terraform-aws-eks/modules/eks-managed-node-group/main.tf

data "aws_partition" "current" {}

data "aws_caller_identity" "current" {}

################################################################################
# User Data
################################################################################

module "user_data" {
  source = "../_user_data"

  create   = var.create
  platform = var.platform

  cluster_name        = var.cluster_name
  cluster_endpoint    = var.cluster_endpoint
  cluster_auth_base64 = var.cluster_auth_base64

  cluster_service_ipv4_cidr = var.cluster_service_ipv4_cidr

  enable_bootstrap_user_data = var.enable_bootstrap_user_data
  pre_bootstrap_user_data    = var.pre_bootstrap_user_data
  post_bootstrap_user_data   = var.post_bootstrap_user_data
  bootstrap_extra_args       = var.bootstrap_extra_args
  user_data_template_path    = var.user_data_template_path
}

################################################################################
# Launch template
################################################################################

locals {
  use_custom_launch_template = var.launch_template_name != ""
  launch_template_name_int   = coalesce(var.launch_template_name, "${var.name}-eks-node-group")
}

resource "aws_launch_template" "this" {
  count = var.create && var.create_launch_template ? 1 : 0

  name        = var.launch_template_use_name_prefix ? null : local.launch_template_name_int
  name_prefix = var.launch_template_use_name_prefix ? "${local.launch_template_name_int}-" : null
  description = var.launch_template_description

  ebs_optimized = var.ebs_optimized
  image_id      = var.ami_id
  # # Set on node group instead
  # instance_type = var.launch_template_instance_type
  key_name  = var.key_name
  user_data = module.user_data.user_data

  vpc_security_group_ids = compact(concat([try(aws_security_group.this[0].id, "")], var.vpc_security_group_ids))

  default_version         = var.launch_template_default_version
  update_default_version  = var.update_launch_template_default_version
  disable_api_termination = var.disable_api_termination
  # Set on EKS managed node group, will fail if set here
  # https://docs.aws.amazon.com/eks/latest/userguide/launch-templates.html#launch-template-basics
  # instance_initiated_shutdown_behavior = var.instance_initiated_shutdown_behavior
  kernel_id   = var.kernel_id
  ram_disk_id = var.ram_disk_id

  dynamic "block_device_mappings" {
    for_each = var.block_device_mappings
    content {
      device_name  = block_device_mappings.value.device_name
      no_device    = lookup(block_device_mappings.value, "no_device", null)
      virtual_name = lookup(block_device_mappings.value, "virtual_name", null)

      dynamic "ebs" {
        for_each = flatten([lookup(block_device_mappings.value, "ebs", [])])
        content {
          delete_on_termination = lookup(ebs.value, "delete_on_termination", null)
          encrypted             = lookup(ebs.value, "encrypted", null)
          kms_key_id            = lookup(ebs.value, "kms_key_id", null)
          iops                  = lookup(ebs.value, "iops", null)
          throughput            = lookup(ebs.value, "throughput", null)
          snapshot_id           = lookup(ebs.value, "snapshot_id", null)
          volume_size           = lookup(ebs.value, "volume_size", null)
          volume_type           = lookup(ebs.value, "volume_type", null)
        }
      }
    }
  }

  dynamic "capacity_reservation_specification" {
    for_each = var.capacity_reservation_specification != null ? [var.capacity_reservation_specification] : []
    content {
      capacity_reservation_preference = lookup(capacity_reservation_specification.value, "capacity_reservation_preference", null)

      dynamic "capacity_reservation_target" {
        for_each = lookup(capacity_reservation_specification.value, "capacity_reservation_target", [])
        content {
          capacity_reservation_id = lookup(capacity_reservation_target.value, "capacity_reservation_id", null)
        }
      }
    }
  }

  dynamic "cpu_options" {
    for_each = var.cpu_options != null ? [var.cpu_options] : []
    content {
      core_count       = cpu_options.value.core_count
      threads_per_core = cpu_options.value.threads_per_core
    }
  }

  dynamic "credit_specification" {
    for_each = var.credit_specification != null ? [var.credit_specification] : []
    content {
      cpu_credits = credit_specification.value.cpu_credits
    }
  }

  dynamic "elastic_gpu_specifications" {
    for_each = var.elastic_gpu_specifications != null ? [var.elastic_gpu_specifications] : []
    content {
      type = elastic_gpu_specifications.value.type
    }
  }

  dynamic "elastic_inference_accelerator" {
    for_each = var.elastic_inference_accelerator != null ? [var.elastic_inference_accelerator] : []
    content {
      type = elastic_inference_accelerator.value.type
    }
  }

  dynamic "enclave_options" {
    for_each = var.enclave_options != null ? [var.enclave_options] : []
    content {
      enabled = enclave_options.value.enabled
    }
  }

  # Set on EKS managed node group, will fail if set here
  # https://docs.aws.amazon.com/eks/latest/userguide/launch-templates.html#launch-template-basics
  # dynamic "hibernation_options" {
  #   for_each = var.hibernation_options != null ? [var.hibernation_options] : []
  #   content {
  #     configured = hibernation_options.value.configured
  #   }
  # }

  # Set on EKS managed node group, will fail if set here
  # https://docs.aws.amazon.com/eks/latest/userguide/launch-templates.html#launch-template-basics
  # dynamic "iam_instance_profile" {
  #   for_each = [var.iam_instance_profile]
  #   content {
  #     name = lookup(var.iam_instance_profile, "name", null)
  #     arn  = lookup(var.iam_instance_profile, "arn", null)
  #   }
  # }

  dynamic "instance_market_options" {
    for_each = var.instance_market_options != null ? [var.instance_market_options] : []
    content {
      market_type = instance_market_options.value.market_type

      dynamic "spot_options" {
        for_each = lookup(instance_market_options.value, "spot_options", null) != null ? [instance_market_options.value.spot_options] : []
        content {
          block_duration_minutes         = spot_options.value.block_duration_minutes
          instance_interruption_behavior = lookup(spot_options.value, "instance_interruption_behavior", null)
          max_price                      = lookup(spot_options.value, "max_price", null)
          spot_instance_type             = lookup(spot_options.value, "spot_instance_type", null)
          valid_until                    = lookup(spot_options.value, "valid_until", null)
        }
      }
    }
  }

  dynamic "license_specification" {
    for_each = var.license_specifications != null ? [var.license_specifications] : []
    content {
      license_configuration_arn = license_specifications.value.license_configuration_arn
    }
  }

  dynamic "metadata_options" {
    for_each = var.metadata_options != null ? [var.metadata_options] : []
    content {
      http_endpoint               = lookup(metadata_options.value, "http_endpoint", null)
      http_tokens                 = lookup(metadata_options.value, "http_tokens", null)
      http_put_response_hop_limit = lookup(metadata_options.value, "http_put_response_hop_limit", null)
      http_protocol_ipv6          = lookup(metadata_options.value, "http_protocol_ipv6", null)
    }
  }

  dynamic "monitoring" {
    for_each = var.enable_monitoring != null ? [1] : []
    content {
      enabled = var.enable_monitoring
    }
  }

  dynamic "network_interfaces" {
    for_each = var.network_interfaces
    content {
      associate_carrier_ip_address = lookup(network_interfaces.value, "associate_carrier_ip_address", null)
      associate_public_ip_address  = lookup(network_interfaces.value, "associate_public_ip_address", null)
      delete_on_termination        = lookup(network_interfaces.value, "delete_on_termination", null)
      description                  = lookup(network_interfaces.value, "description", null)
      device_index                 = lookup(network_interfaces.value, "device_index", null)
      ipv4_addresses               = lookup(network_interfaces.value, "ipv4_addresses", null) != null ? network_interfaces.value.ipv4_addresses : []
      ipv4_address_count           = lookup(network_interfaces.value, "ipv4_address_count", null)
      ipv6_addresses               = lookup(network_interfaces.value, "ipv6_addresses", null) != null ? network_interfaces.value.ipv6_addresses : []
      ipv6_address_count           = lookup(network_interfaces.value, "ipv6_address_count", null)
      network_interface_id         = lookup(network_interfaces.value, "network_interface_id", null)
      private_ip_address           = lookup(network_interfaces.value, "private_ip_address", null)
      security_groups              = lookup(network_interfaces.value, "security_groups", null) != null ? network_interfaces.value.security_groups : []
      # Set on EKS managed node group, will fail if set here
      # https://docs.aws.amazon.com/eks/latest/userguide/launch-templates.html#launch-template-basics
      # subnet_id                    = lookup(network_interfaces.value, "subnet_id", null)
    }
  }

  dynamic "placement" {
    for_each = var.placement != null ? [var.placement] : []
    content {
      affinity          = lookup(placement.value, "affinity", null)
      availability_zone = lookup(placement.value, "availability_zone", null)
      group_name        = lookup(placement.value, "group_name", null)
      host_id           = lookup(placement.value, "host_id", null)
      spread_domain     = lookup(placement.value, "spread_domain", null)
      tenancy           = lookup(placement.value, "tenancy", null)
      partition_number  = lookup(placement.value, "partition_number", null)
    }
  }

  dynamic "tag_specifications" {
    for_each = toset(["instance", "volume", "network-interface"])
    content {
      resource_type = tag_specifications.key
      tags          = merge(var.tags, { Name = var.name })
    }
  }

  lifecycle {
    create_before_destroy = true
  }

  # Prevent premature access of security group roles and policies by pods that
  # require permissions on create/destroy that depend on nodes
  depends_on = [
    aws_security_group_rule.this,
    aws_iam_role_policy_attachment.this,
  ]

  tags = var.tags
}

################################################################################
# Node Group
################################################################################

locals {
  launch_template_name = try(aws_launch_template.this[0].name, var.launch_template_name, null)
  # Change order to allow users to set version priority before using defaults
  launch_template_version = coalesce(var.launch_template_version, try(aws_launch_template.this[0].default_version, "$Default"))
}

resource "aws_eks_node_group" "this" {
  count = var.create ? 1 : 0

  # Required
  cluster_name  = var.cluster_name
  node_role_arn = var.create_iam_role ? aws_iam_role.this[0].arn : var.iam_role_arn
  subnet_ids    = var.subnet_ids

  scaling_config {
    min_size     = var.min_size
    max_size     = var.max_size
    desired_size = var.desired_size
  }

  # Optional
  node_group_name        = var.use_name_prefix ? null : var.name
  node_group_name_prefix = var.use_name_prefix ? "${var.name}-" : null

  # https://docs.aws.amazon.com/eks/latest/userguide/launch-templates.html#launch-template-custom-ami
  ami_type        = var.ami_id != "" ? null : var.ami_type
  release_version = var.ami_id != "" ? null : var.ami_release_version
  version         = var.ami_id != "" ? null : var.cluster_version

  capacity_type        = var.capacity_type
  disk_size            = local.use_custom_launch_template ? null : var.disk_size # if using LT, set disk size on LT or else it will error here
  force_update_version = var.force_update_version
  instance_types       = var.instance_types
  labels               = var.labels

  dynamic "launch_template" {
    for_each = local.use_custom_launch_template ? [1] : []
    content {
      name    = local.launch_template_name
      version = local.launch_template_version
    }
  }

  dynamic "remote_access" {
    for_each = length(var.remote_access) > 0 ? [var.remote_access] : []
    content {
      ec2_ssh_key               = try(remote_access.value.ec2_ssh_key, null)
      source_security_group_ids = try(remote_access.value.source_security_group_ids, [])
    }
  }

  dynamic "taint" {
    for_each = var.taints
    content {
      key    = taint.value.key
      value  = lookup(taint.value, "value")
      effect = taint.value.effect
    }
  }

  dynamic "update_config" {
    for_each = length(var.update_config) > 0 ? [var.update_config] : []
    content {
      max_unavailable_percentage = try(update_config.value.max_unavailable_percentage, null)
      max_unavailable            = try(update_config.value.max_unavailable, null)
    }
  }

  timeouts {
    create = lookup(var.timeouts, "create", null)
    update = lookup(var.timeouts, "update", null)
    delete = lookup(var.timeouts, "delete", null)
  }

  lifecycle {
    create_before_destroy = true
    ignore_changes = [
      scaling_config[0].desired_size,
    ]
  }

  tags = merge(
    var.tags,
    { Name = var.name }
  )
}

################################################################################
# Security Group
################################################################################

locals {
  security_group_name   = coalesce(var.security_group_name, "${var.name}-eks-node-group")
  create_security_group = var.create && var.create_security_group
}

resource "aws_security_group" "this" {
  count = local.create_security_group ? 1 : 0

  name        = var.security_group_use_name_prefix ? null : local.security_group_name
  name_prefix = var.security_group_use_name_prefix ? "${local.security_group_name}-" : null
  description = var.security_group_description
  vpc_id      = var.vpc_id

  tags = merge(
    var.tags,
    { "Name" = local.security_group_name },
    var.security_group_tags
  )
}

resource "aws_security_group_rule" "this" {
  for_each = { for k, v in var.security_group_rules : k => v if local.create_security_group }

  # Required
  security_group_id = aws_security_group.this[0].id
  protocol          = each.value.protocol
  from_port         = each.value.from_port
  to_port           = each.value.to_port
  type              = each.value.type

  # Optional
  description      = try(each.value.description, null)
  cidr_blocks      = try(each.value.cidr_blocks, null)
  ipv6_cidr_blocks = try(each.value.ipv6_cidr_blocks, null)
  prefix_list_ids  = try(each.value.prefix_list_ids, [])
  self             = try(each.value.self, null)
  source_security_group_id = try(
    each.value.source_security_group_id,
    try(each.value.source_cluster_security_group, false) ? var.cluster_security_group_id : null
  )
}

################################################################################
# IAM Role
################################################################################

locals {
  iam_role_name = coalesce(var.iam_role_name, "${var.name}-eks-node-group")

  iam_role_policy_prefix = "arn:${data.aws_partition.current.partition}:iam::aws:policy"

  cni_policy = var.cluster_ip_family == "ipv6" ? "arn:${data.aws_partition.current.partition}:iam::${data.aws_caller_identity.current.account_id}:policy/AmazonEKS_CNI_IPv6_Policy" : "${local.iam_role_policy_prefix}/AmazonEKS_CNI_Policy"
}

data "aws_iam_policy_document" "assume_role_policy" {
  count = var.create && var.create_iam_role ? 1 : 0

  statement {
    sid     = "EKSNodeAssumeRole"
    actions = ["sts:AssumeRole"]

    principals {
      type        = "Service"
      identifiers = ["ec2.${data.aws_partition.current.dns_suffix}"]
    }
  }
}

resource "aws_iam_role" "this" {
  count = var.create && var.create_iam_role ? 1 : 0

  name        = var.iam_role_use_name_prefix ? null : local.iam_role_name
  name_prefix = var.iam_role_use_name_prefix ? "${local.iam_role_name}-" : null
  path        = var.iam_role_path
  description = var.iam_role_description

  assume_role_policy    = data.aws_iam_policy_document.assume_role_policy[0].json
  permissions_boundary  = var.iam_role_permissions_boundary
  force_detach_policies = true

  tags = merge(var.tags, var.iam_role_tags)
}

# Policies attached ref https://registry.terraform.io/providers/hashicorp/aws/latest/docs/resources/eks_node_group
resource "aws_iam_role_policy_attachment" "this" {
  for_each = var.create && var.create_iam_role ? toset(compact(distinct(concat([
    "${local.iam_role_policy_prefix}/AmazonEKSWorkerNodePolicy",
    "${local.iam_role_policy_prefix}/AmazonEC2ContainerRegistryReadOnly",
    var.iam_role_attach_cni_policy ? local.cni_policy : "",
  ], var.iam_role_additional_policies)))) : toset([])

  policy_arn = each.value
  role       = aws_iam_role.this[0].name
}