Persistent network configuration on Ubuntu — netplan YAML files, static addresses and DHCP, applying safely.
sudo netplan generate
check syntax and build the backend config
sudo netplan try
apply with an automatic rollback after 120s
sudo netplan apply
apply right away
cat /etc/netplan/*.yaml
what the files currently describe
ip commands change the network temporarily, until reboot. Persistent
configuration on Ubuntu Server is set through netplan: YAML files in
/etc/netplan/, which netplan translates into the backend's configuration —
on servers this is usually systemd-networkd.
netplan get doesn't print the current network state — it prints the
merged configuration from the files in /etc/netplan/, i.e. what's
supposed to be applied. Check the actual state of interfaces and addresses
separately: ip -br a, ip -br l, networkctl status.
DHCP on an interface:
A static address:
YAML is sensitive to indentation: spaces only, no tabs, and the nesting level must be consistent throughout the file.
Over SSH on a remote machine, change network configuration with
netplan try: if the new config breaks the connection, after 120 seconds
without confirmation netplan restores the previous file and access comes
back. netplan apply gives you no such safety net — a mistake in an
address or gateway can cut off the only channel to the machine.
All the files are applied together, in name order (01-..., 50-...); a
later file can override an earlier one's settings for the same interface.
Keep one interface in one file, so you don't have to hunt for a conflict
across several.
A route added with ip route add lasts until a reboot. A permanent one goes into
routes: of the interface the neighbour is reachable through: to is the destination
network, via is the neighbour's address on the shared network. The default route
(to: default) stays as it was; the new entry sits next to it:
ip route get shows which way the kernel will send a packet to an address: if the
answer has the neighbour as via, the route works. The via address must be on the
interface's own network, or the route will not be installed. More on the routing table
in ip.
It was added with ip route add, and it is not in the files under /etc/netplan/.
Write the route into the interface's routes: and apply the configuration.
There's an error in the config: a wrong address, gateway, or indentation.
Go through the hypervisor's console, fix the file, and apply it with
netplan try, not apply. If access was lost specifically because of
apply without a rollback, switch to try for every later network change
on this machine.
Check the syntax (sudo netplan generate will show the error and the
file), whether you edited the right file, and whether the interface name
(ip -br a) matches what's in the YAML.
Add a nameservers: addresses: [...] block to the interface's config.
Check which servers are actually in use: resolvectl status (see
dns-resolv).
netplan get only reads the files in /etc/netplan/, not the actual state
of the interfaces. If there's a big gap after manual ip commands or a
failed apply, cross-check against ip addr and networkctl status, bring
the files in line, and apply them again.
Found an inaccuracy? Report an error in the article