Set Up an NFS File Server on Ubuntu 26.04

I needed a shared directory that two servers could both write to — a staging area for automated backups and sync jobs. NFS (Network File System) is the simplest way to do that on a LAN. It is built into the Linux kernel, requires no subscription, and exists on every Ubuntu install natively. Here is how I set up the server side and verified it actually works.

Ubuntu 26.04 · linux-nfs.org · libnfs on GitHub

Requirements

1. Two Ubuntu 26.04 machines on the same LAN (or local loopback for testing)
2. TCP ports 111 (portmap) and 2049 (nfsd) open on the server
3. sudo access on both machines

Step 1 — Install NFS Server

Install the NFS kernel server and common utilities.

sudo apt-get install -y nfs-kernel-server nfs-common
# User sessions running outdated binaries:
  # root @ session #95: php[232825]

  # No VM guests are running outdated hypervisor (qemu) binaries on this host.
  active

sudo apt-get install -y nfs-kernel-server nfs-common
# User sessions running outdated binaries:
  # root @ session #95: php[232825]

  # No VM guests are running outdated hypervisor (qemu) binaries on this host.
  active

Step 2 — Create and Export a Directory

Create the directory you want to share, then write its export rule into /etc/exports. Each line has the form export_path client_net(options, ...).

sudo mkdir -p /srv/nfs/projects
sudo chown root:root /srv/nfs/projects
sudo chmod 2775 /srv/nfs/projects
echo '/srv/nfs/projects 192.168.1.0/24(rw,sync,no_subtree_check,root_squash)' > /tmp/exports
sudo cp /tmp/exports /etc/exports
sudo exportfs -rav
exporting 192.168.1.0/24:/srv/nfs/projects

sudo mkdir -p /srv/nfs/projects
sudo chown root:root /srv/nfs/projects
sudo chmod 2775 /srv/nfs/projects
echo '/srv/nfs/projects 192.168.1.0/24(rw,sync,no_subtree_check,root_squash)' > /tmp/exports
sudo cp /tmp/exports /etc/exports
sudo exportfs -rav
exporting 192.168.1.0/24:/srv/nfs/projects

Verify the export is live:

exportfs -v

exportfs -v

The output shows 192.168.1.0/24 with rw,sync,root_squash,no_subtree_check. Reference the diagram below for the request flow through the stack.

Step 3 — Mount from a Client

On the client machine:

sudo apt-get install -y nfs-common          # install NFS client
sudo mkdir -p /mnt/nfs
sudo mount -t nfs -o vers=4.2 server-ip:/srv/nfs/projects /mnt/nfs
df -hT /mnt/nfs
Filesystem                  Type  Size  Used Avail Use% Mounted on
127.0.0.1:/srv/nfs/projects nfs4   39G  7.8G   31G  21% /mnt/nfs

sudo apt-get install -y nfs-common          # install NFS client
sudo mkdir -p /mnt/nfs
sudo mount -t nfs -o vers=4.2 server-ip:/srv/nfs/projects /mnt/nfs
df -hT /mnt/nfs
Filesystem                  Type  Size  Used Avail Use% Mounted on
127.0.0.1:/srv/nfs/projects nfs4   39G  7.8G   31G  21% /mnt/nfs

Check the mount options at runtime:

mount | grep /mnt/nfs
127.0.0.1:/srv/nfs/projects on /mnt/nfs type nfs4 (rw,relatime,vers=4.2,rsize=524288,wsize=524288,namlen=255,hard,fatal_neterrors=none,proto=tcp,timeo=600,retrans=2,sec=sys,clientaddr=127.0.0.1,local_lock=none,addr=127.0.0.1)

mount | grep /mnt/nfs
127.0.0.1:/srv/nfs/projects on /mnt/nfs type nfs4 (rw,relatime,vers=4.2,rsize=524288,wsize=524288,namlen=255,hard,fatal_neterrors=none,proto=tcp,timeo=600,retrans=2,sec=sys,clientaddr=127.0.0.1,local_lock=none,addr=127.0.0.1)

Step 4 — Test Writes and Understand root_squash

NFS maps the client root to nobody:nogroup by default. This is called root_squash and is the secure default — no client can bypass the server’s file permissions by connecting as root.

# write a file via NFS as root (squashed to nobody)
echo 'some config' | sudo tee /mnt/nfs/squash-test.txt
ls -n /mnt/nfs/squash-test.txt
-rw-r--r-- 1 nobody nogroup 12 Nov 14 11:04 /mnt/nfs/squash-test.txt

# write a file via NFS as root (squashed to nobody)
echo 'some config' | sudo tee /mnt/nfs/squash-test.txt
ls -n /mnt/nfs/squash-test.txt
-rw-r--r-- 1 nobody nogroup 12 Nov 14 11:04 /mnt/nfs/squash-test.txt

A throughput test over the loopback mount (your LAN will be slower):

dd if=/dev/zero of=/mnt/nfs/dd-test bs=1M count=256 conv=fdatasync
268435456 bytes (268 MB, 256 MiB) copied, 2.2447 s, 120 MB/s

dd if=/dev/zero of=/mnt/nfs/dd-test bs=1M count=256 conv=fdatasync
268435456 bytes (268 MB, 256 MiB) copied, 2.2447 s, 120 MB/s

Step 5 — Persistent Mount via /etc/fstab

Add an entry to /etc/fstab and test it with mount -a.

# add to /etc/fstab
echo "server-ip:/srv/nfs/projects /mnt/nfs nfs defaults,_netdev 0 0" | sudo tee -a /etc/fstab
sudo mount -a && mount | grep /mnt/nfs
127.0.0.1:/srv/nfs/projects on /mnt/nfs type nfs4 (rw,relatime,vers=4.2,...,_netdev)

# add to /etc/fstab
echo "server-ip:/srv/nfs/projects /mnt/nfs nfs defaults,_netdev 0 0" | sudo tee -a /etc/fstab
sudo mount -a && mount | grep /mnt/nfs
127.0.0.1:/srv/nfs/projects on /mnt/nfs type nfs4 (rw,relatime,vers=4.2,...,_netdev)

Debugging Note — root_squash Permission Denied

On my first try, writes failed with Permission denied despite running as root. The export used root_squash (the sensible default), which maps the client’s root account to nobody:nogroup. But the export directory /srv/nfs/projects was owned by root:root with mode 2775 — group nogroup had no write permission. NFS squashes before the filesystem permission check. The fix: give the directory a group that nobody belongs to (nogroup) and enable the setgid bit so newly created files inherit the group.

sudo chgrp nogroup /srv/nfs/projects
sudo chmod 2775 /srv/nfs/projects
# Now writes via NFS succeed

sudo chgrp nogroup /srv/nfs/projects
sudo chmod 2775 /srv/nfs/projects
# Now writes via NFS succeed

Wrapping Up

NFS gives you a zero-conf shared filesystem over TCP. The server is a single apt-get, the client is a mount command. If you need to move data between machines, pair NFS with rsync vs rclone for backups, secure it with nftables, and cap its CPU/memory with systemd resource limits. The flow diagram and the export options reference above should help you get the defaults right the first time.