Add an option to create regional clusters on GKE

Less prone to k8s API failure this way, although it costs about
70$ a month

Ref 2i2c-org#1248
Ref 2i2c-org#1102

config/clusters/leap/cluster.yaml

@@ -4,7 +4,7 @@ gcp:
   key: enc-deployer-credentials.secret.json
   project: leap-pangeo
   cluster: leap-cluster
-  zone: us-central1-c
+  zone: us-central1
 support:
   helm_chart_values_files:
     - support.values.yaml

docs/howto/operate/new-cluster/new-cluster.md

@@ -30,6 +30,15 @@ The _minimum_ inputs this file requires are:
   Primary identifier to 'group' together resources.
 - `project_id`: GCP Project ID to create resources in.
   Should be the id, rather than display name of the project.
+- `regional_cluster`: Set to true to provision a [GKE Regional
+  Highly Available cluster](https://cloud.google.com/kubernetes-engine/docs/concepts/regional-clusters).
+  Costs ~70$ a month, but worth it for the added reliability for most
+  cases except when cost saving is an absolute requirement.
+- `zone`: Zone where cluster nodes and filestore for home directory
+  are created.
+- `region`: Region where cluster master (if `regional_cluster` is
+  `true`) is run, as well as any storage buckets created with
+  `user_buckets`.
 
 See the [variables file](https://github.com/2i2c-org/infrastructure/tree/HEAD/terraform/gcp/variables.tf) for other inputs this file can take and their descriptions.
 

terraform/gcp/cluster.tf

@@ -24,9 +24,10 @@ resource "google_container_cluster" "cluster" {
   # Setting cluster autoscaling profile is in google-beta
   provider = google-beta
 
-  name     = "${var.prefix}-cluster"
-  location = var.zone
-  project  = var.project_id
+  name           = "${var.prefix}-cluster"
+  location       = var.regional_cluster ? var.region : var.zone
+  node_locations = [var.zone]
+  project        = var.project_id
 
   initial_node_count       = 1
   remove_default_node_pool = true

terraform/gcp/projects/leap.tfvars

@@ -7,6 +7,8 @@ enable_private_cluster = false
 
 # GPUs not available in us-central1-b
 zone = "us-central1-c"
+region = "us-central1"
+regional_cluster = true
 
 # Multi-tenant cluster, network policy is required to enforce separation between hubs
 enable_network_policy  = true

terraform/gcp/projects/m2lines.tfvars

@@ -6,6 +6,8 @@ enable_network_policy  = true
 
 # GPUs not available in us-central1-b
 zone = "us-central1-c"
+region = "us-central1"
+regional_cluster = true
 
 # Setup a filestore for in-cluster NFS
 enable_filestore = true

terraform/gcp/variables.tf

@@ -88,6 +88,22 @@ variable "region" {
 
 }
 
+variable "regional_cluster" {
+  type        = bool
+  default     = false
+  description = <<-EOT
+  Enable to have a highly available cluster with multi zonal masters
+
+  These are more reliable, as otherwise the k8s API might have small
+  outages now and then - with this set to false, the nodes serving
+  the k8s API can go down periodically for upgrades.
+
+  See https://cloud.google.com/kubernetes-engine/docs/concepts/regional-clusters
+  for more information
+  EOT
+}
+
+
 variable "zone" {
   type        = string
   default     = "us-central1-b"