fix: derive cluster network interface from host facts

The Proxmox Ubuntu clones are exposing their primary NIC as eth0, not ens18.
Use ansible_default_ipv4.interface for k3s flannel and kube-vip so bootstrap
tracks the actual interface name instead of a guessed template default.

ansible/roles/k3s-agent/defaults/main.yml

@@ -4,4 +4,4 @@ k3s_server_url: ""
 k3s_token: ""
 k3s_node_ip: ""
 k3s_kubelet_cloud_provider_external: false
-k3s_flannel_iface: ens18
+k3s_flannel_iface: "{{ ansible_default_ipv4.interface | default('eth0') }}"

ansible/roles/k3s-server/defaults/main.yml

@@ -6,7 +6,7 @@ k3s_primary_public_ip: ""
 k3s_disable_embedded_ccm: false
 k3s_disable_servicelb: true
 k3s_kubelet_cloud_provider_external: false
-k3s_flannel_iface: ens18
+k3s_flannel_iface: "{{ ansible_default_ipv4.interface | default('eth0') }}"
 # Load Balancer endpoint for HA cluster joins (set in inventory)
 kube_api_endpoint: ""
 # Tailscale DNS names for control planes (to enable tailnet access)

ansible/roles/kube-vip-deploy/defaults/main.yml

@@ -1,4 +1,4 @@
 ---
 kube_vip_version: v1.1.2
-kube_vip_interface: ens18
+kube_vip_interface: "{{ ansible_default_ipv4.interface | default('eth0') }}"
 kube_vip_address: "{{ kube_api_endpoint }}"