RMAN是Oracle推出的官方备份还原工具。经过几个大版本的发展，RMAN已经支持多种备份介质和恢复策略的主要工具，也是业界普遍认可是Oracle备份还原官方策略。
Archivelog是Oracle备份还原策略的重要组成元素，不完全备份+连续的归档日志可以让我们将数据库恢复到发生故障点，实现数据的无损失恢复。但是，现实生活中archive log给没有经验的运维人员也带来了不少的问题，归档空间占满引起Hang住、瞬间归档日志过多生成引起问题等。一些前辈也在不断强调“归档模式不美好”。
在RMAN工作参数中，针对archive log，是可以设置专门的删除策略（Deletion）。在实践领域中，已经备份过或者确保安全传输的归档日志，其实就可以删除了，特别是在有限的Fast Recovery Area管理模式下。对于自动删除archive log的策略，比较常见的是applied to standby和shipped to standby，也就是Data Guard场景下。
本篇介绍简单的backed up参数使用情况，并通过一系列实验去研究该参数影响下Oracle和RMAN的工作行为特性。
1、基本参数和实验环境
笔者使用Oracle 11gR2进行测试，具体版本编号为11.2.0.4。
SQL> select * from v$version;
BANNER
--------------------------------------------------------------------------------
Oracle Database 11g Enterprise Edition Release 11.2.0.3.0 - 64bit Production
PL/SQL Release 11.2.0.3.0 - Production
CORE 11.2.0.3.0 Production
TNS for Linux: Version 11.2.0.3.0 - Production
NLSRTL Version 11.2.0.3.0 – Production
默认情况下，archivelog deletion policy参数为NONE。
RMAN> show all;
RMAN configuration parameters for database with db_unique_name XXXXDB are:
CONFIGURE RETENTION POLICY TO REDUNDANCY 1; # default
（篇幅原因，有省略……）
CONFIGURE ARCHIVELOG DELETION POLICY TO NONE; # default
该参数常见的集中取值如下：
configure archivelog deletion policy to backed up 2 times to sbt;
configure archivelog deletion policy to backed up 1 times to device type disk;
configure archivelog deletion policy to applied on standby; --DG专用
configure archivelog deletion policy to shipped on standby; --DG专用
configure archivelog deletion policy clear;
研究archivelog行为最好的工具视图是v$archived_log。很多DBA喜欢从操作系统层面删除归档日志，但是这种方式是不会直接被Oracle控制文件认可，所以建议使用RMAN或者官方工具来做。
--已归档未删除日志
SQL> select count(*) from v$archived_log where archived='YES' and deleted='NO';
COUNT(*)
----------
13
2、第一轮备份测试实验
首先我们修改archivelog deletion policy参数，设置为“两次备份后即可以删除”。
RMAN> configure archivelog deletion policy to backed up 2 times to device type disk;
new RMAN configuration parameters:
CONFIGURE ARCHIVELOG DELETION POLICY TO BACKED UP 2 TIMES TO DISK;
new RMAN configuration parameters are successfully stored
手工备份数据库和归档日志，不进行删除动作。
RMAN> backup database plus archivelog;
Starting backup at 21-SEP-15
current log archived
allocated channel: ORA_DISK_1
channel ORA_DISK_1: SID=16 device type=DISK
channel ORA_DISK_1: starting archived log backup set
channel ORA_DISK_1: specifying archived log(s) in backup set
input archived log thread=1 sequence=100 RECID=12 STAMP=890690423
input archived log thread=1 sequence=101 RECID=13 STAMP=890712061
input archived log thread=1 sequence=102 RECID=14 STAMP=890727732
input archived log thread=1 sequence=103 RECID=15 STAMP=890776815
input archived log thread=1 sequence=104 RECID=16 STAMP=890776833
input archived log thread=1 sequence=105 RECID=17 STAMP=890805616
input archived log thread=1 sequence=106 RECID=18 STAMP=890814181
input archived log thread=1 sequence=107 RECID=19 STAMP=890820201
input archived log thread=1 sequence=108 RECID=20 STAMP=890859629
input archived log thread=1 sequence=109 RECID=21 STAMP=890892046
input archived log thread=1 sequence=110 RECID=22 STAMP=890900632
input archived log thread=1 sequence=111 RECID=23 STAMP=890906655
input archived log thread=1 sequence=112 RECID=24 STAMP=890942416
input archived log thread=1 sequence=113 RECID=25 STAMP=890990204
channel ORA_DISK_1: starting piece 1 at 21-SEP-15
channel ORA_DISK_1: finished piece 1 at 21-SEP-15
piece handle=/u01/app/oracle/fast_recovery_area/XXXXDB/backupset/2015_09_21/o1_mf_annnn_TAG20150921T091644_bzypmwty_.bkp tag=TAG20150921T091644
（篇幅原因，省略部分……）
Finished Control File and SPFILE Autobackup at 21-SEP-15
此时，归档日志被备份，并且没有删除。
--多出来的两个是由于进行备份时候自动会有switch log
SQL> select count(*) from v$archived_log where archived='YES' and deleted='NO';
COUNT(*)
----------
15
下面进行第二次实验。
RMAN> backup database plus archivelog;
Starting backup at 21-SEP-15
current log archived
using channel ORA_DISK_1
channel ORA_DISK_1: starting archived log backup set
channel ORA_DISK_1: specifying archived log(s) in backup set
input archived log thread=1 sequence=100 RECID=12 STAMP=890690423
input archived log thread=1 sequence=101 RECID=13 STAMP=890712061
input archived log thread=1 sequence=102 RECID=14 STAMP=890727732
input archived log thread=1 sequence=103 RECID=15 STAMP=890776815
input archived log thread=1 sequence=104 RECID=16 STAMP=890776833
input archived log thread=1 sequence=105 RECID=17 STAMP=890805616
input archived log thread=1 sequence=106 RECID=18 STAMP=890814181
input archived log thread=1 sequence=107 RECID=19 STAMP=890820201
input archived log thread=1 sequence=108 RECID=20 STAMP=890859629
input archived log thread=1 sequence=109 RECID=21 STAMP=890892046
input archived log thread=1 sequence=110 RECID=22 STAMP=890900632
input archived log thread=1 sequence=111 RECID=23 STAMP=890906655
input archived log thread=1 sequence=112 RECID=24 STAMP=890942416
input archived log thread=1 sequence=113 RECID=25 STAMP=890990204
input archived log thread=1 sequence=114 RECID=26 STAMP=890990263
input archived log thread=1 sequence=115 RECID=27 STAMP=890990391
channel ORA_DISK_1: starting piece 1 at 21-SEP-15
channel ORA_DISK_1: finished piece 1 at 21-SEP-15
piece handle=/u01/app/oracle/fast_recovery_area/XXXXDB/backupset/2015_09_21/o1_mf_annnn_TAG20150921T091951_bzypsqj3_.bkp tag=TAG20150921T091951
（篇幅原因，有省略……）
Finished Control File and SPFILE Autobackup at 21-SEP-15
第二次备份，之前备份过的日志还出现在自动备份的列表中。但是，在第二次备份的时候，已经备份过两次（deletion policy）的日志并没有自动删除。
SQL> select count(*) from v$archived_log where archived='YES' and deleted='NO';
COUNT(*)
----------
17
归档日志还在fast recovery area中。
[oracle@Databaseintrawebpro fast_recovery_area]$ du -h
19M ./XXXXDB/autobackup/2015_09_21
9.4M ./XXXXDB/autobackup/2015_09_17
29M ./XXXXDB/autobackup
151M ./XXXXDB/onlinelog
6.0G ./XXXXDB/backupset/2015_09_21
108K ./XXXXDB/backupset/2015_09_17
6.0G ./XXXXDB/backupset
125M ./XXXXDB/archivelog/2015_09_19
27M ./XXXXDB/archivelog/2015_09_21
4.0K ./XXXXDB/archivelog/2015_09_15
127M ./XXXXDB/archivelog/2015_09_18
121M ./XXXXDB/archivelog/2015_09_20
4.0K ./XXXXDB/archivelog/2015_09_16
32M ./XXXXDB/archivelog/2015_09_17
431M ./XXXXDB/archivelog
9.4M ./XXXXDB/controlfile
6.6G ./XXXXDB
6.6G .
此时，归档日志和备份次数，在v$archived_log中可以方便的找出来。
SQL> alter system switch logfile;
System altered
SQL> select count(*) from v$archived_log where archived='YES' and deleted='NO';
COUNT(*)
----------
18
--注意这些已经备份过两次的recid编号
SQL> select recid, sequence#, archived, deleted, backup_count from v$archived_log where backup_count>1;
RECID SEQUENCE# ARCHIVED DELETED BACKUP_COUNT
---------- ---------- -------- ------- ------------
12 100 YES NO 2
13 101 YES NO 2
14 102 YES NO 2
15 103 YES NO 2
16 104 YES NO 2
17 105 YES NO 2
18 106 YES NO 2
19 107 YES NO 2
20 108 YES NO 2
21 109 YES NO 2
22 110 YES NO 2
23 111 YES NO 2
24 112 YES NO 2
25 113 YES NO 2
26 114 YES NO 2
15 rows selected
进行第三次备份。
RMAN> backup database plus archivelog;
Starting backup at 21-SEP-15
current log archived
allocated channel: ORA_DISK_1
channel ORA_DISK_1: SID=498 device type=DISK
skipping archived logs of thread 1 from sequence 100 to 114; already backed up
channel ORA_DISK_1: starting archived log backup set
channel ORA_DISK_1: specifying archived log(s) in backup set
input archived log thread=1 sequence=115 RECID=27 STAMP=890990391
input archived log thread=1 sequence=116 RECID=28 STAMP=890990481
input archived log thread=1 sequence=117 RECID=29 STAMP=890990667
input archived log thread=1 sequence=118 RECID=30 STAMP=890993128
channel ORA_DISK_1: starting piece 1 at 21-SEP-15
channel ORA_DISK_1: finished piece 1 at 21-SEP-15
piece
（篇幅原因，有省略…….）
Finished Control File and SPFILE Autobackup at 21-SEP-15
注意：备份过两次的日志，没有出现在RMAN自动备份的列表中。这里我们定义到了删除策略的一个行为：当满足删除条件的时候，归档日志是不会进入备份集合列表的。
归档日志信息：
SQL> select recid, sequence#, archived, deleted, backup_count from v$archived_log where backup_count>1;
RECID SEQUENCE# ARCHIVED DELETED BACKUP_COUNT
---------- ---------- -------- ------- ------------
12 100 YES YES 2
13 101 YES YES 2
14 102 YES YES 2
15 103 YES YES 2
16 104 YES YES 2
17 105 YES YES 2
18 106 YES YES 2
19 107 YES YES 2
20 108 YES YES 2
21 109 YES YES 2
22 110 YES YES 2
23 111 YES YES 2
24 112 YES YES 2
25 113 YES YES 2
26 114 YES NO 2
27 115 YES NO 2
28 116 YES NO 2
17 rows selected
注意：一部分归档日志被删除，但是并没有所有上次备份过两次的日志都删除掉了，比如recid=26的日志。此时，备份fast recovery area空间情况发生了变化。
[oracle@Databaseintrawebpro fast_recovery_area]$ du -h
29M ./XXXXDB/autobackup/2015_09_21
4.0K ./XXXXDB/autobackup/2015_09_17
29M ./XXXXDB/autobackup
151M ./XXXXDB/onlinelog
5.5G ./XXXXDB/backupset/2015_09_21
4.0K ./XXXXDB/backupset/2015_09_17
5.5G ./XXXXDB/backupset
4.0K ./XXXXDB/archivelog/2015_09_19
2.5M ./XXXXDB/archivelog/2015_09_21
4.0K ./XXXXDB/archivelog/2015_09_15
4.0K ./XXXXDB/archivelog/2015_09_18
4.0K ./XXXXDB/archivelog/2015_09_20
4.0K ./XXXXDB/archivelog/2015_09_16
4.0K ./XXXXDB/archivelog/2015_09_17
2.6M ./XXXXDB/archivelog
9.4M ./XXXXDB/controlfile
5.7G ./XXXXDB
5.7G .
在alert log中，我们看到了Oracle自动删除的动作。
Mon Sep 21 09:24:27 2015
Expanded controlfile section 11 from 28 to 56 records
Requested to grow by 28 records; added 1 blocks of records
Archived Log entry 29 added for thread 1 sequence 117 ID 0x774e158c dest 1:
Mon Sep 21 10:05:28 2015
ALTER SYSTEM ARCHIVE LOG
Mon Sep 21 10:05:28 2015
Thread 1 advanced to log sequence 119 (LGWR switch)
Current log# 2 seq# 119 mem# 0: /u01/app/oracle/oradata/XXXXDB/onlinelog/o1_mf_2_bxzzjj5w_.log
Current log# 2 seq# 119 mem# 1: /u01/app/oracle/fast_recovery_area/XXXXDB/onlinelog/o1_mf_2_bxzzjj80_.log
Archived Log entry 30 added for thread 1 sequence 118 ID 0x774e158c dest 1:
Mon Sep 21 10:05:47 2015
Deleted Oracle managed file /u01/app/oracle/fast_recovery_area/XXXXDB/backupset/2015_09_17/o1_mf_annnn_TAG20150917T195557_bzoblfck_.bkp
Deleted Oracle managed file /u01/app/oracle/fast_recovery_area/XXXXDB/archivelog/2015_09_17/o1_mf_1_100_bzokvqj0_.arc
Deleted Oracle managed file /u01/app/oracle/fast_recovery_area/XXXXDB/autobackup/2015_09_17/o1_mf_s_890682958_bzoblglw_.bkp
Deleted Oracle managed file /u01/app/oracle/fast_recovery_area/XXXXDB/archivelog/2015_09_18/o1_mf_1_101_bzp6zx31_.arc
Deleted Oracle managed file /u01/app/oracle/fast_recovery_area/XXXXDB/archivelog/2015_09_18/o1_mf_1_102_bzpp9nln_.arc
Deleted Oracle managed file /u01/app/oracle/fast_recovery_area/XXXXDB/archivelog/2015_09_18/o1_mf_1_103_bzr67h1h_.arc
Deleted Oracle managed file /u01/app/oracle/fast_recovery_area/XXXXDB/archivelog/2015_09_18/o1_mf_1_104_bzr6812s_.arc
Deleted Oracle managed file /u01/app/oracle/fast_recovery_area/XXXXDB/archivelog/2015_09_19/o1_mf_1_105_bzs2cj5y_.arc
Deleted Oracle managed file /u01/app/oracle/fast_recovery_area/XXXXDB/archivelog/2015_09_19/o1_mf_1_106_bzsbq54p_.arc
Deleted Oracle managed file /u01/app/oracle/fast_recovery_area/XXXXDB/archivelog/2015_09_19/o1_mf_1_107_bzsjm99v_.arc
Deleted Oracle managed file /u01/app/oracle/fast_recovery_area/XXXXDB/archivelog/2015_09_19/o1_mf_1_108_bztq3f2v_.arc
Deleted Oracle managed file /u01/app/oracle/fast_recovery_area/XXXXDB/archivelog/2015_09_20/o1_mf_1_109_bzvprgf1_.arc
Deleted Oracle managed file /u01/app/oracle/fast_recovery_area/XXXXDB/archivelog/2015_09_20/o1_mf_1_110_bzvz4rj7_.arc
Deleted Oracle managed file /u01/app/oracle/fast_recovery_area/XXXXDB/archivelog/2015_09_20/o1_mf_1_111_bzw50zmb_.arc
Deleted Oracle managed file /u01/app/oracle/fast_recovery_area/XXXXDB/archivelog/2015_09_20/o1_mf_1_112_bzx7yj9g_.arc
Deleted Oracle managed file /u01/app/oracle/fast_recovery_area/XXXXDB/archivelog/2015_09_21/o1_mf_1_113_bzypmw8c_.arc
Deleted Oracle managed file /u01/app/oracle/fast_recovery_area/XXXXDB/backupset/2015_09_21/o1_mf_annnn_TAG20150921T091644_bzypmwty_.bkp
Mon Sep 21 10:05:58 2015
Deleted Oracle managed file /u01/app/oracle/fast_recovery_area/XXXXDB/backupset/2015_09_21/o1_mf_nnndf_TAG20150921T091647_bzypn055_.bkp
Mon Sep 21 10:06:15 2015
ALTER SYSTEM ARCHIVE LOG
Mon Sep 21 10:06:15 2015
Thread 1 advanced to log sequence 120 (LGWR switch)
Current log# 3 seq# 120 mem# 0: /u01/app/oracle/oradata/XXXXDB/onlinelog/o1_mf_3_bxzzjl0z_.log
Current log# 3 seq# 120 mem# 1: /u01/app/oracle/fast_recovery_area/XXXXDB/onlinelog/o1_mf_3_bxzzjl35_.log
Archived Log entry 31 added for thread 1 sequence 119 ID 0x774e158c dest 1:
日志里面，Oracle不仅仅删除了部分备份过两次的日志，而且删除了一个已经obsolete的备份集合。这个体现了一个特性：当fast recovery area空间紧张的时候，obsolete和符合删除deletion policy的归档日志会被自动删除。
更多详情见请继续阅读下一页的精彩内容：