Skip to main content

NFS

Network File System is a filesystem protocol built upon the Open Network Computing Remote Procedure Call system (ONC RPC). RPCs are managed by the portmap service.

Server Configuration​

The main configuration is done in /etc/exports file. The file has:

  • A directory to share.
  • Host with mount options.

The server can be reloaded using:

exportfs -ra

The syntax of the exports file is:

# <DIR> <HOST OR NETWORK> [<OPTIONS>]

/srv/nfs 192.168.122.0/24(rw,sync,root_squash)

Client Configuration​

The NFS client mounts the remote filesystem onto the local system. There a few important commands:

# queries mount daemon on the remote server for information including shares that are available for mounting
showmount -e server

# daemon is a dynamic port mapping daemon designed to reduce usage of well-known port
portmap


# mount command has the filesystem type NFS, which links to the mount.nfs command. There are two formats of the mount command for NFS shares.
# - One option for mounting NFS shares is mount HOST:/export /mount-point where the host:/export portion causes the mount command to process this mount as NFS.
# - The other form of the mount command is mount -t NFS HOST:/export /mountpoint which specifies which NFS is being used.
# mount SERVER:/share /mnt/share

Security Considerations​

The NFS default security is to use the UNIX UID and GID. The challenge of using UID/GID on different systems is the values must match. User Bib with UID 1000 on system A must have the same UID on system B, or the wrong information may be accessed. Having a single sign on system with network available information like NIS, LDAP, or Kerberos will remove the UID confusion issues. NFS4 is Kerberos-aware and an excellent option to eliminate the UID/GID mapping issues.

The root_squash option /etc/exports file translates the root user's UID/GID (0) to an anonymous UID/GID. This is on by default and prevents root level access to the shared files. You should not disable root_squash without a good reason.

Performance Considerations​

Many factors contribute to the speed and performance of an NFS server or client.

Properly setting the values of rsize and wsize will allow for greater speed in a file transfer. However, you can only reasonably increase the block size to the MTU of your network between client and server. Increasing the frame size (Jumbo Frames) is one option.

Moving from a 1G to a 10G Ethernet network would vastly speed up an NFS setup.

The asynchronous mode trades speed for lack of robustness. An unclean shutdown of a server or client operating in an asynchronous mode has the potential to corrupt the data.