Before the new flag got introduced the EKS cluster was implicitely created with bootstrapping addons.
Now when keeping the default value (bootstrap_self_managed_addons is set to true per default), a recreate of the EKS cluster is forced.
When configuring the value bootstrap_self_managed_addons to false, we end up with an EKS cluster without the addons.
Both is imho wrong, this is a breaking change and does not allow us to move forward.
Request: Consider to implement logic which does not force a recreation of existing clusters while not breaking their functionality.
Versions
Module version 5.65.0
Terraform version: OpenTofu v1.8.1 on linux_amd64
Steps to reproduce the behavior:
Apply EKS Cluster using AWS provider version e.g. 5.51.1.
Apply the same code using AWS provider version e.g. 5.65.0.
This results in
# module.eks_cluster.aws_eks_cluster.eks_cluster must be replaced
--
# Warning: this will destroy the imported resource
-/+ resource "aws_eks_cluster" "eks_cluster" {
~ arn = "arn:aws:eks:us-east-1:123456789012:cluster/mycluster" -> (known after apply)
~ bootstrap_self_managed_addons = false -> true # forces replacement
Expected behavior
The cluster is not recreated while maintaining its functionality.
Actual behavior
The cluster is attempted to be recreated.
Flipping configuration value to false causes degraded functionality as pieces such as coredns are no more available as before.
Description
The issue https://github.com/terraform-aws-modules/terraform-aws-eks/issues/3143 and related pull request https://github.com/terraform-aws-modules/terraform-aws-eks/pull/3099 (introduced via provider version 5.58.0 causes a severe incompatibility which we fail to address.
Before the new flag got introduced the EKS cluster was implicitely created with bootstrapping addons. Now when keeping the default value (bootstrap_self_managed_addons is set to true per default), a recreate of the EKS cluster is forced. When configuring the value bootstrap_self_managed_addons to false, we end up with an EKS cluster without the addons.
Both is imho wrong, this is a breaking change and does not allow us to move forward.
Request: Consider to implement logic which does not force a recreation of existing clusters while not breaking their functionality.
Versions
Steps to reproduce the behavior:
Expected behavior
The cluster is not recreated while maintaining its functionality.
Actual behavior
The cluster is attempted to be recreated. Flipping configuration value to
false
causes degraded functionality as pieces such as coredns are no more available as before.