Terraform ECS Scheduler Module

A modular Terraform project for managing ECS service scheduling (Scale Up/Down) with batched Slack notifications.

Features

• Flexible Time Window: Distributes schedule execution over 5 minutes to prevent API throttling
• Batched Notifications: Aggregates multiple ECS events into a single Slack summary message
• Multi-Environment: Separate configurations for dev, stg, prod
• Modular Design: Reusable ecs-scheduler and slack-notifier modules

Directory Structure

terraform-ecs-scheduler-module/
├── modules/
│   ├── ecs-scheduler/           # ECS scheduling module
│   │   ├── main.tf              # IAM Roles, EventBridge Schedule resources
│   │   ├── variables.tf
│   │   └── outputs.tf
│   └── slack-notifier/          # Slack notification module (SQS batching)
│       ├── main.tf              # SQS, Lambda, EventBridge Rule
│       ├── variables.tf
│       ├── outputs.tf
│       └── lambda/
│           └── index.py
├── environments/
│   ├── dev/                     # Development environment
│   └── stg/                     # Staging environment
└── README.md

Architecture

sequenceDiagram
    participant Scheduler as EventBridge Scheduler<br/>(5min flexible window)
    participant ECS as ECS Service
    participant Trail as CloudTrail
    participant Rule as EventBridge Rule
    participant SQS as SQS Queue<br/>(Batching)
    participant Lambda as Lambda Function
    participant Slack as Slack Webhook

    Scheduler->>ECS: UpdateService (Scale Up/Down)
    ECS->>Trail: Log API Call
    Trail->>Rule: Trigger Event
    Rule->>SQS: Send Message
    Note over SQS: Batch messages<br/>(10 msgs or 5 min)
    SQS->>Lambda: Invoke with batch
    Lambda->>Slack: Send Summary Notification

Loading

Usage

1. Select Environment

cd environments/dev   # or stg

2. Configure Backend

Each environment has its own backend.tf with Terraform Cloud workspace:

• dev: ecs-scheduler-dev
• stg: ecs-scheduler-stg

3. Initialize

terraform login
terraform init

4. Configure Variables

Set variables in Terraform Cloud workspace:

• slack_webhook_url (sensitive)
• aws_account_id

5. Plan and Apply

terraform plan
terraform apply

Adding a New Service

Add the following block to environments/<env>/main.tf:

module "new_service_scheduler" {
  source       = "../../modules/ecs-scheduler"
  service_name = "new-service"
  cluster_name = var.cluster_name
  environment  = var.environment

  # Optional: Custom cron expressions (KST timezone)
  # scale_up_cron   = "cron(0 9 ? * MON-FRI *)"
  # scale_down_cron = "cron(0 18 ? * MON-FRI *)"

  # Optional: Disable flexible time window (default: 5 minutes)
  # flexible_time_window_minutes = 0
}

Note: ecs_service_arn is automatically generated from service_name and cluster_name.

Module Variables

ecs-scheduler

Variable	Description	Default
service_name	ECS Service name	required
cluster_name	ECS Cluster name	required
environment	Environment name	"dev"
scale_up_cron	Scale up cron (KST)	"cron(0 0 ? * MON-FRI *)"
scale_down_cron	Scale down cron (KST)	"cron(0 9 ? * MON-FRI *)"
scale_up_count	Desired count for scale up	1
scale_down_count	Desired count for scale down	0
flexible_time_window_minutes	Flexible window (0=OFF)	5

slack-notifier

Variable	Description	Default
environment	Environment name	required
slack_webhook_url	Slack Webhook URL	required
name_prefix	Resource name prefix	"ecs-scheduler"
batch_size	Max messages per batch	10
batching_window_seconds	Max wait time for batching (max 300)	300

Note: batching_window_seconds has a maximum value of 300 seconds (AWS Lambda Event Source Mapping limitation).

Slack Notifications

Summary Notification Example

✅ ECS Scheduled Scaling Completed

Environment: DEV
Total Services: 20 (20 succeeded, 0 failed)

Services:
✅ `nestjs` - Scale Up (-> 1)
✅ `python-api` - Scale Up (-> 1)
✅ `worker` - Scale Up (-> 1)
...

Time: 2025-01-15 09:05:00 KST

Notification States

State	Color	Description
All Succeeded	Green	All services scaled successfully
Partial Failure	Yellow	Some services failed
All Failed	Red	All services failed

Configuration Notes

• Timezone: Cron expressions use Asia/Seoul (KST)
• Backend: Terraform Cloud with per-environment workspaces
• AWS Region: ap-northeast-2 (Seoul)
• Terraform: >= 1.6.0, AWS provider ~> 5.95.0
• Batching: Messages are batched for 5 minutes (max) or until 10 messages accumulate

Known Issues

Notifications may arrive separately when using flexible_time_window

When flexible_time_window_minutes > 0, EventBridge Scheduler executes each service at different times within the window. Combined with CloudTrail event delay (5-15 minutes), this can cause notifications to arrive separately for each service.

Solution: Set flexible_time_window_minutes = 0 to execute all services at the exact same time:

module "service_scheduler" {
  source                       = "../../modules/ecs-scheduler"
  # ...
  flexible_time_window_minutes = 0  # Execute at exact time
}

Trade-off: Disabling flexible time window may increase AWS API call concentration. For most use cases with a small number of services, this is not a concern.