nixos/kubeadm/modules/k8s-common.nix

@@ -195,15 +195,47 @@ in
       env -i PATH=/run/current-system/sw/bin:/usr/bin:/bin kubeadm init \
         --config /tmp/kubeadm/init-config.yaml \
         --upload-certs \
-        --ignore-preflight-errors=NumCPU,HTTPProxyCIDR,Port-10250 || {
+        --ignore-preflight-errors=NumCPU,HTTPProxyCIDR,Port-10250 \
-        echo "==> kubeadm init failed, checking pod status:"
+        --skip-phases=wait-control-plane || {
+        echo "==> kubeadm init phases failed, checking pod status:"
         crictl pods || true
         crictl ps -a || true
+        echo "==> Checking if VIP is bound:"
+        ip -4 addr show | grep "$vip" || echo "VIP NOT BOUND"
+        echo "==> kube-vip logs:"
+        crictl logs $(crictl ps --name kube-vip -q 2>/dev/null | head -1) 2>/dev/null || echo "Could not get kube-vip logs"
         echo "==> kubelet logs:"
         journalctl -xeu kubelet --no-pager -n 50
         exit 1
       }
 
+      echo "==> Waiting for kube-vip to claim VIP $vip"
+      for i in $(seq 1 60); do
+        if ip -4 addr show | grep -q "$vip"; then
+          echo "==> VIP $vip is bound"
+          break
+        fi
+        if [ "$i" -eq 60 ]; then
+          echo "==> WARNING: VIP not bound after 2 minutes, proceeding anyway"
+        fi
+        sleep 2
+      done
+
+      echo "==> Waiting for API server to be ready"
+      for i in $(seq 1 60); do
+        if curl -sk "https://$vip:6443/healthz" 2>/dev/null | grep -q "ok"; then
+          echo "==> API server is healthy"
+          break
+        fi
+        if [ "$i" -eq 60 ]; then
+          echo "==> ERROR: API server not healthy after 2 minutes"
+          crictl pods || true
+          crictl ps -a || true
+          exit 1
+        fi
+        sleep 2
+      done
+
       mkdir -p /root/.kube
       cp /etc/kubernetes/admin.conf /root/.kube/config
       chmod 600 /root/.kube/config