 | ||
| NFSRestart agent notes | ||
|---|---|---|
| Prev | File share agents | Next |
The NFSRestart agent has the following notes:
The high availability fire drill detects discrepancies between the VCS configuration and the underlying infrastructure on a node; discrepancies that might prevent a service group from going online on a specific node. For NFSRestart resources, the high availability drill performs the following, it:
For more information about using the high availability fire drill see the Veritas Cluster Server Administrator's Guide.
You must provide a fully qualified host name (nfsserver.example.edu) for the NFS server while mounting the file system on the NFS client. If you do not use a fully qualified host name, or if you use a virtual IP address (10.122.12.25) or partial host name (nfsserver), NFS lock recovery may fail.
If you want to use the virtual IP address or a partial host name, make the following changes to the service database (hosts) and the nsswitch.conf files:
To use the virtual IP address and partial host name for the NFS server, you need to add an entry to the /etc/hosts file. The virtual IP address and the partial host name should resolve to the fully qualified host name.
You should also modify the hosts entry in this file so that upon resolving a name locally, the host does not first contact NIS/DNS, but instead immediately returns a successful status. Changing the nsswitch.conf file might affect other services running on the system.
hosts: files [SUCCESS=return] dns nis
You have to make sure that the NFS client stores the same information for the NFS server as the client uses while mounting the file system. For example, if the NFS client mounts the file system using fully qualified domain names for the NFS server, then the NFS client directory: /var/lib/nfs/ directory should also have a fully qualified domain name after the acquisition of locks. Otherwise, you need to start and stop the NFS client twice using the /etc/init.d/nfs.client script to clear the lock cache of the NFS client.
A time period exists where the virtual IP address is online but locking services are not registered on the server. Any NFS client trying to acquire a lock in this interval would fail and get ENOLCK error.
Every two seconds, the smsyncd daemon copies the list of clients that hold the locks on the shared filesystem in the service group. If the service group fails before smsyncd has a chance to copy the client list, the clients may not get a notification once the service group is brought up. This causes NFS lock recovery failure.