AIX5 RAC另一节点重启后 单一节点挂起

如题 现象如下

oracle@P570A1:[/oracle/app/oracle/crs/log/p570a1]$sqlplus/ as sysdba
 
SQL*Plus: Release 10.2.0.2.0 -Production on Sun Nov 30 11:52:30 2014
 
Copyright (c) 1982, 2005,Oracle.  All Rights Reserved.
 
NFS server 192.16.10.23 notresponding still trying

问题原因：

参见MOS文档 When NFS Server Is Down, Oracle Server Freezes With No Errors In Alert Log File (文档 ID 1316251.1)

摘录部分以便解释

Oracle code calls a Unix system call, 'getcwd' to get the current working directory. Then, after that, all the control reverts over to the operating system.  From what we can see, the function 'getcwd' calls 'getwd' which in turn calls 'stat'. Once 'stat' is entered it starts processing directory entries in the order shown below by performing a 'statx' call for each entry:

Oracle源代码调用UNIX方法'getcwd'来获取当前工作目录。
之后Oracle将控制权还到操作系统。根据上面的trace
'getcwd' 调用 'getwd'，'getwd'将调用'stat'.方法。
一旦调用'stat'，将使用'statx'方法获取每个文件夹的状态。
注：此时/下/archlog1目录不正常。
所以会使sqlplus hang

解决方法

将挂在根目录下的NFS文件系统挂在到下级目录