/terraform-aws-ecs-service

Creates an ECS Service.

Primary LanguageHCLBSD 3-Clause "New" or "Revised" LicenseBSD-3-Clause

Terraform module that creates an ECS service with the following features

  • Runs an ECS service with or without an AWS load balancer.
  • Stream logs to a CloudWatch log group encrypted with a KMS key.
  • Associate multiple target groups with Network Load Balancers (NLB) and Application Load Balancers (ALB).
  • Supports running ECS tasks on EC2 instances or Fargate.

Default container definition (hello world app)

We create an initial task definition using the golang:alpine image as a way to validate the initial infrastructure is working: visiting the site shows a simple Go hello world page listening on two configurable ports. This is meant to get a proof of concept instance up and running and to help with testing.

If you want to customize the listener ports for the hello world app, you can modify the hello_world_container_ports variable.

In production usage, we expect deployment tooling to manage the container definitions going forward, not Terraform.

Terraform Versions

Terraform 0.13 and 0.14. Pin module version to ~> 6.0. Submit pull-requests to master branch.

Terraform 0.12 is deprecated.

Usage

ECS service associated with an Application Load Balancer (ALB)

module "app_ecs_service" {
  source = "trussworks/ecs-service/aws"

  name        = "app"
  environment = "prod"

  ecs_cluster                   = aws_ecs_cluster.mycluster
  ecs_vpc_id                    = module.vpc.vpc_id
  ecs_subnet_ids                = module.vpc.private_subnets
  kms_key_id                    = aws_kms_key.main.arn
  tasks_desired_count           = 2

  associate_alb      = true
  alb_security_group = module.security_group.id
  lb_target_groups =
  [
    {
      container_port              = 8443
      container_health_check_port = 8443
      lb_target_group_arn         = module.alb.arn
    }
  ]
}

ECS Service associated with a Network Load Balancer(NLB)

module "app_ecs_service" {
  source = "trussworks/ecs-service/aws"

  name        = "app"
  environment = "prod"

  ecs_cluster                   = aws_ecs_cluster.mycluster
  ecs_vpc_id                    = module.vpc.vpc_id
  ecs_subnet_ids                = module.vpc.private_subnets
  kms_key_id                    = aws_kms_key.main.arn
  tasks_desired_count           = 2

  associate_nlb          = true
  nlb_subnet_cidr_blocks = ["10.0.0.0/24", "10.0.1.0/24", "10.0.2.0/24"]

  lb_target_groups =
  [
    {
      container_port              = 8443
      container_health_check_port = 8080
      lb_target_group_arn         = module.nlb.arn
    }
  ]
}

ECS Service without any AWS load balancer

module "app_ecs_service" {
  source = "trussworks/ecs-service/aws"

  name        = "app"
  environment = "prod"

  ecs_cluster                   = aws_ecs_cluster.mycluster
  ecs_vpc_id                    = module.vpc.vpc_id
  ecs_subnet_ids                = module.vpc.private_subnets
  kms_key_id                    = aws_kms_key.main.arn
}

Requirements

Name Version
terraform >= 0.13
aws >= 3.0

Providers

Name Version
aws >= 3.0

Modules

No Modules.

Resources

Name
aws_cloudwatch_log_group
aws_cloudwatch_metric_alarm
aws_ecs_service
aws_ecs_task_definition
aws_ecs_task_definition
aws_iam_policy_document
aws_iam_role
aws_iam_role_policy
aws_region
aws_security_group
aws_security_group_rule

Inputs

Name Description Type Default Required
additional_security_group_ids In addition to the security group created for the service, a list of security groups the ECS service should also be added to. list(string) [] no
alb_security_group Application Load Balancer (ALB) security group ID to allow traffic from. string "" no
assign_public_ip Whether this instance should be accessible from the public internet. Default is false. bool false no
associate_alb Whether to associate an Application Load Balancer (ALB) with the ECS service. bool false no
associate_nlb Whether to associate a Network Load Balancer (NLB) with the ECS service. bool false no
cloudwatch_alarm_actions The list of actions to take for cloudwatch alarms list(string) [] no
cloudwatch_alarm_cpu_enable Enable the CPU Utilization CloudWatch metric alarm bool true no
cloudwatch_alarm_cpu_threshold The CPU Utilization threshold for the CloudWatch metric alarm number 80 no
cloudwatch_alarm_mem_enable Enable the Memory Utilization CloudWatch metric alarm bool true no
cloudwatch_alarm_mem_threshold The Memory Utilization threshold for the CloudWatch metric alarm number 80 no
cloudwatch_alarm_name Generic name used for CPU and Memory Cloudwatch Alarms string "" no
container_definitions Container definitions provided as valid JSON document. Default uses golang:alpine running a simple hello world. string "" no
container_image The image of the container. string "golang:alpine" no
ecr_repo_arns The ARNs of the ECR repos. By default, allows all repositories. list(string) 
[
  "*"
]
 no
ecs_cluster ECS cluster object for this task. 
object({
    arn  = string
    name = string
  })
 n/a yes
ecs_instance_role The name of the ECS instance role. string "" no
ecs_subnet_ids Subnet IDs for the ECS tasks. list(string) n/a yes
ecs_use_fargate Whether to use Fargate for the task definition. bool false no
ecs_vpc_id VPC ID to be used by ECS. string n/a yes
environment Environment tag, e.g prod. string n/a yes
fargate_platform_version The platform version on which to run your service. Only applicable when using Fargate launch type. string "LATEST" no
fargate_task_cpu Number of cpu units used in initial task definition. Default is minimum. number 256 no
fargate_task_memory Amount (in MiB) of memory used in initial task definition. Default is minimum. number 512 no
health_check_grace_period_seconds Grace period within which failed health checks will be ignored at container start. Only applies to services with an attached loadbalancer. number null no
hello_world_container_ports List of ports for the hello world container app to listen on. The app currently supports listening on two ports. list(number) 
[
  8080,
  8081
]
 no
kms_key_id KMS customer managed key (CMK) ARN for encrypting application logs. string n/a yes
lb_target_groups List of load balancer target group objects containing the lb_target_group_arn, container_port and container_health_check_port. The container_port is the port on which the container will receive traffic. The container_health_check_port is an additional port on which the container can receive a health check. The lb_target_group_arn is either Application Load Balancer (ALB) or Network Load Balancer (NLB) target group ARN tasks will register with. 
list(
    object({
      container_port              = number
      container_health_check_port = number
      lb_target_group_arn         = string
      }
    )
  )
 [] no
logs_cloudwatch_group CloudWatch log group to create and use. Default: /ecs/{name}-{environment} string "" no
logs_cloudwatch_retention Number of days you want to retain log events in the log group. number 90 no
name The service name. string n/a yes
nlb_subnet_cidr_blocks List of Network Load Balancer (NLB) CIDR blocks to allow traffic from. list(string) [] no
service_registries List of service registry objects as per https://www.terraform.io/docs/providers/aws/r/ecs_service.html#service_registries-1. List can only have a single object until hashicorp/terraform-provider-aws#9573 is resolved. 
list(object({
    registry_arn   = string
    container_name = string
    container_port = number
    port           = number
  }))
 [] no
target_container_name Name of the container the Load Balancer should target. Default: {name}-{environment} string "" no
tasks_desired_count The number of instances of a task definition. number 1 no
tasks_maximum_percent Upper limit on the number of running tasks. number 200 no
tasks_minimum_healthy_percent Lower limit on the number of running tasks. number 100 no

Outputs

Name Description
awslogs_group Name of the CloudWatch Logs log group containers should use.
awslogs_group_arn ARN of the CloudWatch Logs log group containers should use.
ecs_security_group_id Security Group ID assigned to the ECS tasks.
task_definition_arn Full ARN of the Task Definition (including both family and revision).
task_definition_family The family of the Task Definition.
task_execution_role The role object of the task execution role that the Amazon ECS container agent and the Docker daemon can assume.
task_execution_role_arn The ARN of the task execution role that the Amazon ECS container agent and the Docker daemon can assume.
task_execution_role_name The name of the task execution role that the Amazon ECS container agent and the Docker daemon can assume.
task_role The IAM role object assumed by Amazon ECS container tasks.
task_role_arn The ARN of the IAM role assumed by Amazon ECS container tasks.
task_role_name The name of the IAM role assumed by Amazon ECS container tasks.

Upgrade Path

5.x.x to 6.0.0

In versions 5.x.x and prior, the following resources existed as arrays (toggled by a count meta-argument). With 6.0.0, each pair has been merged into a single resource.

  • aws_cloudwatch_metric_alarm.alarm_cpu[0] xor aws_cloudwatch_metric_alarm.alarm_cpu_no_lb[0] -> aws_cloudwatch_metric_alarm.alarm_cpu
  • aws_cloudwatch_metric_alarm.alarm_mem[0] xor aws_cloudwatch_metric_alarm.alarm_mem_no_lb[0] -> aws_cloudwatch_metric_alarm.alarm_mem
  • aws_ecs_service.main[0] xor aws_ecs_service.main_no_lb[0] -> aws_ecs_service.main

To upgrade to 6.0.0, you will need to perform a terraform state mv for any affected resources to avoid destruction and recreation. Alternatively, you can let Terraform delete/recreate the deployed resources.

For example, if you are using this module and naming it example, you could run one or more of the commands as appropriate given your environment:

# Example alarm_cpu state mv commands (pick the relevant one for your environment):
terraform state mv 'module.example.aws_cloudwatch_metric_alarm.alarm_cpu[0]' 'module.example.aws_cloudwatch_metric_alarm.alarm_cpu'
terraform state mv 'module.example.aws_cloudwatch_metric_alarm.alarm_cpu_no_lb[0]' 'module.example.aws_cloudwatch_metric_alarm.alarm_cpu'

# Example alarm_mem state mv commands (pick the relevant one for your environment):
terraform state mv 'module.example.aws_cloudwatch_metric_alarm.alarm_mem[0]' 'module.example.aws_cloudwatch_metric_alarm.alarm_mem'
terraform state mv 'module.example.aws_cloudwatch_metric_alarm.alarm_mem_no_lb[0]' 'module.example.aws_cloudwatch_metric_alarm.alarm_mem'

# Example main state mv commands (pick the relevant one for your environment):
terraform state mv 'module.example.aws_ecs_service.main[0]' 'module.example.aws_ecs_service.main'
terraform state mv 'module.example.aws_ecs_service.main_no_lb[0]' 'module.example.aws_ecs_service.main'

5.x.x to 5.1.1

With 5.1.1, the hashicorp/aws provider must be a minimum version of 3.0. It no longer has a maximum version. Therefore any code calling this will need to accomodate for that minimum version change.

4.0.0 to 5.0.0

Prior to 5.x, the hashicorp/aws provider required the use of version 2.70. 5.x changes the hashicorp/aws provider so that it can be greater than or equal to 2.70, but must be less than 4.0. Therefore any code calling this will need to accomodate for that version change.

3.x.x to 4.0.0

3.x.x uses Terraform 0.12.x and 4.0.0 uses Terraform 0.13.x, so this requires upgrading any code that uses this module to Terraform 0.13.x.

2.x.x to 3.0.0

In 3.0.0 the module added support for multiple load balancer target groups. To support this change, container_port, container_health_check_port and lb_target_group are being replaced with lb_target_groups.

Without a load balancer

If you are using this module without an ALB or NLB then you can remove any references to container_port, container_health_check_port and lb_target_group if you were doing so.

Using with ALB or NLB target groups

If you are using an NLB or NLB target groups with this module then you will need replace the values of container_port, container_health_check_port and lb_target_group with

Below is an example of how the module would be instantiated prior to version 3.0.0:

module "app_ecs_service" {
  source = "trussworks/ecs-service/aws"
  ...
  container_port                  = 8443
  container_health_check_port     = 8080
  lb_target_group_arn             = module.alb.arn
  ...
}

In 3.0.0 the same example will look like the following

module "app_ecs_service" {
  source = "trussworks/ecs-service/aws"
  ...
  lb_target_groups =
  [
    {
      container_port                  = 8443
      container_health_check_port     = 8080
      lb_target_group_arn             = module.alb.arn
    }
  ]
  ...
}

2.0.0 to 2.1.0

In 2.1.0 KMS log encryption is required by default. This requires that you create and attach a new AWS KMS key ARN. As an example here is how to set that up (please review on your own):

data "aws_iam_policy_document" "cloudwatch_logs_allow_kms" {
  statement {
    sid    = "Enable IAM User Permissions"
    effect = "Allow"

    principals {
      type = "AWS"
      identifiers = [
        "arn:aws:iam::${data.aws_caller_identity.current.account_id}:root",
      ]
    }

    actions = [
      "kms:*",
    ]
    resources = ["*"]
  }

  statement {
    sid    = "Allow logs KMS access"
    effect = "Allow"

    principals {
      type        = "Service"
      identifiers = ["logs.us-west-2.amazonaws.com"]
    }

    actions = [
      "kms:Encrypt*",
      "kms:Decrypt*",
      "kms:ReEncrypt*",
      "kms:GenerateDataKey*",
      "kms:Describe*"
    ]
    resources = ["*"]
  }
}

resource "aws_kms_key" "main" {
  description         = "Key for ECS log encryption"
  enable_key_rotation = true

  policy = data.aws_iam_policy_document.cloudwatch_logs_allow_kms.json
}

NOTE: Best practice is to use a separate KMS key per ECS Service. Do not re-use KMS keys if it can be avoided.

1.15.0 to 2.0.0

v2.0.0 of this module is built against Terraform v0.12. In addition to requiring this upgrade, the v1.15.0 version of the module took the name of the ECS cluster as a parameter; v2.0.0 takes the actual object of the ECS cluster as a parameter instead. You will need to update previous instances of this module with the altered parameter.

Developer Setup

Install dependencies (macOS)

brew install pre-commit go terraform terraform-docs

Testing

Terratest is being used for automated testing with this module. Tests in the test folder can be run locally by running the following command:

make test

Or with aws-vault:

AWS_VAULT_KEYCHAIN_NAME=<NAME> aws-vault exec <PROFILE> -- make test