RAC 出现“Connected to an idle instance.”问题。

整理网友错误现象，其实我在配置的时候，也出现过好几次，总结总结。

Connected to an idle instance

oracle@node2 bin]$ export oracle_sid=RACDB
[oracle@node2 bin]$ sqlplus /nolog

SQL*Plus: Release 10.2.0.1.0 - Production on Tue Jul 8 03:08:59 2008

Copyright (c) 1982, 2005, Oracle.  All rights reserved.

SQL> conn / as sysdba
Connected to an idle instance.
SQL>


但pl/sql中可以登陆。
select instance_name,host_name,archiver,thread#,status from gv$instance;
1        RACDB1        node2        STARTED        1        OPEN
2        RACDB2        node1        STARTED        2        OPEN

如果输入：会出现这样的报错：　

[oracle@node2 bin]$ export oracle_sid=RACDB2
[oracle@node2 bin]$ sqlplus /nolog
SQL*Plus: Release 10.2.0.1.0 - Production on Tue Jul 8 02:45:03 2008
Copyright (c) 1982, 2005, Oracle.  All rights reserved.

SQL> conn / as sysdba
Connected to an idle instance.
SQL> startup
ORA-01078: failure in processing system parameters
LRM-00109: could not open parameter file '/opt/ora10g/product/10.2.0/db_1/dbs/initRACDB2.ora'

解决的办法：　

[oracle@node2 bin]$ export oracle_sid=RACDB
[oracle@node2 bin]$ sqlplus /nolog
SQL*Plus: Release 10.2.0.1.0 - Production on Tue Jul 8 18:58:34 2008
Copyright (c) 1982, 2005, Oracle.  All rights reserved.

SQL> conn system/oracle@RACDB
Connected.
SQL> select instance_name from v$instance;
INSTANCE_NAME
----------------
RACDB1

其实只要

rac1-> srvctl status database -d devdb
能启动，说明数据库没问题．

据说这个是个BUG。

主题:  "Connected to an Idle Instance" Message when Connecting Bequeath to a Running Instance 
  文档 ID:  注释:435044.1 类型:  PROBLEM 
  上次修订日期:  23-APR-2008 状态:  PUBLISHED 

In this Document
  Symptoms
  Cause
  Solution
  References



--------------------------------------------------------------------------------



Applies to: 
Oracle Server - Enterprise Edition - Version: 8.0.6 to 11.1.0.6
This problem can occur on any platform.

Symptoms
While connecting bequeath to a running instance the following message appears : 

$ export ORACLE_SID=test 
$ sqlplus "/ as sysdba" 

SQL*Plus: Release 10.2.0.1.0 - Production on Wed Jun 6 12:17:14 2007 

Copyright (c) 1982, 2005, Oracle. All rights reserved. 

Connected to an idle instance. 

SQL> 
The "test" instance was running and this was verified with the ps command:

$ ps -ef | grep pmon 
oracle 9714 1 0 May28 ? 00:00:07 ora_pmon_test 
oracle 32635 1 0 May30 ? 00:00:01 ora_pmon_orcl115b 
Furthermore the test instance can be started up:

$ export ORACLE_SID=test 
$ sqlplus "/ as sysdba" 

SQL*Plus: Release 10.2.0.1.0 - Production on Wed Jun 6 12:17:14 2007 

Copyright (c) 1982, 2005, Oracle. All rights reserved. 

Connected to an idle instance. 

SQL> startup nomount 
ORACLE instance started. 

Total System Global Area 866291712 bytes 
Fixed Size 1302036 bytes 
Variable Size 239075820 bytes 
Database Buffers 620756992 bytes 
Redo Buffers 5156864 bytes 
SQL> 


The alert.log was immediately written with the informations pertaining to the startup of the operation. When a log switch was performed in the "real" test instance (within a SQL*NET connection)  then the alert.log was appended with the corresponding informations. So at that point on the system there were two instances "test" running from the same Oracle Home

$ ps -ef | grep pmon 
oracle 9714 1 0 May28 ? 00:00:07 ora_pmon_test 
oracle 32635 1 0 May30 ? 00:00:01 ora_pmon_orcl115b 
oracle 10382 1 0 12:27 ? 00:00:00 ora_pmon_test



Cause
The issue was caused by a wrong setting of the ORACLE_HOME environment variable used when starting up the instance. This can be easily discovered if one checks the values of the environment variables set within the background processes of the first "test"  instance started : 



$ ps -ef | grep pmon 
oracle 9714 1 0 May28 ? 00:00:07 ora_pmon_test 
oracle 32635 1 0 May30 ? 00:00:01 ora_pmon_orcl115b 
oracle 10382 1 0 12:27 ? 00:00:00 ora_pmon_test

cat /proc/9714/environ

HOSTNAME=ro-rac4 
SHELL=/bin/bash 
TERM=xterm 
HISTSIZE=1000 
USER=oracle 
LS_COLORS=no=00:fi=00:di=00;34:ln=00;36i=40;33:so=00;35:bd=40;33;01:cd=40;33;01:... 
ORACLE_SID=test 
MAIL=/var/spool/mail/oracle 
PATH=INPUTRC=/etc/inputrc 
PWD=/home/oracle 
LANG=en_US.UTF-8SSH_ASKPASS=/usr/libexec/openssh/gnome-ssh-askpass 
SHLVL=1 
HOME=/home/oracleLOGNAME=oracleLESSOPEN=|/usr/bin/lesspipe.sh %s 
ORACLE_HOME=/u01/app/oracle/product//10.2.0/db_1 
G_BROKEN_FILENAMES=1_=/u01/app/oracle/product/10.2.0/db_1/bin/sqlplus 
ORA_NET2_DESC=8,11ORACLE_SPAWNED_PROCESS=1 
SKGP_HIDDEN_ARGS=0

$ORACLE_HOME of the old test instance points to "ORACLE_HOME=/u01/app/oracle/product//10.2.0/db_1" . Note the double slash ("//".

whereas  $ORACLE_HOME of the new test instance points to "ORACLE_HOME=/u01/app/oracle/product/10.2.0/db_1"




ORACLE_SID and ORACLE_HOME are used to generate the IDs of the shared memory segments for the SGA. When using a different ORACLE_HOME value a different ID will get generated and since this does not match any of the existing segments' IDs it is assumed that this is a new idle instance. 

The above example shows how to get the environment variables for a process running on Linux. In order to get these variables for other platforms please read Note 373303.1-How to Check the Environment Variables for an Oracle Process 




Solution
Shutdown the instance and start it up from an OS session where all the environment variables are correctly set. If the instance is to be started via SRVCTL make sure that all the CRS configuration files are using a correct value for the ORACLE_HOME.

-OR-

If the running instance can't be shutdown set the ORACLE_HOME to the value found in the envrionment variables set of the running instance. 

References
Note 373303.1 - How to Check the Environment Variables for an Oracle Process

Keywords
START~INSTANCE; BACKGROUND~PROCESSES; ORACLE~PROCESS; NOMOUNT; SYSDBA; BEQUEATH; SRVCTL; START~INSTANCE; 
--------------------------------------------------------------------------------

Help us improve our service. Please email us your comments for this document. .

来自 “ ITPUB博客 ” ，链接：http://blog.itpub.net/8334342/viewspace-374541/，如需转载，请注明出处，否则将追究法律责任。

转载于:http://blog.itpub.net/8334342/viewspace-374541/