CNFS setup

You can set up a clustered NFS environment within a GPFS cluster.

To do this, follow these steps:

Designate a separate directory for the CNFS shared files:
```
mmchconfig cnfsSharedRoot=directory
```
where:

cnfsSharedRoot=directory

Is the path name to a GPFS directory, preferably on a small separate file system that is not exported by NFS. The GPFS file system that contains the directory must be configured to be mounted automatically upon GPFS start on each of the CNFS nodes (-A yes option on the mmchfs command). cnfsSharedRoot is a mandatory parameter and must be defined first. This directory must be readable by the rpcuser ID.
Add all GPFS file systems that need to be exported to /etc/exports. For NFS export considerations, see Exporting a GPFS file system using NFS.
If the shared directory from step 1 is in an exported file system, restrict access to that directory.
Use the mmchnode command to add nodes to the CNFS cluster:
```
mmchnode --cnfs-interface=ip_address_list -N node
```
where:

ip_address_list

Is a comma-separated list of host names or IP addresses to be used for GPFS cluster NFS serving.

node

Identifies a GPFS node to be added to the CNFS cluster.

For more information, see mmchnode command.
Use the mmchconfig command to configure the optional CNFS parameters.
cnfsMountdPort=mountd_port

Specifies the port number to be used for the rpc.mountd daemon.

For CNFS to work correctly with the automounter (AMD), the rpc.mountd daemon on the different nodes must be bound to the same port.

cnfsNFSDprocs=nfsd_procs

Specifies the number of nfsd kernel threads. The default is 32.

cnfsVersions=nfs_versions
Specifies a comma-separated list of protocol versions that CNFS should start and monitor. The default is 3,4. If you are not using NFS v3 and NFS v4, specify this parameter with the appropriate values for your configuration.
Note: If you are not using NFS v3 and NFS v4, and you do not explicitly specify cnfsVersions with the protocol versions on your system, the following message will continually appear in the mmfs.log:
```
Found NFS version mismatch between CNFS and current running config, check the OS config files.
```
If multiple failover groups are desired, assign a group ID to each NFS node:
```
mmchnode --cnfs-groupid=groupid -N node
```
To assign NFS nodes to different groups, use a group ID that is in a different range of ten. For example, a node with group ID 2n will fail over only to nodes in the same range of ten (which means any node with group ID 20 to 29). Failover in the same group will first look for one of the nodes with the same group ID. If none are found, any node in the group range starting at n0 to n9 is selected.