Empty list in cluster_compute_config.nodepools forces cluster replacement

[ChrisEke]

Description

Setting cluster_compute_config.node_pools = [] in an existing EKS Auto mode cluster that was created with built-in nodepools cluster_compute_config.node_pools = ["general-purpose", "system"] forces replacement of the whole cluster.

From what I can tell the replacement is caused by the module changing compute_config.node_role_arn to null.

The use case here is that some EKS Auto mode clusters can start out with the built-in nodepools but might later opt to to disable the built-in and only use custom nodepools via k8s API. The AWS documentation mentions how to do this with the CLI and I would hope to achieve the same using Terraform: https://docs.aws.amazon.com/eks/latest/userguide/set-builtin-node-pools.html

• ✋ I have searched the open/closed issues and my issue is not listed.

Versions

• Module version [Required]:
  20.34.0
• Terraform version:
  Terraform v1.11.1
  on darwin_arm64
• Provider version(s):
  registry.terraform.io/hashicorp/aws v5.84.0

Reproduction Code [Required]

Steps to reproduce the behavior:

1. Have a previously created cluster with:

   cluster_compute_config = {
     enabled       = true
     node_pools    = ["general-purpose", "system"]
   }

2. Remove the default node_pools:

   cluster_compute_config = {
     enabled       = true
     node_pools    = []
   }

Expected behavior

The built-in node pools should be disabled in the cluster while compute_config.node_role_arn is unchanged.

Actual behavior

terraform plan
# module.eks.aws_eks_cluster.this[0] must be replaced
+/- resource "aws_eks_cluster" "this" {
  ...
      ~ compute_config {
          - node_pools    = [
              - "general-purpose",
              - "system",
            ] -> null
          - node_role_arn = "arn:aws:iam::123456789123:role/my-cluster-eks-auto-20250217081502451000000001" -> null # forces replacement
            # (1 unchanged attribute hidden)
        }
  ...
}

[bryantbiggs]

duplicate #3273

[ChrisEke]

A workaround to at least disable the "general-purpose" nodepool is to remove it as an element while "system" remains in the array. This successfully removes the "general-purpose" from the EKS cluster on the next terraform apply, without forcing replacement of the EKS cluster.

  cluster_compute_config = {
    enabled = true
    node_pools =  ["system"]
  }

[github-actions]

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.