ECS queues based autoscaling

Description

This module creates an ECS cluster and service with an autoscaling configuration based on a custom metric that represents the number of messages per task. To update the metric value, the module creates a Event bridge scheduler that runs in a given cron expression and sends a message to a SNS topic. The topic triggers a lambda function that updates the metric value.

Usage

provider "aws" {
  region = "us-east-1"
}

module "test" {
  source = "../"

  account_id          = "12345678912"
  aws_region          = "us-east-1"
  messages_per_task   = 4
  schedule_expression = "rate(2 minutes)"
  queues_prefix       = "teste__"

  topic_name     = "process_queues"
  cluster_name   = "test-scaling"
  create_cluster = true
  service = {
    name          = "test-service"
    desired_count = 1
    autoscaling = {
      max_capacity            = 10
      min_capacity            = 1
      use_alarms              = false
    }
    task_definition = {
      cpu         = 512
      memory      = 1024
      family_name = "test-task"
      container_definitions = [
        {
          name                    = "test-container"
          repository_name         = "example-repository"
          create_repository_setup = false
          portMappings = [
            {
              containerPort = 80
              hostPort      = 80
              protocol      = "tcp"
            }
          ]
        }
      ]
    }
    network = {
      subnets_tags         = {
				key	 = "Name"
				values = ["test-subnet"]
			}
      security_groups_tags = {
				key	 = "Name"
				values = ["test-sg"]
			}
    }
  }
}

Autoscaling method

The autoscaling can be triggered by alarms or using the Autoscaling lambda. This is defined by the variable use_alarms on the example below.

example using alarm:

  service = {
    name          = "test-service"
    desired_count = 1
    autoscaling = {
      max_capacity            = 10
      min_capacity            = 1
      metric_target_value     = 4
      scale_up_cooldown       = 30
      scale_down_cooldown     = 60
      scale_up_alarm_period   = 60
      scale_down_alarm_period = 120
    }

example using only the lambda:

  service = {
    name          = "test-service"
    desired_count = 1
    autoscaling = {
      max_capacity            = 10
      min_capacity            = 1
      use_alarm               = false
    }

Network configuration

You can use the vpc property inside the service to create a new VPC or use an existing one by setting the network property with your subnets and security groups.

Create autoscaling only

If you only want to create autoscaling for an existing service. Set the create_cluster and create_service variables to false. After that, define the cluster name and service name and autoscaling rules