The terraform-aws-ecs-cluster from obax

About

Terraform module to run ECS cluster, with ASG + Launch Template + Scaling policies via capacity provider. See details in the corresponding AWS blog post Amazon ECS Cluster Auto Scaling is Now Generally Available.

Features

ECS cluster manages ASG capacity automatically.
ASG with optional spot instances support.
It's possible to specify various instance types for your cluster.
EC2 instance profile with SSM policy - you can connect to the instances using the Session Manager.
Default ECS task role allows creating a log group.
Default security group for ECS nodes allow inbound connections from configurable list of network CIDRs.
It's possible to specify additional security groups for ECS nodes.
Latest ECS Optimized AMI with amd64 or arm64 architectures.
Additional EBS disks.
ASG lifecycle hooks.

Usage

Minimal

module "example_ecs_cluster" {
  source       = "github.com/jetbrains-infra/terraform-aws-ecs-cluster?ref=vX.X.X" // see https://github.com/jetbrains-infra/terraform-aws-ecs-cluster/releases
  cluster_name = "FooBar"

  // subnets where the ECS nodes are hosted
  subnets_ids = [
    aws_subnet.private_subnet_1.id,
    aws_subnet.private_subnet_2.id
  ]
}

Full example

module "example_ecs_cluster" {
  source          = "github.com/jetbrains-infra/terraform-aws-ecs-cluster?ref=vX.X.X" // see https://github.com/jetbrains-infra/terraform-aws-ecs-cluster/releases
  cluster_name    = "FooBar"
  spot            = true
  arm64           = true
  target_capacity = 100

  instance_types = {
    "t4g.large"  = 1
    "t4g.xlarge" = 2
  }

  // subnets with ALB and bastion host e.g..
  trusted_cidr_blocks = [
    aws_subnet.public_subnet_1.cidr_block,
    aws_subnet.public_subnet_2.cidr_block
  ]

  ebs_disks = {
    "/dev/sda" = 100
  }

  // subnets where the ECS nodes are hosted
  subnets_ids = [
    aws_subnet.private_subnet_1.id,
    aws_subnet.private_subnet_2.id
  ]

  lifecycle_hooks = [
    {
      name                    = "Example"
      lifecycle_transition    = "autoscaling:EC2_INSTANCE_LAUNCHING"
      default_result          = "CONTINUE"
      heartbeat_timeout       = 2000
      role_arn                = aws_iam_role.example.arn
      notification_target_arn = "arn:aws:sqs:us-east-1:444455556666:queue1"
      notification_metadata   = <<EOF
{
  "foo": "bar"
}
EOF
    }
  ]
}

Required Inputs

The following input variables are required:

cluster_name

Description: Cluster name.

Type: any

subnets_ids

Description: IDs of subnets. Use subnets from various availability zones to make the cluster more reliable.

Type: list(string)

Optional Inputs

The following input variables are optional (have default values):

arm64

Description: ECS node architecture.

Type: bool

Default: false

asg_max_size

Description: The maximum size the auto scaling group (measured in EC2 instances).

Type: number

Default: 100

asg_min_size

Description: The minimum size the auto scaling group (measured in EC2 instances).

Type: number

Default: 0

ebs_disks

Description: A list of additional EBS disks.

Type: map(string)

Default: {}

instance_types

Description: ECS node instance types. Maps of pairs like type = weight. Where weight gives the instance type a proportional weight to other instance types.

Type: map(any)

Default:

{
  "t3a.small": 2
}

lifecycle_hooks

Description: A list of maps containing the name,lifecycle_transition,default_result,heartbeat_timeout,role_arn,notification_target_arn keys.

Type:

list(object({
    name                    = string
    lifecycle_transition    = string
    default_result          = string
    heartbeat_timeout       = number
    role_arn                = string
    notification_target_arn = string
    notification_metadata   = string
  }))