Terraform 0.13 Brings Powerful Meta-Arguments to Modular Workflows

The upcoming Terraform 0.13 release includes powerful new meta-arguments for modular workflows including count, for_each and depends_on. These language features can simplify and streamline Terraform configurations at the resource-level, and they are now available for module-centric workflows.

In previous Terraform versions, the for_each and count features have allowed for systematically creating multiple resource instances from a single resource block based on data from elsewhere in the module:

variable "vpc_id" {
    type = string
}

variable "subnets" {
  type = map(object({
    cidr_block        = string
    availability_zone = string
  }))
}

resource "aws_subnet" "example" {
  for_each = var.subnets

  cidr_block        = each.value.cidr_block
  availability_zone = each.value.availability_zone
  tags = {
    Name = each.key
  }
}

Terraform 0.13 introduces a similar capability for entire modules, allowing a single module block to produce multiple module instances systematically:

# Illustrative example only

locals {
  resources = {
    eks-prod = "prod-eks"
    eks-qa   = "qa-eks"
    eks-dev  = "dev-eks"
  }
}

module "my-cluster" {
  source   = "terraform-aws-modules/eks/aws"
  for_each = local.resources

  cluster_name    = each.value
  cluster_version = "1.17"
  ...
  worker_groups = [
    {
      name          = each.key
      instance_type = var.instance_type
      asg_max_size  = 5
    }
  ]
}

In previous versions of Terraform, module instances have served only as separate namespaces, and have not been nodes in Terraform's dependency graph themselves. Terraform has always tracked dependencies via the input variables and output values of a module, but users have frequently requested a concise way to declare that all objects inside a module share a particular dependency in the calling module. Terraform 0.13 introduces this capability by allowing depends_on as a meta-argument inside module blocks:

resource "aws_iam_policy_attachment" "example" {
  name       = "example"
  roles      = [aws_iam_role.example.name]
  policy_arn = aws_iam_policy.example.arn
}
 
module "uses-role" {
  # ...
 
  depends_on = [aws_iam_policy_attachment.example]
}

While we are happy to have made depends_on available for modular workflows, we recommend considering this a last resort and using data flow to imply dependencies wherever possible. Allowing Terraform to infer dependencies automatically will tend to make your configuration easier to maintain, and will allow Terraform to maximize concurrency when making many changes in a single operation.

We are very excited to be able to bring this first iteration of these in-demand features to our community of Terraform practitioners. Download Terraform 0.13 today, and try count and for_each by following our step-by-step tutorials. Learn more by reading the documentation, and draft upgrade guide, and join our community forum to tell us all about it.