What is the issue?

I recently migrated from a K3s cluster running Ubuntu 22.04 to a Talos v1.8 cluster.

I have been running Tailscale operator for over a year and have a mix of Ingresses, a connector and a service type LoadBalancer exposed through Tailscale, that all worked on the K3s-cluster. However, on the Talos v1.8 cluster, the connector (subnet router) and LoadBalancer service did not work. Please note that Ingresses (HTTPS) works.

After a bit of research, I found out about the TUN device changes with the new version of runc included in Talos v1.8. I have followed the documentation and deployed the generic-device-plugin and added the needed resource request to a Tailscale proxyclass.

This fixed the connector (subnet router), but not the loadbalancer service.

I don't really see any obvious errors in the log output, but trying to access the exposed port doesn't work.

The logs from the proxy are here.

Steps to reproduce

1. Setup a Talos v1.8 cluster
2. Deploy generic-device-plugin for /dev/net/tun access
3. Deploy Tailscale operator with proxyclass that gets access to tun device
4. Deploy echo-server with a service type LoadBalancer
5. Try to access the exposed port with curl and observe that nothing happens.

Are there any recent changes that introduced the issue?

No response

OS

Linux

OS version

Talos v1.8.2

Tailscale version

1.76.6

Other software

No response

Bug report

No response