netplan generates duplicate dns

[Tommy]

Hello,
I have a server with Ubuntu 22.04LTS
Netplan version: netplan.io 0.104-0ubuntu2

When bonds section upper than ethernets section I see duplicate bonds.ifname.nameservers.addresses and bonds.ifname.nameservers.search:

# netplan get
network:
  version: 2
  renderer: networkd
  ethernets:
    ens3f0:
      match:
        macaddress: "some-mac"
      dhcp4: false
      set-name: "ens3f0"
    ens2f0:
      match:
        macaddress: "some-mac"
      dhcp4: false
      set-name: "ens2f0"
  bonds:
    aggi:
      addresses:
      - "10.17.0.73/26"
      nameservers:
        addresses:
        - 8.8.8.8
        - 8.8.4.4
        - 8.8.8.8
        - 8.8.4.4
        search:
        - example.com
        - example.com
      dhcp4: false
      interfaces:
      - ens2f0
      - ens3f0
      parameters:
        mode: "802.3ad"
        mii-monitor-interval: "100"
        up-delay: "200"
        down-delay: "200"
        lacp-rate: "slow"
        transmit-hash-policy: "layer3+4"
      routes:
      - to: "default"
        via: "10.17.0.67"

But when I swap sections everything is fine:
# netplan get
network:
  version: 2
  renderer: networkd
  ethernets:
    ens3f0:
      match:
        macaddress: "some-mac"
      dhcp4: false
      set-name: "ens3f0"
    ens2f0:
      match:
        macaddress: "some-mac"
      dhcp4: false
      set-name: "ens2f0"
  bonds:
    aggi:
      addresses:
      - "10.17.0.73/26"
      nameservers:
        addresses:
        - 8.8.8.8
        - 8.8.4.4
        search:
        - example.com
      dhcp4: false
      interfaces:
      - ens2f0
      - ens3f0
      parameters:
        mode: "802.3ad"
        mii-monitor-interval: "100"
        up-delay: "200"
        down-delay: "200"
        lacp-rate: "slow"
        transmit-hash-policy: "layer3+4"
      routes:
      - to: "default"
        via: "10.17.0.67"

[Danilo Egea Gondolfo (danilogondolfo)]

Hi, thanks for your bug report.

I can confirm that the problem still exists on netplan.io 0.106.

Here is a simpler reproducer:

$ cat /tmp/fakeroot/etc/netplan/90-configs.yaml
network:
  bonds:
    aggi:
      nameservers:
        addresses:
        - 8.8.8.8
        search:
        - example.com
      interfaces:
      - ens3f0
  ethernets:
    ens3f0: {}

$ netplan get --root-dir /tmp/fakeroot/
network:
  version: 2
  ethernets:
    ens3f0: {}
  bonds:
    aggi:
      nameservers:
        addresses:
        - 8.8.8.8
        - 8.8.8.8
        search:
        - example.com
        - example.com
      interfaces:
      - ens3f0

And indeed the duplication doesn't happen if "ethernets" is defined before "bonds".

[Danilo Egea Gondolfo (danilogondolfo)]

It also generates duplicated entries so the problem is not only with the "netplan get" command:

$ cat /tmp/fakeroot/etc/netplan/90-config.yaml
network:
  bonds:
    aggi:
      nameservers:
        addresses:
        - 8.8.8.8
        search:
        - example.com
      interfaces:
      - ens3f0
  ethernets:
    ens3f0: {}

$ cat /tmp/fakeroot/run/systemd/network/10-netplan-aggi.network
[Match]
Name=aggi

[Network]
LinkLocalAddressing=ipv6
DNS=8.8.8.8
DNS=8.8.8.8
Domains=example.com example.com
ConfigureWithoutCarrier=yes

[Danilo Egea Gondolfo (danilogondolfo)]

The problem happens when the parser needs a second pass to find all the interfaces. It doesn't check if the values are already in their lists.

The problem is not limited to the DNS entries.

Here is another example:

network:
  bonds:
    aggi:
      routing-policy:
        - table: 1
          from: 1.2.3.4
      nameservers:
        addresses:
        - 8.8.8.8
        search:
        - example.com
      interfaces:
      - ens3f0
  ethernets:
    ens3f0: {}

The .netdev file will look like this:

[Match]
Name=aggi

[Network]
LinkLocalAddressing=ipv6
DNS=8.8.8.8
DNS=8.8.8.8
Domains=example.com example.com
ConfigureWithoutCarrier=yes

[RoutingPolicyRule]
From=1.2.3.4
Table=1

[RoutingPolicyRule]
From=1.2.3.4
Table=1

[Danilo Egea Gondolfo (danilogondolfo)]

Another instance of duplication:

network:
  bridges:
    br0:
      openvswitch:
        protocols:
          - OpenFlow10
          - OpenFlow11
          - OpenFlow12
        controller:
          addresses:
            - tcp:127.0.0.1:6653
            - tcp:127.0.0.2:6653
      interfaces:
        - nic1
  ethernets:
    nic1: {}

$ tail -n 4 /tmp/fakeroot/run/systemd/system/netplan-ovs-br0.service
ExecStart=/usr/bin/ovs-vsctl set Bridge br0 protocols=OpenFlow10,OpenFlow11,OpenFlow12,OpenFlow10,OpenFlow11,OpenFlow12
ExecStart=/usr/bin/ovs-vsctl set Bridge br0 external-ids:netplan/protocols=OpenFlow10,OpenFlow11,OpenFlow12,OpenFlow10,OpenFlow11,OpenFlow12
ExecStart=/usr/bin/ovs-vsctl set-controller br0 tcp:127.0.0.1:6653 tcp:127.0.0.2:6653 tcp:127.0.0.1:6653 tcp:127.0.0.2:6653
ExecStart=/usr/bin/ovs-vsctl set Bridge br0 external-ids:netplan/global/set-controller=tcp:127.0.0.1:6653,tcp:127.0.0.2:6653,tcp:127.0.0.1:6653,tcp:127.0.0.2:6653

[Danilo Egea Gondolfo (danilogondolfo)]

PR addressing these issues: https://github.com/canonical/netplan/pull/329

[Lukas Märdian (slyon)]

Fixed https://launchpad.net/ubuntu/+source/netplan.io/0.106.1-1