SHUTDOWN IMMEDIATE长时间不能完成的问题
4月24日预定下班后更换外网网络交换模块，更换外网交换模块前按计划预先关闭外网的数据库和业务中间件，避免系统受到网络中断的影响。主要关闭过程为，首先关闭了业务中间件，然后连接数据库，中止仍然连接的会话，再使用SQLPLUS登录数据库，分别对RAC库下的A库和B库执行SHUTDOWN IMMDIATE，A库立即切换到DISMOUNT状态，但再不往下走了，等待超过1小时仍没有结果。查看alert日志，A库出现Trace dumping is performing id 的信息，查看cdmp文件夹下的trw文件，没有发现异常信息。使用TOPAS查看系统资源使用情况，ora_lms1进程占用CPU 16.7%，最新的TRC文件正是ora_lms1进程的，查看其内容，包含error 4031 detected信息。该错误表明ORACLE在分配和回收内存时遇到问题，有可能为ORACLE的BUG，也有可能确实有SGA不足的情况。强行KILL掉ora_lms1进程后观察数分钟，SHUTDOWN仍无进展，执行SHUTDOWN ABORT，A库关闭，INSTANCE SHUTDOWN，之后B库的SHUTDOWN IMMEDIATE也自动恢复，直到INSTANCE SHUTDOWN。更换完外网交换模块后，重新启动数据库和中间件，一切正常。
关闭数据库的注意事项：
Stopping Oracle Users Before Database Shutdown
The command SHUTDOWN IMMEDIATE may hang if you have persisting connections in the environment; for example, if you are running processes associated with monitoring tools such as OEM Agent. In that case, it is advisable to terminate the connections from the operating system level by issuing the following command:
"$ stop/id="
SQL> select sid,serial#,process from v$session;
SQL>host stop/id=20C0018B
You are strongly discouraged from stopping the sessions within ORACLE by using the ALTER SYSTEM KILL SESSION command followed by a stop/id on the operating system level.
When you issue an ALTER SYSTEM KILL SESSION command, it marks the process for deletion by PMON. If you then kill the process before PMON can get to it, confusions results and a clean process deletion does not occur. The deleted process appears to still be connected. Thus, the SHUTDOWN IMMEDIATE hangs and the partially dead process can't respond to the logoff command issued by the SHUTDOWN.
Use either ALTER SESSION KILL SESSION or HOST STOP/ID=. Don't use both.
De-installing Shareable Images
After shutting down all Oracle9i instances on a node, if you intend to do a machine shutdown or code relink, remove the shareable images by issuing the following command:
$ REMORACLE
4月24日预定下班后更换外网网络交换模块，更换外网交换模块前按计划预先关闭外网的数据库和业务中间件，避免系统受到网络中断的影响。主要关闭过程为，首先关闭了业务中间件，然后连接数据库，中止仍然连接的会话，再使用SQLPLUS登录数据库，分别对RAC库下的A库和B库执行SHUTDOWN IMMDIATE，A库立即切换到DISMOUNT状态，但再不往下走了，等待超过1小时仍没有结果。查看alert日志，A库出现Trace dumping is performing id 的信息，查看cdmp文件夹下的trw文件，没有发现异常信息。使用TOPAS查看系统资源使用情况，ora_lms1进程占用CPU 16.7%，最新的TRC文件正是ora_lms1进程的，查看其内容，包含error 4031 detected信息。该错误表明ORACLE在分配和回收内存时遇到问题，有可能为ORACLE的BUG，也有可能确实有SGA不足的情况。强行KILL掉ora_lms1进程后观察数分钟，SHUTDOWN仍无进展，执行SHUTDOWN ABORT，A库关闭，INSTANCE SHUTDOWN，之后B库的SHUTDOWN IMMEDIATE也自动恢复，直到INSTANCE SHUTDOWN。更换完外网交换模块后，重新启动数据库和中间件，一切正常。
关闭数据库的注意事项：
Stopping Oracle Users Before Database Shutdown
The command SHUTDOWN IMMEDIATE may hang if you have persisting connections in the environment; for example, if you are running processes associated with monitoring tools such as OEM Agent. In that case, it is advisable to terminate the connections from the operating system level by issuing the following command:
"$ stop/id="
SQL> select sid,serial#,process from v$session;
SQL>host stop/id=20C0018B
You are strongly discouraged from stopping the sessions within ORACLE by using the ALTER SYSTEM KILL SESSION command followed by a stop/id on the operating system level.
When you issue an ALTER SYSTEM KILL SESSION command, it marks the process for deletion by PMON. If you then kill the process before PMON can get to it, confusions results and a clean process deletion does not occur. The deleted process appears to still be connected. Thus, the SHUTDOWN IMMEDIATE hangs and the partially dead process can't respond to the logoff command issued by the SHUTDOWN.
Use either ALTER SESSION KILL SESSION or HOST STOP/ID=. Don't use both.
De-installing Shareable Images
After shutting down all Oracle9i instances on a node, if you intend to do a machine shutdown or code relink, remove the shareable images by issuing the following command:
$ REMORACLE
来自 “ ITPUB博客 ” ，链接：http://blog.itpub.net/10096205/viewspace-700712/，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/10096205/viewspace-700712/
扫一扫
2万+
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
