谈谈Oracle日志文件的INVALID状态
在Oracle数据库中,三大文件(Data File、Control File和Online Redo Log)扮演着极其重要的地位和作用。其中,Redo Log和Oracle Redo Log机制更是Oracle正常、安全运行的保证。
Oracle Online Redo Log采用日志组(Logfile Group)和组成员(Group Member)的方式。每个Log File Group中的成员内容保持镜像相同,生成的每一条Redo Log Entry都需要写入到每个成员中。原则上,各个日志成员保存在不同的存储位置,这样一旦发生介质损坏,可以使用其他成员进行恢复。
Oracle数据库提供了对日志组和日志成员管理的语句操作支持。我们可以方便的添加、删除日志组,向日志组中添加、删除成员文件。但是,在一般的情况下,我们还是希望能够一次性将日志组所有成员确定添加。这样可以避免很多问题,其中之一就是日志文件的INVALID状态。
1、实验环境介绍
本文选择Oracle 11g进行实验测试。
SQL> select * from v$version;
BANNER
--------------------------------------------------------------------------------
Oracle Database 11g Enterprise Edition Release 11.2.0.1.0 - Production
PL/SQL Release 11.2.0.1.0 - Production
CORE 11.2.0.1.0 Production
当前系统中有三个日志组,每个组成员两个。当前第三组是Current状态Redo Log Group。
SQL> select group#, sequence#, members, status, first_change#, next_change# from v$log;
GROUP# SEQUENCE# MEMBERS STATUS FIRST_CHANGE# NEXT_CHANGE#
---------- ---------- ---------- ---------------- ------------- ------------
1 73 2 INACTIVE 1560017 1580625
2 74 2 INACTIVE 1580625 1600939
3 75 2 CURRENT 1600939 281474976710
SQL> select group#, status, type, member from v$logfile;
GROUP# STATUS TYPE MEMBER
---------- ------- ------- --------------------------------------------------------------------------------
3 ONLINE /u01/oradata/WILSON/onlinelog/o1_mf_3_7xt456o0_.log
3 ONLINE /u01/flash_recovery_area/WILSON/onlinelog/o1_mf_3_7xt45bvp_.log
2 ONLINE /u01/oradata/WILSON/onlinelog/o1_mf_2_7xt44w3k_.log
2 ONLINE /u01/flash_recovery_area/WILSON/onlinelog/o1_mf_2_7xt450rv_.log
1 ONLINE /u01/oradata/WILSON/onlinelog/o1_mf_1_7xt44nq1_.log
1 ONLINE /u01/flash_recovery_area/WILSON/onlinelog/o1_mf_1_7xt44qt6_.log
6 rows selected
在v$logfile中,有一个状态列status,默认情况下是为空。
2、非Current日志组成员添加
如果当前我们希望给日志组添加一个member的时候,Oracle是提供了相应的语句的。当前日志组是group 3,我们先给非current组添加。
SQL> alter database add logfile member '/u01/flash_recovery_area/WILSON/onlinelog/redolog01a.log' to group 1;
Database altered
SQL> select group#, status, type, member from v$logfile;
GROUP# STATUS TYPE MEMBER
---------- ------- ------- --------------------------------------------------------------------------------
3 ONLINE /u01/oradata/WILSON/onlinelog/o1_mf_3_7xt456o0_.log
3 ONLINE /u01/flash_recovery_area/WILSON/onlinelog/o1_mf_3_7xt45bvp_.log
2 ONLINE /u01/oradata/WILSON/onlinelog/o1_mf_2_7xt44w3k_.log
2 ONLINE /u01/flash_recovery_area/WILSON/onlinelog/o1_mf_2_7xt450rv_.log
1 ONLINE /u01/oradata/WILSON/onlinelog/o1_mf_1_7xt44nq1_.log
1 ONLINE /u01/flash_recovery_area/WILSON/onlinelog/o1_mf_1_7xt44qt6_.log
1 INVALID ONLINE /u01/flash_recovery_area/WILSON/onlinelog/redolog01a.log
7 rows selected
SQL> select group#, sequence#, members, status, first_change#, next_change# from v$log;
GROUP# SEQUENCE# MEMBERS STATUS FIRST_CHANGE# NEXT_CHANGE#
---------- ---------- ---------- ---------------- ------------- ------------
1 73 3 INACTIVE 1560017 1580625
2 74 2 INACTIVE 1580625 1600939
3 75 2 CURRENT 1600939 281474976710
注意,我们的确给group 1添加了文件。但是该文件的状态是INVALID。在OS层面,我们观察一下:
[oracle@bspdev onlinelog]$ pwd
/u01/flash_recovery_area/WILSON/onlinelog
[oracle@bspdev onlinelog]$ ls -l
total 205040
-rw-r----- 1 oracle oinstall 52429312 Sep 10 03:56 o1_mf_1_7xt44qt6_.log
-rw-r----- 1 oracle oinstall 52429312 Sep 10 03:56 o1_mf_2_7xt450rv_.log
-rw-r----- 1 oracle oinstall 52429312 Sep 10 04:04 o1_mf_3_7xt45bvp_.log
-rw-r----- 1 oracle oinstall 52429312 Sep 10 04:03 redolog01a.log
[oracle@bspdev onlinelog]$
在OS层面上,Oracle的确创建了文件redolog01a.log,而且大小和其他组成员大小相同。那么这个INVALID状态是什么意思呢?如果我们切换日志组,转换到文件上,有什么问题吗?
SQL> alter system switch logfile;
System altered
SQL> select group#, sequence#, members, status, first_change#, next_change# from v$log;
GROUP# SEQUENCE# MEMBERS STATUS FIRST_CHANGE# NEXT_CHANGE#
---------- ---------- ---------- ---------------- ------------- ------------
1 76 3 CURRENT 1601735 281474976710
2 74 2 INACTIVE 1580625 1600939
3 75 2 ACTIVE 1600939 1601735
SQL> select group#, status, type, member from v$logfile;
GROUP# STATUS TYPE MEMBER
---------- ------- ------- --------------------------------------------------------------------------------
(篇幅原因,有省略……)
2 ONLINE /u01/flash_recovery_area/WILSON/onlinelog/o1_mf_2_7xt450rv_.log
1 ONLINE /u01/oradata/WILSON/onlinelog/o1_mf_1_7xt44nq1_.log
1 ONLINE /u01/flash_recovery_area/WILSON/onlinelog/o1_mf_1_7xt44qt6_.log
1 ONLINE /u01/flash_recovery_area/WILSON/onlinelog/redolog01a.log
7 rows selected
当进行switch logfile之后,redolog01a.log文件的状态变为空,表示一切正常。可以猜想:当日志切换到一个新的日志组的时候,会将日志原有的内容覆盖刷新。这个过程中,伴随着各个成员Member保持一致。此时,Invalid状态的日志文件就会得到同步的机会。
进一步想,标记为Invalid状态的日志文件,虽然大小已经确定,但是内容上没有与其他同Group Member同步。所以Oracle将其标记为Invalid。
根据Oracle的工作规则,如果处在Invalid状态的日志文件组不是Current日志组,那么在日志切换到该组的时候,内容都会被覆盖。由此,Invalid状态消失。
那么,如果当前Invalid所在是一个Current日志组,Lgwr不断的写入内容。这个过程是如何呢?
3、Current日志组成员添加
接着上面的实验结果,如果往当前Current日志组中添加成员,效果如何呢?
SQL> alter database add logfile member '/u01/flash_recovery_area/WILSON/onlinelog/redolog01b.log' to group 1;
Database altered
SQL> select group#, status, type, member from v$logfile;
GROUP# STATUS TYPE MEMBER
---------- ------- ------- --------------------------------------------------------------------------------
(篇幅原因,省略….)
1 ONLINE /u01/flash_recovery_area/WILSON/onlinelog/redolog01a.log
1 INVALID ONLINE /u01/flash_recovery_area/WILSON/onlinelog/redolog01b.log
8 rows selected
SQL> select group#, sequence#, members, status, first_change#, next_change# from v$log;
GROUP# SEQUENCE# MEMBERS STATUS FIRST_CHANGE# NEXT_CHANGE#
---------- ---------- ---------- ---------------- ------------- ------------
1 76 4 CURRENT 1601735 281474976710
2 74 2 INACTIVE 1580625 1600939
3 75 2 ACTIVE 1600939 1601735
之后,我们实验生成Redo Entry之后,Lgwr会不会书写到Invalid状态的日志中。
SQL> col name for a10;
SQL> select * from v$statname where name='redo size';
STATISTIC# NAME CLASS STAT_ID
---------- ---------- ---------- ----------
169 redo size 2 1236385760
SQL> select * from v$mystat where STATISTIC#=169;
SID STATISTIC# VALUE
---------- ---------- ----------
43 169 8585072
SQL> delete t;
72591 rows deleted
SQL> commit;
Commit complete
SQL> select * from v$mystat where STATISTIC#=169;
SID STATISTIC# VALUE
---------- ---------- ----------
43 169 36099756
删除数据表T,生成了(36099756-858072)的redo log。那么,此时Invalid状态的日志文件如何呢?
SQL> select group#, status, type, member from v$logfile;
GROUP# STATUS TYPE MEMBER
---------- ------- ------- --------------------------------------------------------------------------------
(篇幅原因,省略部分……)
1 ONLINE /u01/flash_recovery_area/WILSON/onlinelog/o1_mf_1_7xt44qt6_.log
1 ONLINE /u01/flash_recovery_area/WILSON/onlinelog/redolog01a.log
1 INVALID ONLINE /u01/flash_recovery_area/WILSON/onlinelog/redolog01b.log
8 rows selected
SQL> select group#, sequence#, members, status, first_change#, next_change# from v$log;
GROUP# SEQUENCE# MEMBERS STATUS FIRST_CHANGE# NEXT_CHANGE#
---------- ---------- ---------- ---------------- ------------- ------------
1 76 4 CURRENT 1601735 281474976710
2 74 2 INACTIVE 1580625 1600939
3 75 2 INACTIVE 1600939 1601735
虽然生成了redo log,并且写入到了文件中,但是Invalid状态没有变化。说明,即使是当前的日志组,如果有处在Invalid状态日志,内容不会同步,新日志也不会写入到其中。
只有在日志组切换到下一个循环之后,才能最后实现同步。
SQL> alter system switch logfile;
System altered
SQL> alter system switch logfile;
System altered
SQL> alter system switch logfile;(执行时间很长)
System altered
SQL> select group#, status, type, member from v$logfile;
GROUP# STATUS TYPE MEMBER
---------- ------- ------- --------------------------------------------------------------------------------
(篇幅原因,省略部分……)
1 ONLINE /u01/flash_recovery_area/WILSON/onlinelog/redolog01b.log
8 rows selected
SQL> select group#, sequence#, members, status, first_change#, next_change# from v$log;
GROUP# SEQUENCE# MEMBERS STATUS FIRST_CHANGE# NEXT_CHANGE#
---------- ---------- ---------- ---------------- ------------- ------------
1 79 4 CURRENT 1603673 281474976710
2 77 2 ACTIVE 1603650 1603669
3 78 2 ACTIVE 1603669 1603673
可见,只有在循环之后才有消除Invalid状态。接下来,我们讨论Invalid对Oracle启动过程的影响。
Oracle Online Redo Log采用日志组(Logfile Group)和组成员(Group Member)的方式。每个Log File Group中的成员内容保持镜像相同,生成的每一条Redo Log Entry都需要写入到每个成员中。原则上,各个日志成员保存在不同的存储位置,这样一旦发生介质损坏,可以使用其他成员进行恢复。
Oracle数据库提供了对日志组和日志成员管理的语句操作支持。我们可以方便的添加、删除日志组,向日志组中添加、删除成员文件。但是,在一般的情况下,我们还是希望能够一次性将日志组所有成员确定添加。这样可以避免很多问题,其中之一就是日志文件的INVALID状态。
1、实验环境介绍
本文选择Oracle 11g进行实验测试。
SQL> select * from v$version;
BANNER
--------------------------------------------------------------------------------
Oracle Database 11g Enterprise Edition Release 11.2.0.1.0 - Production
PL/SQL Release 11.2.0.1.0 - Production
CORE 11.2.0.1.0 Production
当前系统中有三个日志组,每个组成员两个。当前第三组是Current状态Redo Log Group。
SQL> select group#, sequence#, members, status, first_change#, next_change# from v$log;
GROUP# SEQUENCE# MEMBERS STATUS FIRST_CHANGE# NEXT_CHANGE#
---------- ---------- ---------- ---------------- ------------- ------------
1 73 2 INACTIVE 1560017 1580625
2 74 2 INACTIVE 1580625 1600939
3 75 2 CURRENT 1600939 281474976710
SQL> select group#, status, type, member from v$logfile;
GROUP# STATUS TYPE MEMBER
---------- ------- ------- --------------------------------------------------------------------------------
3 ONLINE /u01/oradata/WILSON/onlinelog/o1_mf_3_7xt456o0_.log
3 ONLINE /u01/flash_recovery_area/WILSON/onlinelog/o1_mf_3_7xt45bvp_.log
2 ONLINE /u01/oradata/WILSON/onlinelog/o1_mf_2_7xt44w3k_.log
2 ONLINE /u01/flash_recovery_area/WILSON/onlinelog/o1_mf_2_7xt450rv_.log
1 ONLINE /u01/oradata/WILSON/onlinelog/o1_mf_1_7xt44nq1_.log
1 ONLINE /u01/flash_recovery_area/WILSON/onlinelog/o1_mf_1_7xt44qt6_.log
6 rows selected
在v$logfile中,有一个状态列status,默认情况下是为空。
2、非Current日志组成员添加
如果当前我们希望给日志组添加一个member的时候,Oracle是提供了相应的语句的。当前日志组是group 3,我们先给非current组添加。
SQL> alter database add logfile member '/u01/flash_recovery_area/WILSON/onlinelog/redolog01a.log' to group 1;
Database altered
SQL> select group#, status, type, member from v$logfile;
GROUP# STATUS TYPE MEMBER
---------- ------- ------- --------------------------------------------------------------------------------
3 ONLINE /u01/oradata/WILSON/onlinelog/o1_mf_3_7xt456o0_.log
3 ONLINE /u01/flash_recovery_area/WILSON/onlinelog/o1_mf_3_7xt45bvp_.log
2 ONLINE /u01/oradata/WILSON/onlinelog/o1_mf_2_7xt44w3k_.log
2 ONLINE /u01/flash_recovery_area/WILSON/onlinelog/o1_mf_2_7xt450rv_.log
1 ONLINE /u01/oradata/WILSON/onlinelog/o1_mf_1_7xt44nq1_.log
1 ONLINE /u01/flash_recovery_area/WILSON/onlinelog/o1_mf_1_7xt44qt6_.log
1 INVALID ONLINE /u01/flash_recovery_area/WILSON/onlinelog/redolog01a.log
7 rows selected
SQL> select group#, sequence#, members, status, first_change#, next_change# from v$log;
GROUP# SEQUENCE# MEMBERS STATUS FIRST_CHANGE# NEXT_CHANGE#
---------- ---------- ---------- ---------------- ------------- ------------
1 73 3 INACTIVE 1560017 1580625
2 74 2 INACTIVE 1580625 1600939
3 75 2 CURRENT 1600939 281474976710
注意,我们的确给group 1添加了文件。但是该文件的状态是INVALID。在OS层面,我们观察一下:
[oracle@bspdev onlinelog]$ pwd
/u01/flash_recovery_area/WILSON/onlinelog
[oracle@bspdev onlinelog]$ ls -l
total 205040
-rw-r----- 1 oracle oinstall 52429312 Sep 10 03:56 o1_mf_1_7xt44qt6_.log
-rw-r----- 1 oracle oinstall 52429312 Sep 10 03:56 o1_mf_2_7xt450rv_.log
-rw-r----- 1 oracle oinstall 52429312 Sep 10 04:04 o1_mf_3_7xt45bvp_.log
-rw-r----- 1 oracle oinstall 52429312 Sep 10 04:03 redolog01a.log
[oracle@bspdev onlinelog]$
在OS层面上,Oracle的确创建了文件redolog01a.log,而且大小和其他组成员大小相同。那么这个INVALID状态是什么意思呢?如果我们切换日志组,转换到文件上,有什么问题吗?
SQL> alter system switch logfile;
System altered
SQL> select group#, sequence#, members, status, first_change#, next_change# from v$log;
GROUP# SEQUENCE# MEMBERS STATUS FIRST_CHANGE# NEXT_CHANGE#
---------- ---------- ---------- ---------------- ------------- ------------
1 76 3 CURRENT 1601735 281474976710
2 74 2 INACTIVE 1580625 1600939
3 75 2 ACTIVE 1600939 1601735
SQL> select group#, status, type, member from v$logfile;
GROUP# STATUS TYPE MEMBER
---------- ------- ------- --------------------------------------------------------------------------------
(篇幅原因,有省略……)
2 ONLINE /u01/flash_recovery_area/WILSON/onlinelog/o1_mf_2_7xt450rv_.log
1 ONLINE /u01/oradata/WILSON/onlinelog/o1_mf_1_7xt44nq1_.log
1 ONLINE /u01/flash_recovery_area/WILSON/onlinelog/o1_mf_1_7xt44qt6_.log
1 ONLINE /u01/flash_recovery_area/WILSON/onlinelog/redolog01a.log
7 rows selected
当进行switch logfile之后,redolog01a.log文件的状态变为空,表示一切正常。可以猜想:当日志切换到一个新的日志组的时候,会将日志原有的内容覆盖刷新。这个过程中,伴随着各个成员Member保持一致。此时,Invalid状态的日志文件就会得到同步的机会。
进一步想,标记为Invalid状态的日志文件,虽然大小已经确定,但是内容上没有与其他同Group Member同步。所以Oracle将其标记为Invalid。
根据Oracle的工作规则,如果处在Invalid状态的日志文件组不是Current日志组,那么在日志切换到该组的时候,内容都会被覆盖。由此,Invalid状态消失。
那么,如果当前Invalid所在是一个Current日志组,Lgwr不断的写入内容。这个过程是如何呢?
3、Current日志组成员添加
接着上面的实验结果,如果往当前Current日志组中添加成员,效果如何呢?
SQL> alter database add logfile member '/u01/flash_recovery_area/WILSON/onlinelog/redolog01b.log' to group 1;
Database altered
SQL> select group#, status, type, member from v$logfile;
GROUP# STATUS TYPE MEMBER
---------- ------- ------- --------------------------------------------------------------------------------
(篇幅原因,省略….)
1 ONLINE /u01/flash_recovery_area/WILSON/onlinelog/redolog01a.log
1 INVALID ONLINE /u01/flash_recovery_area/WILSON/onlinelog/redolog01b.log
8 rows selected
SQL> select group#, sequence#, members, status, first_change#, next_change# from v$log;
GROUP# SEQUENCE# MEMBERS STATUS FIRST_CHANGE# NEXT_CHANGE#
---------- ---------- ---------- ---------------- ------------- ------------
1 76 4 CURRENT 1601735 281474976710
2 74 2 INACTIVE 1580625 1600939
3 75 2 ACTIVE 1600939 1601735
之后,我们实验生成Redo Entry之后,Lgwr会不会书写到Invalid状态的日志中。
SQL> col name for a10;
SQL> select * from v$statname where name='redo size';
STATISTIC# NAME CLASS STAT_ID
---------- ---------- ---------- ----------
169 redo size 2 1236385760
SQL> select * from v$mystat where STATISTIC#=169;
SID STATISTIC# VALUE
---------- ---------- ----------
43 169 8585072
SQL> delete t;
72591 rows deleted
SQL> commit;
Commit complete
SQL> select * from v$mystat where STATISTIC#=169;
SID STATISTIC# VALUE
---------- ---------- ----------
43 169 36099756
删除数据表T,生成了(36099756-858072)的redo log。那么,此时Invalid状态的日志文件如何呢?
SQL> select group#, status, type, member from v$logfile;
GROUP# STATUS TYPE MEMBER
---------- ------- ------- --------------------------------------------------------------------------------
(篇幅原因,省略部分……)
1 ONLINE /u01/flash_recovery_area/WILSON/onlinelog/o1_mf_1_7xt44qt6_.log
1 ONLINE /u01/flash_recovery_area/WILSON/onlinelog/redolog01a.log
1 INVALID ONLINE /u01/flash_recovery_area/WILSON/onlinelog/redolog01b.log
8 rows selected
SQL> select group#, sequence#, members, status, first_change#, next_change# from v$log;
GROUP# SEQUENCE# MEMBERS STATUS FIRST_CHANGE# NEXT_CHANGE#
---------- ---------- ---------- ---------------- ------------- ------------
1 76 4 CURRENT 1601735 281474976710
2 74 2 INACTIVE 1580625 1600939
3 75 2 INACTIVE 1600939 1601735
虽然生成了redo log,并且写入到了文件中,但是Invalid状态没有变化。说明,即使是当前的日志组,如果有处在Invalid状态日志,内容不会同步,新日志也不会写入到其中。
只有在日志组切换到下一个循环之后,才能最后实现同步。
SQL> alter system switch logfile;
System altered
SQL> alter system switch logfile;
System altered
SQL> alter system switch logfile;(执行时间很长)
System altered
SQL> select group#, status, type, member from v$logfile;
GROUP# STATUS TYPE MEMBER
---------- ------- ------- --------------------------------------------------------------------------------
(篇幅原因,省略部分……)
1 ONLINE /u01/flash_recovery_area/WILSON/onlinelog/redolog01b.log
8 rows selected
SQL> select group#, sequence#, members, status, first_change#, next_change# from v$log;
GROUP# SEQUENCE# MEMBERS STATUS FIRST_CHANGE# NEXT_CHANGE#
---------- ---------- ---------- ---------------- ------------- ------------
1 79 4 CURRENT 1603673 281474976710
2 77 2 ACTIVE 1603650 1603669
3 78 2 ACTIVE 1603669 1603673
可见,只有在循环之后才有消除Invalid状态。接下来,我们讨论Invalid对Oracle启动过程的影响。
相关推荐
lklong 2020-11-22
oraclemch 2020-11-06
shilukun 2020-10-10
周嘉笙 2020-11-09
zhuzhufxz 2020-09-16
bfcady 2020-08-16
Hody 2020-08-16
FightFourEggs 2020-08-16
数据库设计 2020-08-16
yanghuatong 2020-08-16
dbasunny 2020-08-16
罗罗 2020-08-16
ihuaqiang 2020-08-16
choice0 2020-07-30
娜娜 2020-07-28
solarspot 2020-07-28
踩风火轮的乌龟 2020-07-26
娜娜 2020-07-20
xwb 2020-07-19
娜娜 2020-07-18
流云追风 2020-07-04
dataminer 2020-06-25
娜娜 2020-06-22
zhangchaoming 2020-06-21