TF 0.13 bug with The "count" value depends on resource attributes that cannot be determined until apply

[bharathkkb]

Terraform Version

0.13.4

Terraform Configuration Files

• Apply this config which will fail due to machine_type = "foo"

provider "google" {
  version = "~> 3.42.0"
  project  = var.project
}

variable "project" {
    type = string
}

resource "google_project_service" "container_api" {
  service                    = "container.googleapis.com"
  disable_on_destroy         = false
  disable_dependent_services = false
}

resource "google_container_cluster" "primary" {
  name     = "my-gke-cluster-1"
  location = "us-central1-a"
  remove_default_node_pool = true
  initial_node_count       = 1
  depends_on = [google_project_service.container_api]
}

resource "google_container_node_pool" "pools" {
  count = 2
  name       = "my-node-pool"
  location   = "us-central1-a"
  cluster    = google_container_cluster.primary.name
  node_count = 1
  node_config {
    machine_type = "foo"
  }
}

resource "random_pet" "pets" {
count = length([for pool in google_container_node_pool.pools : pool.name])
}

• Subsequent terraform apply fails with

The "count" value depends on resource attributes that cannot be determined
until apply, so Terraform cannot predict how many instances will be created.
To work around this, use the -target argument to first apply only the
resources that the count depends on.

Debug Output

Crash Output

N/A

Expected Behavior

Plan should be successful as it is in 0.12.29

Actual Behavior

The "count" value depends on resource attributes that cannot be determined
until apply, so Terraform cannot predict how many instances will be created.
To work around this, use the -target argument to first apply only the
resources that the count depends on.

Steps to Reproduce

Listed above

Additional Context

For 0.13.4 using

terraform apply -refresh=false 

which bypasses the initial refresh seems to be a workaround.

References

x-ref: terraform-google-modules/terraform-google-kubernetes-engine#690
Might be related to

• The "count" value depends on resource attributes that cannot be determined until apply, but resource attributes are already applied #26078

[Gael101]

I do have the same behaviour when running "terraform destroy"on this resource

resource "aws_efs_mount_target" "efs_target" {
    count = local.subnet_count
    file_system_id = aws_efs_file_system.efs.id
    subnet_id      = element(data.terraform_remote_state.vpc_state.outputs.private_subnet_id, count.index)
    security_groups = [data.terraform_remote_state.sg_state.outputs.sg_back_web_id]
}

subnet_count="${length(data.terraform_remote_state.vpc_state.outputs.private_subnet_id)}"

From what i'm seeing in the console , the mount targets were destroyed during the destroy of the below resource

resource "aws_efs_file_system" "efs" {
    creation_token = "${local.project_name}-${local.workspace}-efs"
    tags = {
        project_name   = local.project_name
        env            = local.workspace
    }
}

But this should not rise an error

[jbardin]

Thanks @bharathkkb

The fact that -refresh=false confirms that this is something resolved by the changes in 0.14. There will no longer be a separate refresh phase internally, and this type of skew between configuration and state should no longer be a factor.

Thanks!

[ghost]

I'm going to lock this issue because it has been closed for 30 days ⏳. This helps our maintainers find and focus on the active issues.

If you have found a problem that seems similar to this, please open a new issue and complete the issue template so we can capture all the details necessary to investigate further.