adding a vlan to talos

[AEnterprise]

I'm trying to add a vlan to talos. The servers are hetzner dedicated servers with a vswitch (linked to a cloud switch) to get a private network link between them.

I've booted into other oses and this works without any issue there. Except when i try to do it in talos it loses all network connectivity, effecitvely bricking it.

For context these are the commands that work on debian to get the link up an running without issue.

ip link add link enp7s0 name enp7s0.4000 type vlan id 4000
ip link set enp7s0.4000 mtu 1400
ip link set dev enp7s0.4000 up
ip addr add 172.16.1.2/24 brd 172.16.1.255 dev enp7s0.4000
route add -net 172.16.0.0 netmask 255.255.0.0 gw 172.16.1.1

I tried to do the following equivalent and this just kills the network

apiVersion: v1alpha1
kind: VLANConfig
name: enp7s0.4000
vlanID: 4000
vlanMode: 802.1q
parent: eth0
up: true
addresses:
  - address: 172.16.1.2/32
routes:
  - destination: 172.16.0.0/16
    gateway: 172.16.1.1

i've also tried to explicity define the regular interface in case this config was removing that one, but same result

machine:
  network:
    interfaces:
      - interface: enp7s0
        addresses:
          - x.x.x.x/26
        routes:
          - network: 0.0.0.0/0
            gateway: y.y.y.y

Does anyone have any working examples or know where i'm going wrong with this?

[smira]

Please make sure you read the note in the documentation: https://docs.siderolabs.com/talos/v1.12/networking/configuration/overview

Note: Default Talos Linux behavior to run DHCP on any physical link is disabled automatically when any link is configured explicitly in the machine configuration. In order to keep DHCP running, it must be explicitly enabled on the desired link(s).

[AEnterprise]

yes, and i've also tried manually configuring it with the last snippet. still doesn't work with that

[smira]

It's the same story about DHCP being disabled either way.

Make sure you still configure DHCP explicitly on the links you need it (I guess it's enp7s0).

Either way you can use console logs to inspect what changed when you applied your config change.

You can also do talosctl apply-config --mode=try to try changes which will be rolled back automatically.

[AEnterprise]

managed to play around with mode try a bit to get it working. something about the fixed config didn't work, but if i did dhcp instead of fixed ip it works. thanks

[ArjonBu]

@smira
I came across this issue too on 1.12.7 on Hetzner hybrid. The same setup worked fine in 1.11.x. Either the newer versions introduces a bug or 1.12.x is adding unexpected config.

I tried to troubleshoot by getting links and routes and everything looks fine. Couldn't pinpoint the problem. Basically there is nothing going on on the vlan. It's not reachable. talosctl pcap -i enp1s0.4001 during a ping from a cloud node shows only ARP requests going out and no ICMP traffic.

[smira]

See also #12860 for pcap specifically.

But there has been no real bug reported so far. If you look above, you can see that the "Debian" instructions and Talos don't match - MTU is not set correctly in the Talos config.

If you hit a problem, please provide all details.