Update cluster as regional and update ip_cidr_range.

Author: yangyuwei

tools/kubernetes/terraform/examples/v5e/main.tf

@@ -1,7 +1,6 @@
 variable "project_id" {}
 variable "resource_name_prefix" {}
 variable "region" {}
-variable "location" {}
 variable "tpu_node_pools" {}
 variable "maintenance_interval" {}
 
@@ -11,7 +10,6 @@ module "tpu-gke" {
   project_id           = var.project_id
   resource_name_prefix = var.resource_name_prefix
   region               = var.region
-  location             = var.location
   tpu_node_pools       = var.tpu_node_pools
   maintenance_interval = var.maintenance_interval
 }

tools/kubernetes/terraform/examples/v5e/terraform.tfvars

@@ -1,8 +1,25 @@
-project_id           = "project-id"
+project_id           = "project_id"
 resource_name_prefix = "tpu-v5e-test"
 region               = "us-east5"
-location             = "us-east5-b"
 tpu_node_pools = [{
+  zone         = "us-east5-b"
+  node_count   = 64
+  machine_type = "ct5lp-hightpu-4t"
+  topology     = "16x16"
+  policy       = "sb-compact-1"
+  }, {
+  zone         = "us-east5-b"
+  node_count   = 64
+  machine_type = "ct5lp-hightpu-4t"
+  topology     = "16x16"
+  policy       = "sb-compact-1"
+  }, {
+  zone         = "us-east5-b"
+  node_count   = 64
+  machine_type = "ct5lp-hightpu-4t"
+  topology     = "16x16"
+  policy       = "sb-compact-1"
+  }, {
   zone         = "us-east5-b"
   node_count   = 64
   machine_type = "ct5lp-hightpu-4t"

tools/kubernetes/terraform/module/main.tf

@@ -16,7 +16,7 @@
 
 # GKE cluster
 data "google_container_engine_versions" "gke_version" {
-  location       = var.location
+  location       = var.region
   version_prefix = "1.27."
 }
 
@@ -36,12 +36,12 @@ resource "google_compute_subnetwork" "subnet" {
   name          = "${var.resource_name_prefix}-subnet"
   region        = var.region
   network       = google_compute_network.vpc.name
-  ip_cidr_range = "10.10.0.0/24"
+  ip_cidr_range = "10.10.0.0/19"
 }
 
 resource "google_container_cluster" "tpu_cluster" {
   name     = "${var.resource_name_prefix}-gke-cluster"
-  location = var.location
+  location = var.region
 
   # We can't create a cluster with no node pool defined, but we want to only use
   # separately managed node pools. So we create the smallest possible default
@@ -70,7 +70,7 @@ resource "google_container_node_pool" "multihost_tpu" {
   name           = "${google_container_cluster.tpu_cluster.name}-${count.index}"
   provider       = google-beta
   project        = var.project_id
-  location       = var.location
+  location       = var.region
   node_locations = [var.tpu_node_pools[count.index].zone]
   cluster        = google_container_cluster.tpu_cluster.name
 

tools/kubernetes/terraform/module/variables.tf

@@ -22,10 +22,6 @@ variable "region" {
   description = "region"
 }
 
-variable "location" {
-  description = "location"
-}
-
 variable "resource_name_prefix" {
   default     = ""
   description = "prefix for all the resouce naming"