In a high availability environment such as RMS, an exported file system must be able to failover transparently when its server node is taken out of service: clients that mounted the file system before the failover should experience no access problems after the failover. NFS file systems require special preparation to achieve this result.
When a client mounts a remote NFS file system, it creates an internal file handle that it uses for future operations with the file system. To comply with NFS architecture, the client file handle includes the server's major and minor device numbers for the file system. This design can create access problems in the RMS environment. If the file system goes offline on the original server, and then comes back online on a second server that assigns different major and minor device numbers, the file handle will no longer be valid. This condition is called a stale file handle. The solution is to assign the same major and minor device numbers to the file system on every NFS server that may advertise that file system.
The above discussion refers to file systems in general, but in a high availability environment, the file system will actually be a shared disk volume that is accessible from any node that will export it. Preparing a shared disk volume with the same major and minor device number may require changes in the hardware or software configuration. If the shared disk volume is built on top of volume management software, additional steps may be necessary when the volume manager is installed.
This section provides some tips for preparing volume managers for use as NFS servers in the RMS environment.
