Merge branch 'release-candidate' into health_out

community/modules/scheduler/schedmd-slurm-gcp-v6-controller/README.md

@@ -238,7 +238,6 @@ limitations under the License.
 | <a name="module_daos_network_storage_scripts"></a> [daos\_network\_storage\_scripts](#module\_daos\_network\_storage\_scripts) | ../../../../modules/scripts/startup-script | n/a |
 | <a name="module_nodeset_cleanup"></a> [nodeset\_cleanup](#module\_nodeset\_cleanup) | ./modules/cleanup_compute | n/a |
 | <a name="module_nodeset_cleanup_tpu"></a> [nodeset\_cleanup\_tpu](#module\_nodeset\_cleanup\_tpu) | ./modules/cleanup_tpu | n/a |
-| <a name="module_slurm_controller_instance"></a> [slurm\_controller\_instance](#module\_slurm\_controller\_instance) | github.com/GoogleCloudPlatform/slurm-gcp.git//terraform/slurm_cluster/modules/_slurm_instance | 6.8.5 |
 | <a name="module_slurm_controller_template"></a> [slurm\_controller\_template](#module\_slurm\_controller\_template) | github.com/GoogleCloudPlatform/slurm-gcp.git//terraform/slurm_cluster/modules/slurm_instance_template | 6.8.5 |
 | <a name="module_slurm_files"></a> [slurm\_files](#module\_slurm\_files) | ./modules/slurm_files | n/a |
 | <a name="module_slurm_login_instance"></a> [slurm\_login\_instance](#module\_slurm\_login\_instance) | github.com/GoogleCloudPlatform/slurm-gcp.git//terraform/slurm_cluster/modules/_slurm_instance | 6.8.5 |
@@ -250,6 +249,7 @@ limitations under the License.
 
 | Name | Type |
 |------|------|
+| [google_compute_instance_from_template.controller](https://registry.terraform.io/providers/hashicorp/google/latest/docs/resources/compute_instance_from_template) | resource |
 | [google_secret_manager_secret.cloudsql](https://registry.terraform.io/providers/hashicorp/google/latest/docs/resources/secret_manager_secret) | resource |
 | [google_secret_manager_secret_iam_member.cloudsql_secret_accessor](https://registry.terraform.io/providers/hashicorp/google/latest/docs/resources/secret_manager_secret_iam_member) | resource |
 | [google_secret_manager_secret_version.cloudsql_version](https://registry.terraform.io/providers/hashicorp/google/latest/docs/resources/secret_manager_secret_version) | resource |

community/modules/scheduler/schedmd-slurm-gcp-v6-controller/controller.tf

@@ -73,7 +73,6 @@ module "slurm_controller_template" {
   metadata         = local.metadata
   min_cpu_platform = var.min_cpu_platform
 
-  # network_ip = TODO: add support for network_ip
   on_host_maintenance = var.on_host_maintenance
   preemptible         = var.preemptible
   service_account     = local.service_account
@@ -82,40 +81,38 @@ module "slurm_controller_template" {
   source_image_project = local.source_image_project_normalized # requires source_image_logic.tf
   source_image         = local.source_image                    # requires source_image_logic.tf
 
-  # spot = TODO: add support for spot (?)
   subnetwork = var.subnetwork_self_link
 
   tags = concat([local.slurm_cluster_name], var.tags)
   # termination_action = TODO: add support for termination_action (?)
 }
 
 # INSTANCE
-locals {
-  # TODO: add support for proper access_config
-  access_config = {
-    nat_ip       = null
-    network_tier = null
+resource "google_compute_instance_from_template" "controller" {
+  name                     = "${local.slurm_cluster_name}-controller"
+  project                  = var.project_id
+  zone                     = var.zone
+  source_instance_template = module.slurm_controller_template.self_link
+
+  allow_stopping_for_update = true
+
+  # Can't rely on template to specify nics due to usage of static_ip
+  network_interface {
+    dynamic "access_config" {
+      for_each = var.enable_controller_public_ips ? ["unit"] : []
+      content {
+        nat_ip       = null
+        network_tier = null
+      }
+    }
+    network_ip = length(var.static_ips) == 0 ? "" : var.static_ips[0]
+    subnetwork = var.subnetwork_self_link
   }
 }
 
-module "slurm_controller_instance" {
-  source = "github.com/GoogleCloudPlatform/slurm-gcp.git//terraform/slurm_cluster/modules/_slurm_instance?ref=6.8.5"
-
-  access_config       = var.enable_controller_public_ips ? [local.access_config] : []
-  add_hostname_suffix = false
-  hostname            = "${local.slurm_cluster_name}-controller"
-  instance_template   = module.slurm_controller_template.self_link
-
-  project_id          = var.project_id
-  region              = var.region
-  slurm_cluster_name  = local.slurm_cluster_name
-  slurm_instance_role = "controller"
-  static_ips          = var.static_ips
-  subnetwork          = var.subnetwork_self_link
-  zone                = var.zone
-  metadata            = var.metadata
-
-  labels = local.labels
+moved {
+  from = module.slurm_controller_instance.google_compute_instance_from_template.slurm_instance[0]
+  to   = google_compute_instance_from_template.controller
 }
 
 # SECRETS: CLOUDSQL

community/modules/scheduler/schedmd-slurm-gcp-v6-controller/login.tf

@@ -78,5 +78,5 @@ module "slurm_login_instance" {
   zone                = each.value.zone
 
   # trigger replacement of login nodes when the controller instance is replaced
-  replace_trigger = module.slurm_controller_instance.instances_self_links[0]
+  replace_trigger = google_compute_instance_from_template.controller.self_link
 }

community/modules/scheduler/schedmd-slurm-gcp-v6-controller/outputs.tf

@@ -19,7 +19,7 @@ output "slurm_cluster_name" {
 
 output "slurm_controller_instance" {
   description = "Compute instance of controller node"
-  value       = module.slurm_controller_instance.slurm_instances[0]
+  value       = google_compute_instance_from_template.controller
 }
 
 output "slurm_login_instances" {
@@ -36,6 +36,6 @@ output "instructions" {
   description = "Post deployment instructions."
   value       = <<-EOT
     To SSH to the controller (may need to add '--tunnel-through-iap'):
-      gcloud compute ssh ${module.slurm_controller_instance.instances_self_links[0]}
+      gcloud compute ssh ${google_compute_instance_from_template.controller.self_link}
   EOT
 }

modules/compute/gke-node-pool/main.tf

@@ -93,7 +93,8 @@ resource "google_container_node_pool" "node_pool" {
         gpu_partition_size = try(ga.value.gpu_partition_size, null)
 
         dynamic "gpu_driver_installation_config" {
-          for_each = try([ga.gpu_driver_installation_config], [{ gpu_driver_version = "DEFAULT" }])
+          # in case user did not specify guest_accelerator settings, we need a try to default to []
+          for_each = try([ga.value.gpu_driver_installation_config], [{ gpu_driver_version = "DEFAULT" }])
           iterator = gdic
           content {
             gpu_driver_version = gdic.value.gpu_driver_version