CNFS administration

There are some common CNFS administration tasks in this topic along with a sample configuration.

To query the current CNFS configuration, enter:

mmlscluster --cnfs
To temporarily disable CNFS on one or more nodes, enter:
mmchnode --cnfs-disable -N NodeList
Note: This operation affects only the high-availability aspects of the CNFS functionality. Normal NFS exporting of the data from the node is not affected. All currently defined CNFS IP addresses remain unchanged. There will be no automatic failover from or to this node in case of a failure. If failover is desired, GPFS™ should be shut down on the affected node prior to issuing the mmchnode command.
To re-enable previously-disabled CNFS member nodes, enter:
mmchnode --cnfs-enable -N NodeList
Note: If the GPFS daemon is running on a node on which CNFS is being re-enabled, the node will try to activate its CNFS IP address. If the IP address was currently on some other CNFS-enabled node, that activation would include a takeover.
To permanently remove nodes from the CNFS cluster, enter:
mmchnode --cnfs-interface=DELETE -N NodeList
Note: This operation affects only the high-availability aspects of the CNFS functionality. Normal NFS exporting of the data from the node is not affected. All currently defined CNFS IP addresses remain unchanged. There will be no automatic failover from or to this node in case of a failure. If failover is desired, GPFS should be shut down on the affected node prior to issuing the mmchnode command.

A sample CNFS configuration

Here is a CNFS configuration example, which assumes the following:
  • Your GPFS cluster contains three nodes: fin18, fin19, and fin20
  • The host names for NFS serving are: fin18nfs, fin19nfs, and fin20nfs
To define a CNFS cluster made up of these nodes, follow these steps:
  1. Add the desired GPFS file systems to /etc/exports on each of the nodes.
  2. Create a directory called ha in one of the GPFS file systems by entering:
    mkdir /gpfs/fs1/ha
  3. Create a temporary file called /tmp/hanfs-list, which contains the following lines:
    fin18 --cnfs-interface=fin18nfs
    fin19 --cnfs-interface=fin19nfs
    fin20 --cnfs-interface=fin20nfs
  4. Set the CNFS shared directory by entering:
    mmchconfig cnfsSharedRoot=/gpfs/fs1/ha
  5. Create the CNFS cluster with the mmchnode command, by entering:
    mmchnode -S /tmp/hanfs-list
  6. Access the exported GPFS file systems over NFS. If one or more GPFS nodes fail, the NFS clients should continue uninterrupted.