Oracle DG(Dataguard)是目前比较常见的数据库HA配置策略。通过实现PhysicalStandby和Logical Standby，可以实现数据冗余容错机制。防止在主库出现严重故障，不能支持服务的时候，没有快速的后备支持环境。
在DG中，switchover和failover是两个重要的概念，也是DG实现的核心。两者共同点都是Primary和Standby角色切 换，差异在于Planned和UnPlanned之分。Switchover关键点在于Planned，这个切换动作是在运维机构规划范围内的动作。比 如，进行定期系统软硬件升级、设备维修等动作。而Failover是真正出现严重系统故障，如数据库宕机、软硬件故障导致的Primary不能支持服务，从而进行的切换动作。
根据不同的DG配置，switchover和failover也是有差异的。理论上，Switchover是不会造成数据丢失的，Primary在 切换之后也是在DG配置环境中，作为Standby存在的。但是Failover则不同，除了运行在最大保护(Maximum Protection)模式下，Primary突发的故障可能引起一部分Redo Log不能及时的传递到Standby端，切换之后很可能有数据损失的情况。更重要的是，Primary端在发生Failover之后，是不能够直接加入 回DG配置的!也就是说，Failover之后，Primary实际上就是被“抛出”了DG环境。
那么，有什么方法实现Primary回到原有的环境呢?这个问题的困难在于保持Primary和Standby一致。在正常情况下，Primary 和Standby之间是关联同步的，即使发生了Switchover，也在可控情况下。Failover过程中有数据的缺失，还有Primary修复问 题。在目前流行版本(11g)中，有三个方法：
环境重建：一种最简单的方法就是直接删除原来的Primary库，引用DG重建方法，重新搭建Standby端;
RMAN备份恢复：如果Primary端保留过一份Failover之前的备份，则可以强制原来的Primary端恢复到进行Failover的时间点，之后作为Standby接收当前Primary的redo log传递，应用后可以跟上进度;
Flashback Database恢复：Flashback技术是作为传统备份还原技术的补充，提供了更加便捷的恢复策略。使用flashback，可以将数据库恢复到failover之前的时间点。之后的过程和RMAN备份恢复策略相同;
案例分析：
一、在主库端模拟数据库意外宕机
1. 7scott@bjdb>conn /as sysdba
2. Connected.
3. sys@bjdb>alter system switch logfile;
4. System altered.
5. sys@bjdb>shutdown abort
6. ORACLE instance shut down.
二、在备库端
1、查看切换信息
1. 5sys@shdb>select name,database_role,switchover_status from v$database;
2. NAME DATABASE_ROLE SWITCHOVER_STATUS
3. --------- ---------------- --------------------
4. TESTDB12 PHYSICAL STANDBY NOT ALLOWED
5. 可以看到此时备库处于无法切换状态
2、直接切换
1. sys@shdb>alter database commit to switchover to primary;
2. alert_log:(告警日志)
3. Fatal NI connect error 12514, connecting to:
4. (DESCRIPTION=(ADDRESS_LIST=(ADDRESS=(PROTOCOL=TCP)(HOST=shsrv)(PORT=1521)))(CONNECT_DATA=(SERVICE_NAME=shdb)(CID=(PROGRAM=oracle)(HOST=bjsrv)(USER=oracle))))
5. VERSION INFORMATION:
6. TNS for Linux: Version 11.2.0.3.0 - Production
7. TCP/IP NT Protocol Adapter for Linux: Version 11.2.0.3.0 - Production
8. Time: 04-MAR-2015 21:25:13
9. Tracing not turned on.
10.Tns error struct:
11.ns main err code: 12564
12.TNS-12564: TNS:connection refused
13.ns secondary err code: 0
14.nt main err code: 0
15.nt secondary err code: 0
16.nt OS err code: 0
17.Error 12514 received logging on to the standby
18.FAL[client, MRP0]: Error 12514 connecting to shdb for fetching gap sequence
19.Wed Mar 04 21:26:00 2015
20.alter database commit to switchover to primary
21.ALTER DATABASE SWITCHOVER TO PRIMARY (TestDB12)
22.Maximum wait for role transition is 15 minutes.
23.Switchover: Media recovery is still active
24.Database not available for switchover
25.End-Of-REDO archived log file has not been recovered
26.Database not available for switchover
27.End-Of-REDO archived log file has not been recovered
28.Database not available for switchover
3、关闭standby MPR进程
1. 35sys@shdb>ALTER DATABASE RECOVER managed standby database finish;
2. ALTER DATABASE RECOVER managed standby database finish
3. Terminal Recovery: request posted (TestDB12)
4. Wed Mar 04 21:34:34 2015
5. Begin: Standby Redo Logfile archival
6. End: Standby Redo Logfile archival
7. Terminal Recovery timestamp is '03/04/2015 21:34:34'
8. Terminal Recovery: applying standby redo logs.
9. Terminal Recovery: thread 1 seq# 34 redo required
10.Media Recovery Waiting for thread 1 sequence 34
11.Terminal Recovery: End-Of-Redo log allocation
12.Terminal Recovery: standby redo logfile 4 created '/dsk4/arch_bj/arch_1_0_820054583.log'
13.This standby redo logfile is being created as part of the
14.failover operation. This standby redo logfile should be
15.deleted after the switchover to primary operation completes.
16.Media Recovery Log /dsk4/arch_bj/arch_1_0_820054583.log
17.Terminal Recovery: log 4 reserved for thread 1 sequence 34
18.Recovery of Online Redo Log: Thread 1 Group 4 Seq 34 Reading mem 0
19.Mem# 0: /dsk4/arch_bj/arch_1_0_820054583.log
20.Identified End-Of-Redo (failover) for thread 1 sequence 34 at SCN 0xffff.ffffffff
21.Incomplete Recovery applied until change 1234252 time 03/04/2015 21:23:43
22.MRP0: Media Recovery Complete (TestDB12)
23.Terminal Recovery: successful completion
24.Wed Mar 04 21:34:35 2015
25.ARCH: Archival stopped, error occurred. Will continue retrying
26.ORACLE Instance TestDB12 - Archival Error
27.ORA-16014: log 4 sequence# 34 not archived, no available destinations
28.ORA-00312: online log 4 thread 1: '/dsk4/arch_bj/arch_1_0_820054583.log'
29.Forcing ARSCN to IRSCN for TR 0:1234252
30.Attempt to set limbo arscn 0:1234252 irscn 0:1234252
31.Resetting standby activation ID 2865247982 (0xaac836ee)
32.MRP0: Background Media Recovery process shutdown (TestDB12)
33.Terminal Recovery: completion detected (TestDB12)
34.Completed: ALTER DATABASE RECOVER managed standby database finish
4、切换数据库到Primary
1. sys@shdb>select status from v$instance;
2. STATUS
3. ------------
4. OPEN
5. sys@shdb>select name,database_role,switchover_status from v$database;
6. NAME DATABASE_ROLE SWITCHOVER_STATUS
7. --------- ---------------- --------------------
8. TESTDB12 PHYSICAL STANDBY TO PRIMARY
9. sys@shdb>alter database commit to switchover to primary;
10.Database altered.
11.sys@shdb>alter database open;
12.Database altered.
13.告警日志：
14.alter database commit to switchover to primary
15.ALTER DATABASE SWITCHOVER TO PRIMARY (TestDB12)
16.Maximum wait for role transition is 15 minutes.
17.All dispatchers and shared servers shutdown
18.CLOSE: killing server sessions.
19.CLOSE: all sessions shutdown successfully.
20.Wed Mar 04 21:35:47 2015
21.SMON: disabling cache recovery
22.Backup controlfile written to trace file /u01/app/oracle/diag/rdbms/bjdb/TestDB12/trace/TestDB12_ora_3146.trc
23.Standby terminal recovery start SCN: 1234251
24.RESETLOGS after incomplete recovery UNTIL CHANGE 1234252
25.Online log /dsk2/oradata/bjdb/redo01b.log: Thread 1 Group 1 was previously cleared
26.Online log /dsk1/oradata/bjdb/redo01a.log: Thread 1 Group 1 was previously cleared
27.Online log /dsk2/oradata/bjdb/redo02b.log: Thread 1 Group 2 was previously cleared
28.Online log /dsk1/oradata/bjdb/redo02a.log: Thread 1 Group 2 was previously cleared
29.Online log /dsk2/oradata/bjdb/redo03b.log: Thread 1 Group 3 was previously cleared
30.Online log /dsk1/oradata/bjdb/redo03a.log: Thread 1 Group 3 was previously cleared
31.Standby became primary SCN: 1234250
32.Wed Mar 04 21:35:47 2015
33.Setting recovery target incarnation to 3
34.AUDIT_TRAIL initialization parameter is changed back to its original value as specified in the parameter file.
35.Switchover: Complete - Database mounted as primary
36.Completed: alter database commit to switchover to primary
三、原主库修复后，开机
1. sys@bjdb>startup
2. ORACLE instance started.
3. Total System Global Area 442601472 bytes
4. Fixed Size 2229184 bytes
5. Variable Size 281021504 bytes
6. Database Buffers 155189248 bytes
7. Redo Buffers 4161536 bytes
8. Database mounted.
9. Database opened.
10.sys@bjdb>select name,database_role,switchover_status from v$database;
11.NAME DATABASE_ROLE SWITCHOVER_STATUS
12.--------- ---------------- --------------------
13.TESTDB12 PRIMARY FAILED DESTINATION
现在原来的主库被修复后，整个DataGuara架构已经被破坏了，所以必须把原来的主库构建成新的备库，重新恢复DataGuard的环境。
四、重新构建DataGuard
1. 1sys@bjdb>select name,database_role from v$database;
NAME DATABASE_ROLE
-------------------------------------------------- ----------------
TESTDB12 PHYSICAL STANDBY
Oready即为“One Technology Ready”！
Oready不仅仅是数据库领域的专家，同时也是一个社区，我们汇集百位数据库领域资深顾问、讲师，服务于数据中心的技术支持与技能传授，注重对客户的关键数据库运营支撑，强调经验的分享-Skill Transfer，聚焦数据中心的完美结合.
