resource "google_compute_network" "network" { name = var.cluster_name description = "Network for the ${var.cluster_name} cluster" auto_create_subnetworks = true timeouts { delete = "6m" } } resource "google_compute_firewall" "allow-ssh" { name = "${var.cluster_name}-allow-ssh" network = google_compute_network.network.name allow { protocol = "tcp" ports = [22] } source_ranges = ["0.0.0.0/0"] target_tags = ["${var.cluster_name}-controller", "${var.cluster_name}-worker"] } resource "google_compute_firewall" "internal-etcd" { name = "${var.cluster_name}-internal-etcd" network = google_compute_network.network.name allow { protocol = "tcp" ports = [2379, 2380] } source_tags = ["${var.cluster_name}-controller"] target_tags = ["${var.cluster_name}-controller"] } # Allow Prometheus to scrape etcd metrics resource "google_compute_firewall" "internal-etcd-metrics" { name = "${var.cluster_name}-internal-etcd-metrics" network = google_compute_network.network.name allow { protocol = "tcp" ports = [2381] } source_tags = ["${var.cluster_name}-worker"] target_tags = ["${var.cluster_name}-controller"] } # Allow Prometheus to scrape kube-scheduler and kube-controller-manager metrics resource "google_compute_firewall" "internal-kube-metrics" { name = "${var.cluster_name}-internal-kube-metrics" network = google_compute_network.network.name allow { protocol = "tcp" ports = [10257, 10259] } source_tags = ["${var.cluster_name}-worker"] target_tags = ["${var.cluster_name}-controller"] } resource "google_compute_firewall" "allow-apiserver" { name = "${var.cluster_name}-allow-apiserver" network = google_compute_network.network.name allow { protocol = "tcp" ports = [6443] } source_ranges = ["0.0.0.0/0"] target_tags = ["${var.cluster_name}-controller"] } # BGP and IPIP # https://docs.projectcalico.org/latest/reference/public-cloud/gce resource "google_compute_firewall" "internal-bgp" { count = var.networking != "flannel" ? 1 : 0 name = "${var.cluster_name}-internal-bgp" network = google_compute_network.network.name allow { protocol = "tcp" ports = ["179"] } allow { protocol = "ipip" } source_tags = ["${var.cluster_name}-controller", "${var.cluster_name}-worker"] target_tags = ["${var.cluster_name}-controller", "${var.cluster_name}-worker"] } # flannel VXLAN resource "google_compute_firewall" "internal-vxlan" { count = var.networking == "flannel" ? 1 : 0 name = "${var.cluster_name}-internal-vxlan" network = google_compute_network.network.name allow { protocol = "udp" ports = [4789] } source_tags = ["${var.cluster_name}-controller", "${var.cluster_name}-worker"] target_tags = ["${var.cluster_name}-controller", "${var.cluster_name}-worker"] } # Cilium VXLAN resource "google_compute_firewall" "internal-linux-vxlan" { count = var.networking == "cilium" ? 1 : 0 name = "${var.cluster_name}-linux-vxlan" network = google_compute_network.network.name allow { protocol = "udp" ports = [8472] } # Cilium health allow { protocol = "icmp" } allow { protocol = "tcp" ports = [4240] } source_tags = ["${var.cluster_name}-controller", "${var.cluster_name}-worker"] target_tags = ["${var.cluster_name}-controller", "${var.cluster_name}-worker"] } # Allow Prometheus to scrape node-exporter daemonset resource "google_compute_firewall" "internal-node-exporter" { name = "${var.cluster_name}-internal-node-exporter" network = google_compute_network.network.name allow { protocol = "tcp" ports = [9100] } source_tags = ["${var.cluster_name}-worker"] target_tags = ["${var.cluster_name}-controller", "${var.cluster_name}-worker"] } # Allow Prometheus to scrape kube-proxy metrics resource "google_compute_firewall" "internal-kube-proxy" { name = "${var.cluster_name}-internal-kube-proxy" network = google_compute_network.network.name allow { protocol = "tcp" ports = [10249] } source_tags = ["${var.cluster_name}-worker"] target_tags = ["${var.cluster_name}-controller", "${var.cluster_name}-worker"] } # Allow apiserver to access kubelets for exec, log, port-forward resource "google_compute_firewall" "internal-kubelet" { name = "${var.cluster_name}-internal-kubelet" network = google_compute_network.network.name allow { protocol = "tcp" ports = [10250] } # allow Prometheus to scrape kubelet metrics too source_tags = ["${var.cluster_name}-controller", "${var.cluster_name}-worker"] target_tags = ["${var.cluster_name}-controller", "${var.cluster_name}-worker"] } # Workers resource "google_compute_firewall" "allow-ingress" { name = "${var.cluster_name}-allow-ingress" network = google_compute_network.network.name allow { protocol = "tcp" ports = [80, 443] } source_ranges = ["0.0.0.0/0"] target_tags = ["${var.cluster_name}-worker"] } resource "google_compute_firewall" "google-worker-health-checks" { name = "${var.cluster_name}-worker-health" network = google_compute_network.network.name allow { protocol = "tcp" ports = [10256] } # https://cloud.google.com/compute/docs/instance-groups/autohealing-instances-in-migs source_ranges = [ "35.191.0.0/16", "130.211.0.0/22", ] target_tags = ["${var.cluster_name}-worker"] } resource "google_compute_firewall" "google-ingress-health-checks" { name = "${var.cluster_name}-ingress-health" network = google_compute_network.network.name allow { protocol = "tcp" ports = [10254] } # https://cloud.google.com/load-balancing/docs/health-check-concepts#method source_ranges = [ "35.191.0.0/16", "130.211.0.0/22", "35.191.0.0/16", "209.85.152.0/22", "209.85.204.0/22", ] target_tags = ["${var.cluster_name}-worker"] }