fix(modules,eks): eks-managed nodepool node labels and taints

[g-iannelli]

FIXES:

• For the eks-managed node pool, move labels from kubelet_extra_args to k8s_labels and move taints from kubelet_extra_args to taints.
• For the eks-managed node pool, I fix also the ami_type parameter that is calculated based on the custom ami_type introduced to choose the default EKS ami OS between "Amazon Linux 2" and "amazon Linux 2023"
• Fix the regexp used to fetch AMI ID from the name to include the Amazon Linux arm64 ones and exclude gpu ones

[sbruzzese902]

Nice catch 😄

[smerlos]

During testing, I observed that the AMI ID selection fluctuates between different values, even when no modifications were made to the example provided by the repository. Running terraform plan and terraform apply multiple times consistently led to variations in the assigned AMI ID. This inconsistency has critical implications for the stability and predictability of the deployment.

The selected AMI IDs vary between:

• AMI ID ami-07e1ee9a4c62970c8: "amazon-eks-gpu-node-1.29-v20241106" — a GPU-enabled EKS node image.
• AMI ID ami-0464f6a0d134ee64c: "amazon-eks-node-1.29-v20241106" — a standard EKS node image without GPU support.

Steps to Reproduce

To replicate this behavior:

1. Use the unmodified example from the repository.
2. Execute terraform plan and terraform apply multiple times.
3. Observe the AMI ID assigned to the EKS node pool. It may alternate between the two AMIs noted above, despite there being no changes to the code or configuration.

This inconsistency can be verified with the following commands:

# To verify the GPU-enabled AMI
export AMI_ID=ami-07e1ee9a4c62970c8
aws ec2 describe-images --image-ids $AMI_ID --query 'Images[0].Name' --output json --region eu-west-1 | jq
# Expected output: "amazon-eks-gpu-node-1.29-v20241106"

# To verify the standard AMI
export AMI_ID=ami-0464f6a0d134ee64c
aws ec2 describe-images --image-ids $AMI_ID --query 'Images[0].Name' --output json --region eu-west-1 | jq
# Expected output: "amazon-eks-node-1.29-v20241106"

[smerlos]

After pulling the latest changes on fix/eks-node-pool-k8s-labels-and-taints, I created a cluster using the repository’s base configuration. I then modified the AMI type for the m5-node-pool-self-managed node pool to ami_type = "alinux2023".

However, running terraform plan now results in an error:

Error: Your query returned no results. Please change your search criteria and try again.
   with module.fury_public_example.data.aws_ami.eks_worker_default_ami["m5-node-pool-self-managed"],
   on ../../modules/eks/data.tf line 30, in data "aws_ami" "eks_worker_default_ami":

Could you check the AMI search criteria or confirm alinux2023 compatibility with the m5-node-pool-self-managed configuration?

[smerlos]

Nice Job 💪

LGTM