feat(update-variant)!: allow updating disk_type, disk_size_gb, machine_type, enable_gcfs in place

autogen/main/cluster.tf.tmpl

@@ -716,8 +716,6 @@ resource "google_container_cluster" "primary" {
 {% if update_variant %}
 locals {
   force_node_pool_recreation_resources = [
-    "disk_size_gb",
-    "disk_type",
     "accelerator_count",
     "accelerator_type",
     "gpu_partition_size",
@@ -730,15 +728,13 @@ locals {
     {% if beta_cluster %}
     "local_ssd_ephemeral_count",
     {% endif %}
-    "machine_type",
     "placement_policy",
     "max_pods_per_node",
     "min_cpu_platform",
     "pod_range",
     "preemptible",
     "spot",
     "service_account",
-    "enable_gcfs",
     "enable_gvnic",
     "boot_disk_kms_key",
     "queued_provisioning",

docs/upgrading_to_v35.0.md

@@ -6,3 +6,47 @@ The Terraform Kubernetes Engine Module now requires version 6 of the Google Clou
 
 ### Private Cluster Sub-Modules Endpoint Output
 The private cluster sub-modules now return the cluster's private endpoint for the `endpoint` output when the `enable_private_endpoint` argument is `true`, regardless of the `deploy_using_private_endpoint` argument value.
+
+## Update variant random ID keepers updated
+
+The v35.0 release updates the keepers for the update variant modules. This will force a recreation of the nodepools.
+
+To avoid this, it is possible to edit the remote state of the `random_id` resource to add the new attributes.
+
+1. Perform a `terraform plan` as normal, identifying the `random_id` resource(s) changing and the new/removed attributes
+```tf
+      ~ keepers     = { # forces replacement
+          - "disk_type"                   = "" -> null
+          - "disk_size_gb"                = "" -> null
+          - "machine_type"                = "" -> null
+          - "enable_gcfs"                 = "" -> null
+            # (19 unchanged elements hidden)
+        }
+        # (2 unchanged attributes hidden)
+    }
+```
+2. Pull the remote state locally: `terraform state pull > default.tfstate`
+3. Back up the original remote state: `cp default.tfstate original.tfstate`
+4. Edit the `random_id` resource(s) to add/remove the attributes from the `terraform plan` step
+```diff
+"attributes": {
+            "b64_std": "pool-02-vb4=",
+            "b64_url": "pool-02-vb4",
+            "byte_length": 2,
+            "dec": "pool-02-48574",
+            "hex": "pool-02-bdbe",
+            "id": "vb4",
+            "keepers": {
+            ...
+              "taints": "",
+-             "disk_size_gb": "",
+-             "enable_gcfs": "",
+-             "machine_type": "",
+-             "disk_type": "",
+            },
+            "prefix": "pool-02-"
+          }
+```
+1. Bump the serial number at the top
+2. Push the modified state to the remote `terraform state push default.tfstate`
+3. Confirm the `random_id` resource(s) no longer changes (or the corresponding `nodepool`) in a `terraform plan`

modules/beta-private-cluster-update-variant/cluster.tf

@@ -610,8 +610,6 @@ resource "google_container_cluster" "primary" {
  *****************************************/
 locals {
   force_node_pool_recreation_resources = [
-    "disk_size_gb",
-    "disk_type",
     "accelerator_count",
     "accelerator_type",
     "gpu_partition_size",
@@ -622,15 +620,13 @@ locals {
     "enable_integrity_monitoring",
     "local_ssd_count",
     "local_ssd_ephemeral_count",
-    "machine_type",
     "placement_policy",
     "max_pods_per_node",
     "min_cpu_platform",
     "pod_range",
     "preemptible",
     "spot",
     "service_account",
-    "enable_gcfs",
     "enable_gvnic",
     "boot_disk_kms_key",
     "queued_provisioning",

modules/beta-public-cluster-update-variant/cluster.tf

@@ -589,8 +589,6 @@ resource "google_container_cluster" "primary" {
  *****************************************/
 locals {
   force_node_pool_recreation_resources = [
-    "disk_size_gb",
-    "disk_type",
     "accelerator_count",
     "accelerator_type",
     "gpu_partition_size",
@@ -601,15 +599,13 @@ locals {
     "enable_integrity_monitoring",
     "local_ssd_count",
     "local_ssd_ephemeral_count",
-    "machine_type",
     "placement_policy",
     "max_pods_per_node",
     "min_cpu_platform",
     "pod_range",
     "preemptible",
     "spot",
     "service_account",
-    "enable_gcfs",
     "enable_gvnic",
     "boot_disk_kms_key",
     "queued_provisioning",

modules/private-cluster-update-variant/cluster.tf

@@ -563,8 +563,6 @@ resource "google_container_cluster" "primary" {
  *****************************************/
 locals {
   force_node_pool_recreation_resources = [
-    "disk_size_gb",
-    "disk_type",
     "accelerator_count",
     "accelerator_type",
     "gpu_partition_size",
@@ -574,15 +572,13 @@ locals {
     "enable_secure_boot",
     "enable_integrity_monitoring",
     "local_ssd_count",
-    "machine_type",
     "placement_policy",
     "max_pods_per_node",
     "min_cpu_platform",
     "pod_range",
     "preemptible",
     "spot",
     "service_account",
-    "enable_gcfs",
     "enable_gvnic",
     "boot_disk_kms_key",
     "queued_provisioning",