NFS driver

The Network File System (NFS) is a distributed file system protocol originally developed by Sun Microsystems in 1984. An NFS server exports one or more of its file systems, known as shares. An NFS client can mount these exported shares on its own file system. You can perform file actions on this mounted remote file system as if the file system were local.

How the NFS driver works

The NFS driver, and other drivers based on it, work quite differently than a traditional block storage driver.

The NFS driver does not actually allow an instance to access a storage device at the block level. Instead, files are created on an NFS share and mapped to instances, which emulates a block device. This works in a similar way to QEMU, which stores instances in the /var/lib/nova/instances directory.

How to use the NFS driver

Creating an NFS server is outside the scope of this document.

Configure with one NFS server

This example assumes access to the following NFS server and mount point:


This example demonstrates the usage of this driver with one NFS server.

Set the nas_host option to the IP address or host name of your NFS server, and the nas_share_path option to the NFS export path:

nas_host =
nas_share_path = /storage

Configure with multiple NFS servers


You can use the multiple NFS servers with cinder multi back ends feature. Configure the enabled_backends option with multiple values, and use the nas_host and nas_share options for each back end as described above.

The below example is another method to use multiple NFS servers, and demonstrates the usage of this driver with multiple NFS servers. Multiple servers are not required. One is usually enough.

This example assumes access to the following NFS servers and mount points:




  1. Add your list of NFS servers to the file you specified with the nfs_shares_config option. For example, if the value of this option was set to /etc/cinder/shares.txt file, then:

    # cat /etc/cinder/shares.txt

    Comments are allowed in this file. They begin with a #.

  2. Configure the nfs_mount_point_base option. This is a directory where cinder-volume mounts all NFS shares stored in the shares.txt file. For this example, /var/lib/cinder/nfs is used. You can, of course, use the default value of $state_path/mnt.

  3. Start the cinder-volume service. /var/lib/cinder/nfs should now contain a directory for each NFS share specified in the shares.txt file. The name of each directory is a hashed name:

    # ls /var/lib/cinder/nfs/
  4. You can now create volumes as you normally would:

    $ openstack volume create --size 5 MYVOLUME
    # ls /var/lib/cinder/nfs/46c5db75dc3a3a50a10bfd1a456a9f3f

This volume can also be attached and deleted just like other volumes.

NFS driver notes

  • cinder-volume manages the mounting of the NFS shares as well as volume creation on the shares. Keep this in mind when planning your OpenStack architecture. If you have one master NFS server, it might make sense to only have one cinder-volume service to handle all requests to that NFS server. However, if that single server is unable to handle all requests, more than one cinder-volume service is needed as well as potentially more than one NFS server.

  • Because data is stored in a file and not actually on a block storage device, you might not see the same IO performance as you would with a traditional block storage driver. Please test accordingly.

  • Despite possible IO performance loss, having volume data stored in a file might be beneficial. For example, backing up volumes can be as easy as copying the volume files.


Regular IO flushing and syncing still stands.