Add Google Cloud worker instances to a target pool

* Background: A managed instance group of workers is used in backend
services for global load balancing (HTTP/HTTPS Ingress) and output
for custom global load balancing use cases
* Add worker instances to a target pool load balancing TCP/UDP
applications (NodePort or proxied). Output as `worker_target_pool`
* Health check for workers with a healthy Ingress controller. Forward
rules (regional) to target pools don't support different external and
internal ports so choosing nodes with Ingress allows proxying as a
workaround
* A target pool is a logical grouping only. It doesn't add costs to
clusters or worker pools

CHANGES.md

@@ -8,6 +8,9 @@ Notable changes between versions.
 
 #### Google Cloud
 
+* Add ability to load balance TCP/UDP applications ([#442](https://github.com/poseidon/typhoon/pull/442))
+  * Add worker instances to a target pool, output as `worker_target_pool`
+  * Health check for workers with Ingress controllers. Forward rules don't support differing internal/external ports, but some Ingress controllers support TCP/UDP proxy as a workaround 
 * Remove Haswell minimum CPU platform requirement ([#439](https://github.com/poseidon/typhoon/pull/439))
   * Google Cloud API implements `min_cpu_platform` to mean "use exactly this CPU". Revert [#405](https://github.com/poseidon/typhoon/pull/405) added in v1.13.4.
   * Fix error creating clusters in new regions without Haswell (e.g. europe-west2) ([#438](https://github.com/poseidon/typhoon/issues/438))

google-cloud/container-linux/kubernetes/network.tf

@@ -138,8 +138,8 @@ resource "google_compute_firewall" "allow-ingress" {
   target_tags   = ["${var.cluster_name}-worker"]
 }
 
-resource "google_compute_firewall" "google-health-checks" {
-  name    = "${var.cluster_name}-google-health-checks"
+resource "google_compute_firewall" "google-ingress-health-checks" {
+  name    = "${var.cluster_name}-ingress-health"
   network = "${google_compute_network.network.name}"
 
   allow {
@@ -147,7 +147,19 @@ resource "google_compute_firewall" "google-health-checks" {
     ports    = [10254]
   }
 
-  # https://cloud.google.com/compute/docs/load-balancing/tcp-ssl/tcp-proxy#health-checking
-  source_ranges = ["130.211.0.0/22", "35.191.0.0/16"]
+  # https://cloud.google.com/load-balancing/docs/health-check-concepts#method
+  source_ranges = [
+    # Global LB health checks
+    "35.191.0.0/16",
+
+    "130.211.0.0/22",
+
+    # Region LB health checks
+    "35.191.0.0/16",
+
+    "209.85.152.0/22",
+    "209.85.204.0/22",
+  ]
+
   target_tags = ["${var.cluster_name}-worker"]
 }

google-cloud/container-linux/kubernetes/outputs.tf

@@ -33,6 +33,11 @@ output "network_self_link" {
 # Outputs for custom load balancing
 
 output "worker_instance_group" {
-  description = "Full URL of the worker managed instance group"
+  description = "Worker managed instance group full URL"
   value       = "${module.workers.instance_group}"
 }
+
+output "worker_target_pool" {
+  description = "Worker target pool self link"
+  value       = "${module.workers.target_pool}"
+}

google-cloud/container-linux/kubernetes/workers/outputs.tf

@@ -1,4 +1,13 @@
+# Outputs for global load balancing
+
 output "instance_group" {
-  description = "Full URL of the worker managed instance group"
+  description = "Worker managed instance group full URL"
   value       = "${google_compute_region_instance_group_manager.workers.instance_group}"
 }
+
+# Outputs for regional load balancing
+
+output "target_pool" {
+  description = "Worker target pool self link"
+  value       = "${google_compute_target_pool.workers.self_link}"
+}

google-cloud/container-linux/kubernetes/workers/target_pool.tf

@@ -0,0 +1,21 @@
+# Target pool for TCP/UDP load balancing
+resource "google_compute_target_pool" "workers" {
+  name             = "${var.name}-worker-pool"
+  session_affinity = "NONE"
+
+  health_checks = [
+    "${google_compute_http_health_check.workers.name}",
+  ]
+}
+
+# HTTP Health Check (for TCP/UDP load balancing)
+# Forward rules (regional) to target pools don't support different external
+# and internal ports. Health check for nodes with Ingress controllers that
+# may support proxying or otherwise satisfy the check.
+resource "google_compute_http_health_check" "workers" {
+  name        = "${var.name}-target-pool-health"
+  description = "Health check for the worker target pool"
+
+  port         = 10254
+  request_path = "/healthz"
+}

google-cloud/container-linux/kubernetes/workers/workers.tf

@@ -1,4 +1,4 @@
-# Regional managed instance group of workers
+# Managed instance group of workers
 resource "google_compute_region_instance_group_manager" "workers" {
   name        = "${var.name}-worker-group"
   description = "Compute instance group of ${var.name} workers"
@@ -9,6 +9,7 @@ resource "google_compute_region_instance_group_manager" "workers" {
   region             = "${var.region}"
 
   target_size  = "${var.count}"
+  target_pools = ["${google_compute_target_pool.workers.self_link}"]
 
   named_port {
     name = "http"