERE GRANTEE IN ('USERNAME1','USERNAME2');
核对job：
--JOB
SET LINES200 PAGES 2000
SELECT JOB,WHAT,BROKEN FROM DBA_JOBS WHERE LOG_USER IN ('USERNAME1','USERNAME2');
--SCHEDULER
SET LINES200 PAGES 2000
SELECT JOB_NAME,PROGRAM_OWNER,PROGRAM_NAME FROM DBA_SCHEDULER_JOBS WHERE OWNER IN ('USERNAME1','USERNAME2');
--CRONTAB(ROOT,GRID,ORACLE)
crontab -l
核对外部表：
SELECT DISTINCT DEFAULT_DIRECTORY_NAME FROM DBA_EXTERNAL_TABLES WHERE OWNER IN ('USERNAME1','USERNAME2');
核对DBLINK：
--业务用户
SET LINES 200 PAGES 2000
SELECT DB_LINK,USERNAME,HOST FROM DBA_DB_LINKS WHERE OWNER IN ('USERNAME1','USERNAME2');
--PUBLIC
SET LINES 200 PAGES 2000
COL DB_LINK FOR a10
COL USERNAME, FOR a10
COL HOST FOR a50
SELECT DB_LINK,USERNAME,HOST FROM DBA_DB_LINKS WHERE OWNER='PUBLIC';
4.5 编译无效对象
SQL> @?/rdbms/admin/utlrp.sql
两端无效对象对比：
SET LINES200 PAGES 2000
COL OWNER FOR A20
COL OBJECT_TYPE FOR A30
SELECT OWNER,OBJECT_TYPE,COUNT(*) TOTAL,SUM(CASE WHEN STATUS<>'VALID' THEN 1 ELSE 0 END) INVALID FROM DBA_OBJECTS WHERE OWNER IN ('USERNAME1','USERNAME2') GROUP BY OWNER,OBJECT_TYPE ORDER BY 1,2;
4.6 目标端收集统计信息
收集数据字典统计信息：
SQL> EXEC DBMS_STATS.GATHER_DICTIONARY_STATS(DEGREE=>4);
SQL> EXEC DBMS_STATS.GATHER_FIXED_OBJECTS_STATS;
收集全库统计信息：
SQL> begin
dbms_stats.gather_database_stats(
ESTIMATE_PERCENT=>DBMS_STATS.AUTO_SAMPLE_SIZE, method_opt => 'for all columns size auto', cascade=>true, degree=>72);
end;
/
4.7 应用连接新库验证
五、迁移回退
回退步骤目标端关闭当前数据泵导入工作，原生产端开启应用和数据库监听即可。