engine/lib/aws/bootstrap/eks-workers-nodes.j2.tf

{% for eks_worker_node in eks_worker_nodes %}
resource "aws_eks_node_group" "eks_cluster_workers_{{ loop.index }}" {
  cluster_name           = aws_eks_cluster.eks_cluster.name
  version                = var.eks_k8s_versions.workers
  node_role_arn          = aws_iam_role.eks_workers.arn
  node_group_name_prefix = "qovery-"
  subnet_ids             = flatten([aws_subnet.eks_zone_a[*].id, aws_subnet.eks_zone_b[*].id, aws_subnet.eks_zone_c[*].id])
  instance_types   = ["{{ eks_worker_node.instance_type }}"]
  ami_type         = "AL2_x86_64"
  disk_size = "{{ eks_worker_node.disk_size_in_gib }}"

  tags = merge(
  local.tags_eks,
  {
    "QoveryNodeGroupId" = "${var.kubernetes_cluster_id}-{{ loop.index }}"
    "QoveryNodeGroupName" = "{{ eks_worker_node.name }}"
  }
  )

  scaling_config {
    desired_size = "{{ eks_worker_node.min_nodes }}"
    max_size     = "{{ eks_worker_node.max_nodes }}"
    min_size     = "{{ eks_worker_node.min_nodes }}"
  }

  lifecycle {
    // don't update the desired size and let the cluster-autoscaler do the job
    ignore_changes = [scaling_config[0].desired_size]
    create_before_destroy = true
  }

  update_config {
    max_unavailable_percentage = 10
  }

  timeouts {
    create = "60m"
    delete = "60m"
    update = "60m"
  }

  # Ensure that IAM Role permissions are created before and deleted after EKS Node Group handling.
  # Otherwise, EKS will not be able to properly delete EC2 Instances and Elastic Network Interfaces.
  depends_on = [
    aws_iam_role_policy_attachment.node_AmazonEKSWorkerNodePolicy,
    aws_iam_role_policy_attachment.node_AmazonEKS_CNI_Policy,
    aws_iam_role_policy_attachment.node_AmazonEC2ContainerRegistryReadOnly,
    aws_eks_cluster.eks_cluster,
  ]
}
{% endfor %}