Linux
Below are a collection of example netplan configurations for common scenarios 常见方案. If you see a scenario missing or have one to contribute 贡献, please file a bug 提交错误 against 针对 this documentation with the example using the links at the bottom of this page. Thank you!
Configuration
To configure netplan, save configuration files under /etc/netplan/ with a .yaml extension (e.g. /etc/netplan/config.yaml), then run sudo netplan apply. This command parses and applies the configuration to the system. Configuration written to disk under /etc/netplan/ will persist 持久化 between reboots.
Using DHCP and static addressing
To let 让 the interface named 'enp3s0' get an address via 通过 DHCP, create a YAML file with the following:
要让名为“enp3s0”的接口通过 DHCP 获取地址，请使用以下内容创建 YAML 文件：
network:
version: 2
renderer: networkd
ethernets:
enp3s0:
dhcp4: true
To instead 改为 set a static IP address, use the addresses key, which takes a list of (IPv4 or IPv6), addresses along with 以及 the subnet prefix length (e.g. /24). DNS information can be provided as well, and the gateway can be defined via a default route:
要改为设置静态 IP 地址，请使用addresses关键字，关键字值是一个包含地址以及子网前缀（例如 0.0.0.0/24）(IPv4 or IPv6)的列表。也可以提供 DNS 信息，并且可以通过默认路由定义网关：
network:
version: 2
renderer: networkd
ethernets:
enp3s0:
addresses:
- 10.10.10.2/24
nameservers:
search: [mydomain, otherdomain]
addresses: [10.10.10.1, 1.1.1.1]
routes:
- to: default
via: 10.10.10.1
Connecting multiple interfaces with DHCP
Many systems now include more than one network interface. Servers will commonly 通常 need to connect to multiple networks, and may require that traffic to 连接到 the Internet goes through a specific interface despite 尽管 all of them providing a valid gateway.
许多系统现在包含多个网络接口。服务器通常需要连接到多个网络，并且可能只通过特定的接口链接到英特网，尽管所有这些接口都提供有效的网关。
One can achieve 实现 the exact 确切的 routing desired 期望 over DHCP by specifying a metric 指标 for the routes retrieved 检索 over DHCP, which will ensure some routes are preferred over others. In this example, 'enred' is preferred over 'engreen', as it has a lower route metric:
可以通过为通过 DHCP 检索的路由指定指标来实现所需的 DHCP 上的确切路由，这将确保某些路由优先于其他路由。在此示例中，“enred”优先于“engreen”，因为它的路由指标较低：
network:
version: 2
ethernets:
enred:
dhcp4: yes
dhcp4-overrides:
route-metric: 100
engreen:
dhcp4: yes
dhcp4-overrides:
route-metric: 200
Connecting to an open wireless network
Netplan easily supports connecting to an open wireless network (one that 一种 is not secured by a password), only requiring that the access point is defined:
network:
version: 2
wifis:
wl0:
access-points:
opennetwork: {}
dhcp4: yes
Connecting to a WPA Personal wireless network
Wireless devices use the 'wifis' key and share the same configuration options with wired ethernet devices. The wireless access point name and password should also be specified:
network:
version: 2
renderer: networkd
wifis:
wlp2s0b1:
dhcp4: no
dhcp6: no
addresses: [192.168.0.21/24]
nameservers:
addresses: [192.168.0.1, 8.8.8.8]
access-points:
"network_ssid_name":
password: "**********"
routes:
- to: default
via: 192.168.0.1
Connecting to WPA Enterprise wireless networks
It is also common to find wireless networks secured using WPA or WPA2 Enterprise, which requires additional authentication parameters.
For example, if the network is secured using WPA-EAP and TTLS:
network:
version: 2
wifis:
wl0:
access-points:
workplace:
auth:
key-management: eap
method: ttls
anonymous-identity: "@internal.example.com"
identity: "joe@internal.example.com"
password: "v3ryS3kr1t"
dhcp4: yes
Or, if the network is secured using WPA-EAP and TLS:
network:
version: 2
wifis:
wl0:
access-points:
university:
auth:
key-management: eap
method: tls
anonymous-identity: "@cust.example.com"
identity: "cert-joe@cust.example.com"
ca-certificate: /etc/ssl/cust-cacrt.pem
client-certificate: /etc/ssl/cust-crt.pem
client-key: /etc/ssl/cust-key.pem
client-key-password: "d3cryptPr1v4t3K3y"
dhcp4: yes
Many different modes of encryption are supported. See the Netplan reference page.
Using multiple addresses on a single interface
The addresses key can take a list of addresses to assign to an interface:
network:
version: 2
renderer: networkd
ethernets:
enp3s0:
addresses:
- 10.100.1.37/24
- 10.100.1.38/24:
label: enp3s0:0
- 10.100.1.39/24:
label: enp3s0:some-label
routes:
- to: default
via: 10.100.1.1
Using multiple addresses with multiple gateways
Similar to 类似 the example above, interfaces with multiple addresses can be configured with multiple gateways.
network:
version: 2
renderer: networkd
ethernets:
enp3s0:
addresses:
- 10.0.0.10/24
- 11.0.0.11/24
routes:
- to: default
via: 10.0.0.1
metric: 200
- to: default
via: 11.0.0.1
metric: 300
We configure individual 个别 routes to default (or 0.0.0.0/0) using the address of the gateway for the subnet. The metric value should be adjusted so the routing happens as expected.
我们使用子网网关的地址将各个路由配置为默认值（或 0.0.0.0/0）。应调整指标值，以便按预期进行路由。
DHCP can be used to receive one of the IP addresses for the interface. In this case, the default route for that address will be automatically configured with a metric value of 100.
Using Network Manager as a renderer
Netplan supports both networkd and Network Manager as backends. You can specify which network backend should be used to configure particular 特定devices by using the renderer key. You can also 还可以 delegate 委托 all configuration of the network to Network Manager itself by specifying only the renderer key:
network:
version: 2
renderer: NetworkManager
Configuring interface bonding
Bonding is configured by declaring a bond interface with a list of physical interfaces and a bonding mode. Below is an example of an active-backup bond that uses DHCP to obtain an address:
network:
version: 2
renderer: networkd
bonds:
bond0:
dhcp4: yes
interfaces:
- enp3s0
- enp4s0
parameters:
mode: active-backup
primary: enp3s0
Below is an example of a system acting as a router with various bonded interfaces and different types. Note the 'optional: true' key declarations that allow booting to occur without waiting for those interfaces to activate fully.
network:
version: 2
renderer: networkd
ethernets:
enp1s0:
dhcp4: no
enp2s0:
dhcp4: no
enp3s0:
dhcp4: no
optional: true
enp4s0:
dhcp4: no
optional: true
enp5s0:
dhcp4: no
optional: true
enp6s0:
dhcp4: no
optional: true
bonds:
bond-lan:
interfaces: [enp2s0, enp3s0]
addresses: [192.168.93.2/24]
parameters:
mode: 802.3ad
mii-monitor-interval: 1
bond-wan:
interfaces: [enp1s0, enp4s0]
addresses: [192.168.1.252/24]
nameservers:
search: [local]
addresses: [8.8.8.8, 8.8.4.4]
parameters:
mode: active-backup
mii-monitor-interval: 1
gratuitious-arp: 5
routes:
- to: default
via: 192.168.1.1
bond-conntrack:
interfaces: [enp5s0, enp6s0]
addresses: [192.168.254.2/24]
parameters:
mode: balance-rr
mii-monitor-interval: 1
Configuring network bridges
To create a very simple bridge consisting of a single device that uses DHCP, write:
network:
version: 2
renderer: networkd
ethernets:
enp3s0:
dhcp4: no
bridges:
br0:
dhcp4: yes
interfaces:
- enp3s0
A more complex example, to get libvirtd to use a specific bridge with a tagged vlan, while continuing to provide an untagged interface as well would involve:
network:
version: 2
renderer: networkd
ethernets:
enp0s25:
dhcp4: true
bridges:
br0:
addresses: [ 10.3.99.25/24 ]
interfaces: [ vlan15 ]
vlans:
vlan15:
accept-ra: no
id: 15
link: enp0s25
Then libvirtd would be configured to use this bridge by adding the following content to a new XML file under /etc/libvirtd/qemu/networks/. The name of the bridge in the <bridge> tag as well as in <name> need to match the name of the bridge device configured using netplan:xml br0
Attaching VLANs to network interfaces
To configure multiple VLANs with renamed interfaces:
network:
version: 2
renderer: networkd
ethernets:
mainif:
match:
macaddress: "de:ad:be:ef:ca:fe"
set-name: mainif
addresses: [ "10.3.0.5/23" ]
nameservers:
addresses: [ "8.8.8.8", "8.8.4.4" ]
search: [ example.com ]
routes:
- to: default
via: 10.3.0.1
vlans:
vlan15:
id: 15
link: mainif
addresses: [ "10.3.99.5/24" ]
vlan10:
id: 10
link: mainif
addresses: [ "10.3.98.5/24" ]
nameservers:
addresses: [ "127.0.0.1" ]
search: [ domain1.example.com, domain2.example.com ]
Reaching a directly connected gateway
This allows setting up a default route, or any route, using the "on-link" keyword where the gateway is an IP address that is directly connected to the network even if the address does not match the subnet configured on the interface.
network:
version: 2
renderer: networkd
ethernets:
ens3:
addresses: [ "10.10.10.1/24" ]
routes:
- to: default # or 0.0.0.0/0
via: 9.9.9.9
on-link: true
For IPv6 the config would be very similar, with the notable difference being an additional scope: link host route to the router's address required:
network:
version: 2
renderer: networkd
ethernets:
ens3:
addresses: [ "2001:cafe:face:beef::dead:dead/64" ]
routes:
- to: "2001:cafe:face::1/128"
scope: link
- to: default # or "::/0"
via: "2001:cafe:face::1"
on-link: true
Configuring source routing
Route tables can be added to particular interfaces to allow routing between two networks:
In the example below, ens3 is on the 192.168.3.0/24 network and ens5 is on the 192.168.5.0/24 network. This enables clients on either network to connect to the other and allow the response to come from the correct interface.
Furthermore, the default route is still assigned to ens5 allowing any other traffic to go through it.
network:
version: 2
renderer: networkd
ethernets:
ens3:
addresses:
- 192.168.3.30/24
dhcp4: no
routes:
- to: 192.168.3.0/24
via: 192.168.3.1
table: 101
routing-policy:
- from: 192.168.3.0/24
table: 101
ens5:
addresses:
- 192.168.5.24/24
dhcp4: no
routes:
- to: default
via: 192.168.5.1
- to: 192.168.5.0/24
via: 192.168.5.1
table: 102
routing-policy:
- from: 192.168.5.0/24
table: 102
Configuring a loopback interface
Networkd does not allow creating new loopback devices, but a user can add new addresses to the standard loopback interface, lo, in order to have it considered a valid address on the machine as well as for custom routing:
network:
version: 2
renderer: networkd
ethernets:
lo:
addresses: [ "127.0.0.1/8", "::1/128", "7.7.7.7/32" ]
Integration with a Windows DHCP Server
For networks where DHCP is provided by a Windows Server using the dhcp-identifier key allows for interoperability:
network:
version: 2
ethernets:
enp3s0:
dhcp4: yes
dhcp-identifier: mac
Connecting an IP tunnel
Tunnels allow an administrator to extend networks across the Internet by configuring two endpoints that will connect a special tunnel interface and do the routing required. Netplan supports SIT, GRE, IP-in-IP (ipip, ipip6, ip6ip6), IP6GRE, VTI and VTI6 tunnels.
A common use of tunnels is to enable IPv6 connectivity on networks that only support IPv4. The example below show how such a tunnel might be configured.
Here, 1.1.1.1 is the client's own IP address; 2.2.2.2 is the remote server's IPv4 address, "2001:dead:beef::2/64" is the client's IPv6 address as defined by the tunnel, and "2001:dead:beef::1" is the remote server's IPv6 address.
Finally, "2001:cafe:face::1/64" is an address for the client within the routed IPv6 prefix:
network:
version: 2
ethernets:
eth0:
addresses:
- 1.1.1.1/24
- "2001:cafe:face::1/64"
routes:
- to: default
via: 1.1.1.254
tunnels:
he-ipv6:
mode: sit
remote: 2.2.2.2
local: 1.1.1.1
addresses:
- "2001:dead:beef::2/64"
routes:
- to: default
via: "2001:dead:beef::1"
Configuring SR-IOV Virtual Functions
For SR-IOV network cards, it is possible to dynamically allocate Virtual Function interfaces for every configured Physical Function. In netplan, a VF is defined by having a link: property pointing to the parent PF.
network:
version: 2
ethernets:
eno1:
mtu: 9000
enp1s16f1:
link: eno1
addresses : [ "10.15.98.25/24" ]
vf1:
match:
name: enp1s16f[2-3]
link: eno1
addresses : [ "10.15.99.25/24" ]