Auto Scaling

Overview

The Express Gateway Service Express Gateway Service
The aws_ecs_express_gateway_service Terraform resource that provisions an ECS service with managed ALB, auto scaling, and simplified configuration for HTTP/HTTPS web applications and APIs. includes built-in target tracking auto scaling through the scaling_target block. Unlike classic ECS where you configure separate Application Auto Scaling resources, Express Mode handles the scaling infrastructure automatically.

Note

Express Mode uses target tracking scaling exclusively. Step scaling and scheduled scaling are not available.

The scaling_target Block

main.tf

resource "aws_ecs_express_gateway_service" "this" {
  # ...

  scaling_target {
    auto_scaling_metric       = "CPU"
    auto_scaling_target_value = 70
    min_task_count            = 1
    max_task_count            = 3
  }
}

Parameter Reference

Parameter	Type	Allowed Values	Source Repo Value
auto_scaling_metric	string	“CPU”, “Memory"	"CPU”
auto_scaling_target_value	number	1–100 (percentage)	70
min_task_count	number	≥ 1	1
max_task_count	number	≥ min_task_count	3

Tip

The source repo uses CPU at 70%. For memory-intensive workloads like Bedrock inference, consider using Memory instead.

How Target Tracking Works

When you set auto_scaling_metric = "CPU" and auto_scaling_target_value = 70, Express Mode:

1. Monitors average CPU utilization across all running tasks
2. Adds tasks when average exceeds 70%
3. Removes tasks when average drops below 70% (with cooldown)
4. Maintains count between min_task_count and max_task_count

Caution

Setting min_task_count to 1 means no redundancy during scale-in. For production, consider a minimum of 2 to maintain availability during task replacements.

Express Mode vs Classic Auto Scaling

Classic ECS requires multiple resources for auto scaling:

• aws_appautoscaling_target — registers the service as scalable
• aws_appautoscaling_policy — defines the tracking policy
• CloudWatch alarms — created automatically by the policy

With Express Mode, the scaling_target block replaces all of these. Express Mode also provisions the RollbackAlarm RollbackAlarm
A CloudWatch metric alarm created by Express Mode that triggers automatic deployment rollback when new tasks fail ALB health checks. that triggers deployment rollback when new tasks fail health checks.