fixing tailscaled routes

README.md

@@ -43,6 +43,25 @@ worker_nodes = {
 Stateful apps currently pin retained local PVs to the `debian` node. Move or
 duplicate those PV manifests when you want storage on another node.
 
+The website NodePort is reachable from the OCI jump box through the Raspberry Pi
+Tailscale interface. `bootstrap/cluster` installs a persistent
+`homelab-tailscale-nodeport.service` on the configured worker to restore the
+route, rp_filter settings, and iptables rules after reboot. Override the
+defaults through `tailscale_nodeport_access` when the jump-box IP, Pi Tailscale
+IP, pod CIDR, or NodePort changes:
+
+```hcl
+tailscale_nodeport_access = {
+  enabled           = true
+  worker_key        = "raspberrypi"
+  peer_ip           = "100.118.255.19"
+  node_tailscale_ip = "100.77.80.72"
+  pod_cidr          = "10.244.0.0/16"
+  node_port         = 30080
+  target_port       = 80
+}
+```
+
 For `./lab.sh nuke`, set `WORKER_SSH_TARGETS` to a space-separated list of
 remote SSH targets when more worker nodes exist. Set it to an empty string for a
 single-node rebuild.

bootstrap/cluster/main.tf

@@ -290,14 +290,21 @@ resource "null_resource" "kubeadm_worker" {
   depends_on = [data.external.kubeadm_join_command]
 
   triggers = {
-    node_name               = each.value.node_name
+    node_name                            = each.value.node_name
-    host                    = each.value.host
+    host                                 = each.value.host
-    user                    = each.value.user
+    user                                 = each.value.user
-    ssh_key_path            = each.value.ssh_key_path
+    ssh_key_path                         = each.value.ssh_key_path
-    registry_endpoint       = var.registry_endpoint
+    registry_endpoint                    = var.registry_endpoint
-    registry_config_version = "6"
+    registry_config_version              = "6"
-    node_dns_servers        = join(" ", var.node_dns_servers)
+    node_dns_servers                     = join(" ", var.node_dns_servers)
-    persistent_volume_dirs  = join(",", var.persistent_volume_dirs)
+    persistent_volume_dirs               = join(",", var.persistent_volume_dirs)
+    tailscale_nodeport_version           = "1"
+    tailscale_nodeport_enabled           = var.tailscale_nodeport_access.enabled && each.key == var.tailscale_nodeport_access.worker_key ? "true" : "false"
+    tailscale_nodeport_peer_ip           = var.tailscale_nodeport_access.peer_ip
+    tailscale_nodeport_node_tailscale_ip = var.tailscale_nodeport_access.node_tailscale_ip
+    tailscale_nodeport_pod_cidr          = var.tailscale_nodeport_access.pod_cidr
+    tailscale_nodeport_node_port         = tostring(var.tailscale_nodeport_access.node_port)
+    tailscale_nodeport_target_port       = tostring(var.tailscale_nodeport_access.target_port)
   }
 
   connection {
@@ -518,6 +525,82 @@ if ! getent hosts "${self.triggers.node_name}" >/dev/null; then
   printf '%s %s\n' "${self.triggers.host}" "${self.triggers.node_name}" | sudo tee -a /etc/hosts >/dev/null
 fi
 
+configure_tailscale_nodeport_access() {
+  local enabled="$1"
+  local peer_ip="$2"
+  local node_tailscale_ip="$3"
+  local pod_cidr="$4"
+  local node_port="$5"
+  local target_port="$6"
+
+  if [ "$enabled" != "true" ]; then
+    return 0
+  fi
+
+  sudo mkdir -p /usr/local/sbin /etc/sysctl.d
+  sudo tee /etc/sysctl.d/98-homelab-tailscale-nodeport.conf >/dev/null <<NODEPORT_SYSCTL_EOT
+net.ipv4.conf.all.rp_filter = 0
+net.ipv4.conf.tailscale0.rp_filter = 0
+NODEPORT_SYSCTL_EOT
+
+  sudo tee /usr/local/sbin/homelab-tailscale-nodeport.sh >/dev/null <<NODEPORT_SCRIPT_EOT
+#!/bin/sh
+set -eu
+
+PEER_IP="$peer_ip"
+NODE_TAILSCALE_IP="$node_tailscale_ip"
+POD_CIDR="$pod_cidr"
+NODE_PORT="$node_port"
+TARGET_PORT="$target_port"
+
+sysctl -w net.ipv4.conf.all.rp_filter=0 >/dev/null
+sysctl -w net.ipv4.conf.tailscale0.rp_filter=0 >/dev/null 2>&1 || true
+
+if ! ip link show tailscale0 >/dev/null 2>&1; then
+  echo "tailscale0 is not present; skipping Tailscale NodePort routing"
+  exit 0
+fi
+
+ip route replace "$PEER_IP/32" dev tailscale0 src "$NODE_TAILSCALE_IP"
+
+iptables -C INPUT -i tailscale0 -p tcp --dport "$NODE_PORT" -j ACCEPT 2>/dev/null ||
+  iptables -I INPUT 1 -i tailscale0 -p tcp --dport "$NODE_PORT" -j ACCEPT
+iptables -C FORWARD -i tailscale0 -d "$POD_CIDR" -p tcp --dport "$TARGET_PORT" -j ACCEPT 2>/dev/null ||
+  iptables -I FORWARD 1 -i tailscale0 -d "$POD_CIDR" -p tcp --dport "$TARGET_PORT" -j ACCEPT
+iptables -C FORWARD -s "$POD_CIDR" -o tailscale0 -m conntrack --ctstate RELATED,ESTABLISHED -j ACCEPT 2>/dev/null ||
+  iptables -I FORWARD 1 -s "$POD_CIDR" -o tailscale0 -m conntrack --ctstate RELATED,ESTABLISHED -j ACCEPT
+iptables -t nat -C POSTROUTING -s 100.64.0.0/10 -d "$POD_CIDR" -p tcp --dport "$TARGET_PORT" -m comment --comment tailscale-nodeport-to-pods -j MASQUERADE 2>/dev/null ||
+  iptables -t nat -I POSTROUTING 1 -s 100.64.0.0/10 -d "$POD_CIDR" -p tcp --dport "$TARGET_PORT" -m comment --comment tailscale-nodeport-to-pods -j MASQUERADE
+NODEPORT_SCRIPT_EOT
+  sudo chmod 0755 /usr/local/sbin/homelab-tailscale-nodeport.sh
+
+  sudo tee /etc/systemd/system/homelab-tailscale-nodeport.service >/dev/null <<'NODEPORT_SERVICE_EOT'
+[Unit]
+Description=Homelab Tailscale NodePort routing
+After=network-online.target tailscaled.service kubelet.service
+Wants=network-online.target
+
+[Service]
+Type=oneshot
+ExecStart=/usr/local/sbin/homelab-tailscale-nodeport.sh
+RemainAfterExit=yes
+
+[Install]
+WantedBy=multi-user.target
+NODEPORT_SERVICE_EOT
+  sudo systemctl daemon-reload
+  sudo systemctl enable homelab-tailscale-nodeport.service >/dev/null
+  sudo systemctl restart homelab-tailscale-nodeport.service
+}
+
+configure_tailscale_nodeport_access \
+  "${self.triggers.tailscale_nodeport_enabled}" \
+  "${self.triggers.tailscale_nodeport_peer_ip}" \
+  "${self.triggers.tailscale_nodeport_node_tailscale_ip}" \
+  "${self.triggers.tailscale_nodeport_pod_cidr}" \
+  "${self.triggers.tailscale_nodeport_node_port}" \
+  "${self.triggers.tailscale_nodeport_target_port}"
+
 configure_containerd_registry "${self.triggers.registry_endpoint}"
 
 pv_dirs="${self.triggers.persistent_volume_dirs}"

bootstrap/cluster/variables.tf

@@ -61,3 +61,25 @@ variable "worker_nodes" {
     }
   }
 }
+
+variable "tailscale_nodeport_access" {
+  type = object({
+    enabled           = bool
+    worker_key        = string
+    peer_ip           = string
+    node_tailscale_ip = string
+    pod_cidr          = string
+    node_port         = number
+    target_port       = number
+  })
+
+  default = {
+    enabled           = true
+    worker_key        = "raspberrypi"
+    peer_ip           = "100.118.255.19"
+    node_tailscale_ip = "100.77.80.72"
+    pod_cidr          = "10.244.0.0/16"
+    node_port         = 30080
+    target_port       = 80
+  }
+}