GKE blue_green_settings batch_node_count & batch_percentage is not idempotent

[tumd]

Community Note

• Please vote on this issue by adding a 👍 reaction to the original issue to help the community and maintainers prioritize this request.
• Please do not leave +1 or me too comments, they generate extra noise for issue followers and do not help prioritize the request.
• If you are interested in working on this issue or have submitted a pull request, please leave a comment.
• If an issue is assigned to a user, that user is claiming responsibility for the issue.
• Customers working with a Google Technical Account Manager or Customer Engineer can ask them to reach out internally to expedite investigation and resolution of this issue.

Terraform Version & Provider Version(s)

Terraform v1.9.8
on darwin_arm64

• provider registry.terraform.io/hashicorp/google v6.10.0

Affected Resource(s)

google_container_node_pool

Terraform Configuration

variable "project_id" {}
variable "location" { default = "us-central1" }
variable "cluster_name" { default = "my-gke-cluster" }

resource "google_service_account" "default" {
  project      = var.project_id
  account_id   = "service-account-id"
  display_name = "Service Account"
}

resource "google_container_cluster" "primary" {
  project  = var.project_id
  name     = var.cluster_name
  location = var.location

  # 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
  # node pool and immediately delete it.
  remove_default_node_pool = true
  initial_node_count       = 1
}

resource "google_container_node_pool" "primary_preemptible_nodes" {
  project    = var.project_id
  name       = "my-node-pool"
  cluster    = google_container_cluster.primary.id
  node_count = 2

  node_config {
    preemptible  = true
    machine_type = "e2-medium"

    # Google recommends custom service accounts that have cloud-platform scope and permissions granted via IAM Roles.
    service_account = google_service_account.default.email
    oauth_scopes = [
      "https://www.googleapis.com/auth/cloud-platform"
    ]
  }
  upgrade_settings {
    blue_green_settings {
      node_pool_soak_duration = "100s"

      standard_rollout_policy {
        batch_node_count    = 0
        batch_soak_duration = "0s"
      }
    }

    strategy = "BLUE_GREEN"
  }
}

data "google_container_cluster" "primary" {
  project  = var.project_id
  name     = var.cluster_name
  location = var.location
}

output "node_pools" {
  value = data.google_container_cluster.primary.node_pool[0].upgrade_settings
}

Debug Output

No response

Expected Behavior

batch_node_count should be set to 0.

Actual Behavior

batch_node_count is set to 1.

Steps to reproduce

1. terraform apply
2. terraform apply

Important Factoids

No response

References

No response

b/379359342

[tumd]

Perhaps far fetched, but wanted to mention it here too, just in case some Googler might pick it up;

The batch_node_count set to 0 seem to be a bit buggy overall, which potentially could cause this bug. See this bug-report.

[ggtisc]

Confirmed issue!

Both terraform apply and terraform plan allow the value 0 for the batch_node_count argument but in the end assign it a value of 1.

In the terraform registry and API documentation there is no mention of whether it is allowed or not. For his part, the user shares this [link](See this bug-report) with additional information

As for the batch_percentage argument, it allows a value of 0 and assigns it correctly in the tfstate file successfully

[wangzhen127]

Both terraform apply and terraform plan allow the value 0 for the batch_node_count argument but in the end assign it a value of 1.

Do you mean the Terraform wouldn't let you to change the batch_node_count field value to 0? But from #20283 (comment), it showed that the value is 0.

I tried with gcloud, the blue pool draining phase is correctly skipped.

[VeraQin]

Zero values for scalar fields (0, 0.0, false, "", the zero value of an enum) are treated the same as unset in terraform. And the default value for "batch_node_count" is 1 if not provided.

Will fix it by adding property ForceSendFields (similar to sendEmptyValue) to the field batch_node_count in file. With the flag, the provider sends “empty” values (such as zero, false, or empty strings) to the API if set explicitly in the user’s configuration.