This is a bit WIP whilst I keep digging into this.
I'm just logging here really.
Two nodes: k3s cluster: dual stack
1 ipv6 only node, runs tailscale operator, initial node 1 ipv4/ipv6 node,
completely separate lans
using k3s --vpn-auth=tailscale flag to get the two nodes connected (tried specifying --iface but this was easier).
I pulled public images, ran them on both nodes, everything was grand.
I then tried to deploy an image that I was hosting on an internal registry on inside k3s itself. For the sake of identifying the two nodes, lets call them A and B. B was where the registry was running A was where the pod was trying to deploy, with a nice little TLS handshake time out error.
Issue, MTU size was exceeded, added a clamp, got through
Need to figure a proper solution