奇怪的RAC归档，求解！

显示全部楼层 · 2013-1-30 22:13:19

数据库：10.2.0.1
二节点RAC，INSTANCE:RACDB1，RACDB2
关闭RACDB2后，手动切换RACDB1的日志，为什么会产生RACDB2的归档？很奇怪，求解！
手动切换：ALTER SYSTEM SWITCH LOGFILE;
也查看过ps -ef|grep oracle，确实没开启RACDB2的实例。

实例情况：
SQL> SELECT STATUS,INSTANCE FROM V$THREAD;
STATUS
------
INSTANCE
--------------------------------------------------------------------------------
OPEN
RACDB1
CLOSED
RACDB2

归档记录：
ASMCMD> ls
thread_1_seq_41.300.795007331
thread_1_seq_42.299.795007425
thread_1_seq_43.297.795007427
thread_1_seq_44.296.795007433
thread_1_seq_45.294.795007517
thread_1_seq_46.295.795007519
thread_1_seq_47.292.795007525
thread_2_seq_20.298.795007427
thread_2_seq_21.293.795007519
thread_2_seq_22.291.795007527
ASMCMD> ls
thread_1_seq_41.300.795007331
thread_1_seq_42.299.795007425
thread_1_seq_43.297.795007427
thread_1_seq_44.296.795007433
thread_1_seq_45.294.795007517
thread_1_seq_46.295.795007519
thread_1_seq_47.292.795007525
thread_1_seq_48.289.795007545
thread_1_seq_49.287.795007545
thread_1_seq_50.285.795007551
thread_1_seq_51.286.795007555
thread_2_seq_20.298.795007427
thread_2_seq_21.293.795007519
thread_2_seq_22.291.795007527
thread_2_seq_23.288.795007547
thread_2_seq_24.284.795007555
ASMCMD> pwd
+rac_disk/racdb/archivelog/2012_09_26

千问 · 2013-1-30 22:13:19

学习了

千问 · 2013-1-30 22:13:19

学习了

千问 · 2013-1-30 22:13:19

tanjiaping 发表于 2012-9-25 23:05
学习了
期待高人！

千问 · 2013-1-30 22:13:19

Can this be reproduced? Can you show us the exact commands and output (text, not image)? Like the following:
select instance_name from v$instance;
select thread#, status, enabled, instance from v$thread;
select max(sequence#) from v$archived_log where thread# = 2; -- assuming instance 2 runs thread 2
select group#, thread#, sequence#, status from v$log;
alter system switch logfile;
select max(sequence#) from v$archived_log where thread# = 2;
select group#, thread#, sequence#, status from v$log;

千问 · 2013-1-30 22:13:19

节点2的归档不是由节点2产生的，是由节点上的Thead 2产生的。
节点2关闭后，Thread 2就接管了原本在节点2上的Redo，但节点2上的Redo将不再包含任何内容。
不过，RAC要求节点1的所有Redo使用一圈后，节点2一定也要切换一次，这次切换，就由节点1上的Thread 2完成。
这是我大概的理解，没做实验验证。

千问 · 2013-1-30 22:13:19

Yong Huang 发表于 2012-9-26 02:08
Can this be reproduced? Can you show us the exact commands and output (text, not image)? Like the fo ...
如下是在RACDB1环境下运行的：

SQL> select instance_name from v$instance;
INSTANCE_NAME
----------------
RACDB1
SQL> select thread#, status, enabled, instance from v$thread;
THREAD# STATUS ENABLEDINSTANCE
---------- ------ -------- ------------------------------
1 OPEN PUBLIC RACDB1
2 CLOSED PUBLIC RACDB2
SQL> select max(sequence#) from v$archived_log where thread# = 2;
MAX(SEQUENCE#)
--------------

23
SQL> select group#, thread#, sequence#, status from v$log;
GROUP#THREAD#SEQUENCE# STATUS
---------- ---------- ---------- ----------------
1
1 45 INACTIVE
2
1 46 CURRENT
3
2 23 INACTIVE
4
2 24 CURRENT
SQL> alter system switch logfile;
System altered.
SQL> alter system switch logfile;
System altered.
SQL> alter system switch logfile;
System altered.
SQL> select max(sequence#) from v$archived_log where thread# = 2;
MAX(SEQUENCE#)
--------------

25
SQL> select group#, thread#, sequence#, status from v$log;
GROUP#THREAD#SEQUENCE# STATUS
---------- ---------- ---------- ----------------
1
1 49 CURRENT
2
1 48 INACTIVE
3
2 25 INACTIVE
4
2 26 CURRENT

千问 · 2013-1-30 22:13:19

vage 发表于 2012-9-26 08:44
节点2的归档不是由节点2产生的，是由节点上的Thead 2产生的。
节点2关闭后，Thread 2就接管了原本在节点2上 ...
确实是节点1的REDO使用一圈，节点2才切换一次。
谢谢vage,学些了！

千问 · 2013-1-30 22:13:19

vage 发表于 2012-9-26 08:44
节点2的归档不是由节点2产生的，是由节点上的Thead 2产生的。
节点2关闭后，Thread 2就接管了原本在节点2上 ...

V哥威武

千问 · 2013-1-30 22:13:19

Kevin__Zhang 发表于 2012-9-26 10:57
V哥威武