环境:
sys@ORCL> select * from v$version where rownum=1; BANNER ---------------------------------------------------------------- Oracle Database 10g Enterprise Edition Release 10.2.0.1.0 - Prod sys@ORCL> !uname -a Linux localhost.localdomain 2.6.18-308.el5xen #1 SMP Fri Jan 27 17:59:00 EST 2012 i686 i686 i386 GNU/Linux
1 操作原理
RVWR每隔δt把数据块的before image flush到flashback log,形成合法的rowid集,我们简称之“基底”,基底是由RVWR动态向前维护,以使rowid集合法有效。在flashback database时,是先向后退一大步到基底,在向前迈一小步。借助flashback log只能将数据库恢复到某个指定时间点的状态,该状态可能并不是一致性的状态,然后需再运用archive log,去填补两个δt时间面之间的缺痕,使数据库恢复到一致性状态。一言以辟之,flashback database其实是把undo拿去再整理。其特性,用周杰伦的歌《回到过去》,“就回到过去,试着让故事继续.....”。
2 必备条件
① 归档模式
② 指定flash recovery area
flash recovery area设置的越大,flashback database的恢复能力就越强。由下面两个初始化参数控制:
1)db_recovery_file_dest:存储路径
2)db_recovery_file_dest_size:最大可用空间
sys@ORCL> show parameter db_recover NAME TYPE VALUE ------------------------------------ ----------- ------------------------------ db_recovery_file_dest string /u01/app/oracle/flash_recovery _area db_recovery_file_dest_size big integer 2G
③ 启动flashback database
启用该模式必须有两个先决条件:
1)归档模式
2)mount阶段
语法:
alter database flashback on/off;
验证是否开启:
select flashback_on from v$database
④ 启用force logging
force logging强制所有操作均产生redo entries
语法:
alter database force logging;
验证是否开启:
select force_logging from v$database
⑤ v$flashback_database_log
00:19:33 hr@ORCL (^ω^) desc v$flashback_database_log 名称 是否为空? 类型 ----------------------------------------- -------- ---------------------------- OLDEST_FLASHBACK_SCN NUMBER OLDEST_FLASHBACK_TIME DATE RETENTION_TARGET NUMBER FLASHBACK_SIZE NUMBER ESTIMATED_FLASHBACK_SIZE NUMBER 00:19:55 hr@ORCL (^ω^) select OLDEST_FLASHBACK_SCN from v$flashback_database_log; 未选定行 00:20:45 hr@ORCL (^ω^) select flashback_on from v$database; FLASHBACK_ON ------------------------------------ NO
3 制约因素
㈠ 控制文件重建
㈡ 数据文件shrunk
㈢ 表空间删除
㈣ 介质损坏
以上改变,其实质,会导致rowid在基底非法无效,便成了制约因素
4 额外补充
① db_flashback_retention_target
控制flashback log数据的保留时间,或者说,你希望flashback database能够恢复的最早时间点。缺省是一天。
sys@ORCL> show parameter db_flashback NAME TYPE VALUE ------------------------------------ ----------- ------------------------------ db_flashback_retention_target integer 1440
② 执行flashback database后,有两种选择:
a)alter database open resetlogs。一旦resetlogs,将不能再flashback至resetlogs之前的时间点,指定的scn之后产生的数据将全部丢失。
b)先alter database open read only;
然后expdp将误操作的表的数据导出
再执行recover database
最后impdp导入
③ 开启flashback on需要消费10%的资源,DBA需均衡利弊。
④ 介词to和until的区别:
例子:
to 20 :包括了20
until 20 :没有包括20,也就是19末。
5 测试实验
㈠ 检查必备条件
sys@ORCL> show parameter db_recover NAME TYPE VALUE ------------------------------------ ----------- ------------------------------ db_recovery_file_dest string /u01/app/oracle/flash_recovery _area db_recovery_file_dest_size big integer 2G sys@ORCL> archive log list Database log mode Archive Mode Automatic archival Enabled Archive destination USE_DB_RECOVERY_FILE_DEST Oldest online log sequence 13 Next log sequence to archive 15 Current log sequence 15 sys@ORCL> select force_logging,flashback_on from v$database; FOR FLASHBACK_ON --- ------------------ YES YES
㈡ 测试
hr@ORCL> select dbms_flashback.get_system_change_number from dual; GET_SYSTEM_CHANGE_NUMBER ------------------------ 731867 hr@ORCL> truncate table t2; Table truncated. hr@ORCL> conn / as sysdba Connected. sys@ORCL> shutdown immediate Database closed. Database dismounted. ORACLE instance shut down. sys@ORCL> startup mount ORACLE instance started. Total System Global Area 419430400 bytes Fixed Size 1219760 bytes Variable Size 83886928 bytes Database Buffers 331350016 bytes Redo Buffers 2973696 bytes Database mounted. sys@ORCL> flashback database to scn 731867; Flashback complete. sys@ORCL> alter database open resetlogs; Database altered. sys@ORCL> select count(*) from hr.t2; COUNT(*) ---------- 19