upgrading_to_v7.0.md

Upgrading to v7.0

The v7.0 release of kubernetes-engine is a backwards incompatible release.

Provider Version

Support for Google provider versions older than v3.1 has been removed due to the introduction of Surge Upgrade features.

Surge Upgrades (beta)

In order to support Surge Upgrades on beta clusters, an upgrade_settings block has been added to all node pools. This will show up in terraform plan:

      + upgrade_settings {
          + max_surge       = 1
          + max_unavailable = 0
        }

The new default behavior will upgrade 1 node at a time. You can tune this behaviour using the new max_surge and max_unavailable settings on the node pool input.

Note that changing upgrade settings can be done in-place (without forcing a node pool recreation).

Node Pool for_each

The google_container_node_pool resource has been updated to use for_each instead of count. This allows adding/removing node pools without causing a diff on unrelated node pools.

Updating to this new format requires running a state migration. Note that this migration must be run with Terraform v0.12.20. You can use a script we provided to automatically make the required state migration.

1. Download the script

   curl -O https://raw.githubusercontent.com/terraform-google-modules/terraform-google-kubernetes-engine/v7.0.0/helpers/migrate7.py
   chmod +x migrate7.py

2. Run the script in dryrun mode to confirm the expected changes:

   $ ./migrate7.py --dryrun
   
   ---- Migrating the following modules:
   -- module.gke-cluster-dev.module.gke
   ---- Commands to run:
   terraform state mv -state terraform.tfstate "module.gke-cluster-dev.module.gke.google_container_cluster.zonal_primary[0]" "module.gke-cluster-dev.module.gke.google_container_cluster.primary[0]"
   terraform state mv "module.gke-cluster-dev.module.gke.google_container_node_pool.zonal_pools[0]" "module.gke-cluster-dev.module.gke.google_container_node_pool.pools[0]"

3. Execute the migration script

   $ ./migrate7.py
   
   ---- Migrating the following modules:
   -- module.gke-cluster-dev.module.gke
   ---- Commands to run:
   Move "module.gke-cluster-dev.module.gke.google_container_cluster.zonal_primary[0]" to "module.gke-cluster-dev.module.gke.google_container_cluster.primary[0]"
   Successfully moved 1 object(s).
   Move "module.gke-cluster-dev.module.gke.google_container_node_pool.zonal_pools[0]" to "module.gke-cluster-dev.module.gke.google_container_node_pool.pools[0]"
   Successfully moved 1 object(s).
   Move "module.gke-cluster-dev.module.gke.null_resource.wait_for_zonal_cluster" to "module.gke-cluster-dev.module.gke.null_resource.wait_for_cluster"
   Successfully moved 1 object(s).

4. Run terraform plan to confirm no changes are expected.

Null Resource Updates

As part of the upgrade, Terraform might indicate that the wait_for_cluster null_resource must be recreated. This is a no-op which can be safely applied:

  # module.gke.null_resource.wait_for_cluster[0] must be replaced
-/+ resource "null_resource" "wait_for_cluster" {
      ~ id       = "8404887862418893500" -> (known after apply)
      + triggers = {
          + "name"       = "REDACTED"
          + "project_id" = "REDACTED"
        } # forces replacement
    }

Alternatively, if you run into changes with this update, you can remove the original resource from the state entirely:

terraform state rm module.gke.null_resource.wait_for_cluster