在Oracle下我们如何正确的执行数据库恢复
xinjingsihai 2014-02-21
当数据库需要进行介质恢复时，为了确保数据库能够顺利的执行恢复过程，恢复数据库到当前状态。我们要做的就是验证！验证什么呢？当然是验证备份集和归档是否能够进行有效的恢复。防止我们restore后，执行recover时却发现归档缺少了一堆，顿时傻眼。
比方说，在数据库当前日志序列号为3时我们完全备份了数据库。在数据库当前联机日志序列号为13时数据库损坏需要恢复。假设数据库联机日志组为3组，则可以推断数据库联机日志序列号分别为11、12、13。因此当数据库执行restore database后，再执行recover时不难推断数据库需要应用归档3、4、5、6、7、8、9、10以及联机日志11、12、13来进行完全恢复。
为了能够顺利的执行完全恢复，我们在执行恢复前，需要对restore调用的备份集进行恢复验证（语句为：restorevalidate database）以及验证recover过程所需的归档3-10（语句为：restore validate archivelog sequence between 3 and10）。
以完全恢复为例，举例如下：
1数据库当前日志seq号为59，我们备份数据库
SQL> selectgroup#,archived,sequence#,status from v$log;
GROUP# ARC SEQUENCE# STATUS
---------- --- ---------- ----------------
1 YES 58 INACTIVE
2 NO 59 CURRENT
3 YES 57 INACTIVE
RMAN> backup database format'/backup/fullbk-%T-%U.bak';
Starting backup at 2014-02-17 12:03:28
using channel ORA_DISK_1
channel ORA_DISK_1: starting full datafilebackup set
channel ORA_DISK_1: specifying datafile(s)in backup set
input datafile file number=00004name=/Oracle/CRM/CRM/users01.dbf
input datafile file number=00001name=/oracle/CRM/CRM/system01.dbf
input datafile file number=00002name=/oracle/CRM/CRM/sysaux01.dbf
input datafile file number=00003name=/oracle/CRM/CRM/undotbs01.dbf
input datafile file number=00005name=/oracle/CRM/CRM/crm.dbf
input datafile file number=00006name=/oracle/CRM/test.dbf
input datafile file number=00008name=/oracle/CRM/jxc.dbf
input datafile file number=00007name=/oracle/CRM/user01.dbf
channel ORA_DISK_1: starting piece 1 at2014-02-17 12:03:29
channel ORA_DISK_1: finished piece 1 at2014-02-17 12:05:57
piecehandle=/backup/fullbk-20140217-3ep0rj0h_1_1.bak tag=TAG20140217T120328 comment=NONE
channel ORA_DISK_1: backup set complete,elapsed time: 00:02:28
channel ORA_DISK_1: starting full datafilebackup set
channel ORA_DISK_1: specifying datafile(s)in backup set
including current control file in backupset
including current SPFILE in backup set
channel ORA_DISK_1: starting piece 1 at2014-02-17 12:06:01
channel ORA_DISK_1: finished piece 1 at2014-02-17 12:06:02
piecehandle=/backup/fullbk-20140217-3fp0rj56_1_1.bak tag=TAG20140217T120328comment=NONE
channel ORA_DISK_1: backup set complete,elapsed time: 00:00:01
Finished backup at 2014-02-17 12:06:02
2 当数据库联机日志为69时数据库崩溃需要进行介质恢复
SQL> selectgroup#,archived,sequence#,status from v$Log;
GROUP# ARC SEQUENCE# STATUS
---------- --- ---------- ----------------
1 YES 67 INACTIVE
2 YES 68 INACTIVE
3 NO 69 CURRENT
注意：这里其实我们可以推断，如果数据库需要恢复到当前状态，那么归档59到归档66的所有归档，必须能够进行有效的恢复。我们只需要发起restore database preview命令，Oracle便可以给出我们归档列表，继续往下看。
3 判定当前数据库恢复所需要备份集和归档条目
注意对于restore database preview列出的归档条目，recover执行完全恢复时并不会完全应用，因为完全恢复recover过程是：应用相关归档+ 所有联机日志，seq号从小到大依次应用。后面会抓取recover过程，这里先暂且提一下。
RMAN> restore database preview;
Starting restore at 2014-02-17 16:14:21
using target database control file insteadof recovery catalog
allocated channel: ORA_DISK_1
channel ORA_DISK_1: SID=63 device type=DISK
List of Backup Sets
===================
BS Key Type LV Size Device TypeElapsed Time Completion Time
------- ---- -- ---------- ----------------------- -------------------
108 Full 2.03G DISK 00:02:26 2014-02-17 12:05:38
BP Key: 108 Status:AVAILABLE Compressed: NO Tag: TAG20140217T120328
Piece Name:/backup/fullbk-20140217-3ep0rj0h_1_1.bak
注意：这里显示备份片总是rman资料库中记录的数据文件最新的备份
List of Datafiles in backup set 108
File LV Type Ckp SCN CkpTime Name
---- -- ---- ---------- ------------------- ----
1 Full 4028039 2014-02-17 12:03:29/oracle/CRM/CRM/system01.dbf
2 Full 4028039 2014-02-17 12:03:29/oracle/CRM/CRM/sysaux01.dbf
3 Full 4028039 2014-02-17 12:03:29/oracle/CRM/CRM/undotbs01.dbf
4 Full 4028039 2014-02-17 12:03:29/oracle/CRM/CRM/users01.dbf
5 Full 4028039 2014-02-17 12:03:29 /oracle/CRM/CRM/crm.dbf
6 Full 4028039 2014-02-17 12:03:29 /oracle/CRM/test.dbf
7 Full 4028039 2014-02-17 12:03:29 /oracle/CRM/user01.dbf
8 Full 4028039 2014-02-17 12:03:29 /oracle/CRM/jxc.dbf
using channel ORA_DISK_1
List of Archived Log Copies for databasewith db_unique_name CRM
=====================================================================
Key Thrd Seq S Low Time
------- ---- ------- - -------------------
131 1 59 A 2014-02-17 11:55:37
Name:/oracle/archivelog/arch_1_59_839098938.arch
132 1 60 A 2014-02-17 12:10:20
Name:/oracle/archivelog/arch_1_60_839098938.arch
133 1 61 A 2014-02-17 12:10:21
Name:/oracle/archivelog/arch_1_61_839098938.arch
134 1 62 A 2014-02-17 12:10:26
Name:/oracle/archivelog/arch_1_62_839098938.arch
135 1 63 A 2014-02-17 12:10:30
Name:/oracle/archivelog/arch_1_63_839098938.arch
136 1 64 A 2014-02-17 12:10:31
Name:/oracle/archivelog/arch_1_64_839098938.arch
137 1 65 A 2014-02-17 12:10:32
Name:/oracle/archivelog/arch_1_65_839098938.arch
138 1 66 A 2014-02-17 12:10:33
Name:/oracle/archivelog/arch_1_66_839098938.arch
139 1 67 A 2014-02-17 12:10:34
Name:/oracle/archivelog/arch_1_67_839098938.arch
140 1 68 A 2014-02-17 12:10:36
Name:/oracle/archivelog/arch_1_68_839098938.arch
Media recovery start SCN is 4028039
Recovery must be done beyond SCN 4028039 toclear datafile fuzziness
Finished restore at 2014-02-17 16:14:24
注意：
1 上面seq号这一列显示的最后一个归档seq为68（从前面可知数据库当前联机日志文件seq号为69）也就是说restore database preview显示的归档列表结果中最后一个归档seq号总是比当前联机日志（当前联机日志也就是查看v$log状态为currnt的日志组）文件seq号小于1.
2 结合当前数据库的联机日志组seq号分别为67 68 69，可以判断：在recover应用最后一个归档seq号为66后，oracle会读取seq号为67、68、69联机日志文件继续推进该数据库来实现整个数据库完全恢复过程。
下面将演示整个验证和恢复过程:
4 验证恢复时需要用到的备份集是否能够正常恢复。
RMAN> restore validate database;
注意：这条命令直接会去rman资料库中找最新的备份集进行验证，也就是restore database preview命令显示的备份集。
Starting restore at 2014-02-17 16:14:59
using channel ORA_DISK_1
channel ORA_DISK_1: starting validation ofdatafile backup set
channel ORA_DISK_1: reading from backuppiece /backup/fullbk-20140217-3ep0rj0h_1_1.bak
channel ORA_DISK_1: piecehandle=/backup/fullbk-20140217-3ep0rj0h_1_1.bak tag=TAG20140217T120328
channel ORA_DISK_1: restored backup piece 1
channel ORA_DISK_1: validation complete,elapsed time: 00:00:36
Finished restore at 2014-02-17 16:15:35
5 验证恢复时应用的归档
RMAN> restore validate archivelogsequence between 59 and 66;
Starting restore at 2014-02-17 16:16:34
using channel ORA_DISK_1
channel ORA_DISK_1: scanning archived log /oracle/archivelog/arch_1_59_839098938.arch
channel ORA_DISK_1: scanning archived log/oracle/archivelog/arch_1_60_839098938.arch
channel ORA_DISK_1: scanning archived log/oracle/archivelog/arch_1_61_839098938.arch
channel ORA_DISK_1: scanning archived log/oracle/archivelog/arch_1_62_839098938.arch
channel ORA_DISK_1: scanning archived log/oracle/archivelog/arch_1_63_839098938.arch
channel ORA_DISK_1: scanning archived log/oracle/archivelog/arch_1_64_839098938.arch
channel ORA_DISK_1: scanning archived log/oracle/archivelog/arch_1_65_839098938.arch
channel ORA_DISK_1: scanning archived log/oracle/archivelog/arch_1_66_839098938.arch
Finished restore at 2014-02-17 16:16:37
6 执行restore和recover过程如下
RMAN> restore database;
Starting restore at 2014-02-17 16:36:23
allocated channel: ORA_DISK_1
channel ORA_DISK_1: SID=63 device type=DISK
channel ORA_DISK_1: starting datafilebackup set restore
channel ORA_DISK_1: specifying datafile(s)to restore from backup set
channel ORA_DISK_1: restoring datafile 00001to /oracle/CRM/CRM/system01.dbf
channel ORA_DISK_1: restoring datafile00002 to /oracle/CRM/CRM/sysaux01.dbf
channel ORA_DISK_1: restoring datafile00003 to /oracle/CRM/CRM/undotbs01.dbf
channel ORA_DISK_1: restoring datafile00004 to /oracle/CRM/CRM/users01.dbf
channel ORA_DISK_1: restoring datafile00005 to /oracle/CRM/CRM/crm.dbf
channel ORA_DISK_1: restoring datafile00006 to /oracle/CRM/test.dbf
channel ORA_DISK_1: restoring datafile00007 to /oracle/CRM/user01.dbf
channel ORA_DISK_1: restoring datafile00008 to /oracle/CRM/jxc.dbf
channel ORA_DISK_1: reading from backuppiece /backup/fullbk-20140217-3ep0rj0h_1_1.bak
channel ORA_DISK_1: piecehandle=/backup/fullbk-20140217-3ep0rj0h_1_1.bak tag=TAG20140217T120328
channel ORA_DISK_1: restored backup piece 1
channel ORA_DISK_1: restore complete,elapsed time: 00:02:08
Finished restore at 2014-02-17 16:38:35
注意：restore后我们通过查询x$kcvfh的redo字节地址(RBA)的seq号(也就是是FHRBA_SEQ字段)可以得到restore database 后数据文件头部记录的rba.seq号, 该值近一步表明recover过程需要从seq号为59归档开始应用。
或者也可以从restore database后数据文件头部的scn值，对比归档的first_change# 和 next_change# 推断出recover 需要应用归档开始。
SQL> select hxfil,fhscn,fhrba_seq fromx$kcvfh;
HXFIL FHSCN FHRBA_SEQ
---------- ---------------- ----------
1 4028039 59
2 4028039 59
3 4028039 59
4 4028039 59
5 4028039 59
6 4028039 59
7 4028039 59
8 4028039 59
8 rows selected.
当然restore database 后，我们也可以直接查询v$recvoery_log来得到recover过程需要应用的归档条目，如下所示：
select * from v$recovery_log;
THREAD# SEQUENCE# TIME ARCHIVE_NAME
---------- ---------- --------- ------------------------------------------------------------
1 59 17-FEB-14/oracle/archivelog/arch_1_59_839098938.arch
1 60 17-FEB-14/oracle/archivelog/arch_1_60_839098938.arch
1 61 17-FEB-14 /oracle/archivelog/arch_1_61_839098938.arch
1 62 17-FEB-14/oracle/archivelog/arch_1_62_839098938.arch
1 63 17-FEB-14/oracle/archivelog/arch_1_63_839098938.arch
1 64 17-FEB-14/oracle/archivelog/arch_1_64_839098938.arch
1 65 17-FEB-14/oracle/archivelog/arch_1_65_839098938.arch
1 66 17-FEB-14/oracle/archivelog/arch_1_66_839098938.arch
RMAN> recover database;
Starting recover at 2014-02-17 16:45:01
using channel ORA_DISK_1
starting media recovery
archived log for thread 1 with sequence 59is already on disk as file /oracle/archivelog/arch_1_59_839098938.arch
archived log for thread 1 with sequence 60is already on disk as file /oracle/archivelog/arch_1_60_839098938.arch
archived log for thread 1 with sequence 61is already on disk as file /oracle/archivelog/arch_1_61_839098938.arch
archived log for thread 1 with sequence 62is already on disk as file /oracle/archivelog/arch_1_62_839098938.arch
archived log for thread 1 with sequence 63is already on disk as file /oracle/archivelog/arch_1_63_839098938.arch
archived log for thread 1 with sequence 64is already on disk as file /oracle/archivelog/arch_1_64_839098938.arch
archived log for thread 1 with sequence 65is already on disk as file /oracle/archivelog/arch_1_65_839098938.arch
archived log for thread 1 with sequence 66is already on disk as file /oracle/archivelog/arch_1_66_839098938.arch
archived log for thread 1 with sequence 67is already on disk as file /oracle/archivelog/arch_1_67_839098938.arch
archived log for thread 1 with sequence 68is already on disk as file /oracle/archivelog/arch_1_68_839098938.arch
archived log filename=/oracle/archivelog/arch_1_59_839098938.arch thread=1 sequence=59
archived log file name=/oracle/archivelog/arch_1_60_839098938.archthread=1 sequence=60
archived log filename=/oracle/archivelog/arch_1_61_839098938.arch thread=1 sequence=61
archived log filename=/oracle/archivelog/arch_1_62_839098938.arch thread=1 sequence=62
archived log filename=/oracle/archivelog/arch_1_63_839098938.arch thread=1 sequence=63
archived log filename=/oracle/archivelog/arch_1_64_839098938.arch thread=1 sequence=64
archived log filename=/oracle/archivelog/arch_1_65_839098938.arch thread=1 sequence=65
archived log filename=/oracle/archivelog/arch_1_66_839098938.arch thread=1 sequence=66
media recovery complete, elapsed time:00:00:08
Finished recover at 2014-02-17 16:45:16
注意：这里可以清楚的看到应用的归档条目（红色标记处）
7 跟踪recover过程内容如下：
alter database recoverlogfile '/oracle/archivelog/arch_1_59_839098938.arch'
Media Recovery Log/oracle/archivelog/arch_1_59_839098938.arch
Mon Feb 17 16:45:12 2014
ORA-279 signalled during: alter databaserecover logfile '/oracle/archivelog/arch_1_59_839098938.arch'...
alter database recoverlogfile '/oracle/archivelog/arch_1_60_839098938.arch'
Media Recovery Log/oracle/archivelog/arch_1_60_839098938.arch
ORA-279 signalled during: alter databaserecover logfile '/oracle/archivelog/arch_1_60_839098938.arch'...
alter database recoverlogfile '/oracle/archivelog/arch_1_61_839098938.arch'
Media Recovery Log/oracle/archivelog/arch_1_61_839098938.arch
ORA-279 signalled during: alter databaserecover logfile '/oracle/archivelog/arch_1_61_839098938.arch'...
alter database recoverlogfile '/oracle/archivelog/arch_1_62_839098938.arch'
Media Recovery Log/oracle/archivelog/arch_1_62_839098938.arch
ORA-279 signalled during: alter databaserecover logfile '/oracle/archivelog/arch_1_62_839098938.arch'...
alter database recoverlogfile '/oracle/archivelog/arch_1_63_839098938.arch'
Media Recovery Log/oracle/archivelog/arch_1_63_839098938.arch
ORA-279 signalled during: alter databaserecover logfile '/oracle/archivelog/arch_1_63_839098938.arch'...
alter database recoverlogfile '/oracle/archivelog/arch_1_64_839098938.arch'
Media Recovery Log/oracle/archivelog/arch_1_64_839098938.arch
ORA-279 signalled during: alter databaserecover logfile '/oracle/archivelog/arch_1_64_839098938.arch'...
alter database recoverlogfile '/oracle/archivelog/arch_1_65_839098938.arch'
Media Recovery Log/oracle/archivelog/arch_1_65_839098938.arch
ORA-279 signalled during: alter databaserecover logfile '/oracle/archivelog/arch_1_65_839098938.arch'...
alter database recoverlogfile '/oracle/archivelog/arch_1_66_839098938.arch'
Media Recovery Log/oracle/archivelog/arch_1_66_839098938.arch
Mon Feb 17 16:45:14 2014
Recovery of Online RedoLog: Thread 1 Group 1 Seq 67 Reading mem 0
Mem# 0: /oracle/CRM/CRM/redo01.log
Recovery of Online RedoLog: Thread 1 Group 2 Seq 68 Reading mem 0
Mem# 0: /oracle/CRM/CRM/redo02.log
Recovery of Online RedoLog: Thread 1 Group 3 Seq 69 Reading mem 0
Mem# 0: /oracle/CRM/CRM/redo03.log
Media Recovery Complete(CRM)
注意：通过跟踪整个恢复过程，可以清楚的观察到在用recover进行完全恢复时，先应用归档，后再通过所有联机日志文件推进整个数据库来实现完全恢复的过程。
8 如果数据库进行不完全恢复如何获取恢复所需要的归档
以基于时间点恢复为例，我们可以这么使用得出恢复到这个时间点数据库需要的归档列表。
run{
sql 'alter session setnls_date_format="yyyy-mm-dd hh24:mi:ss"';
set until time='2013-12-09:05:50:12';
restore database preview;
}
总结：
1 在对数据库进行恢复的时，第一步先看看数据库是否归档，第二步看看数据库是否有备份，第三步验证备份和归档的有效性。最后执行整个恢复过程。
2 完全恢复时，通过对比restore database preview 显示的归档列表seq号和联机日志组的seq号，我们便可以清楚的推出数据库完全恢复时，recover需要应用的归档。
