Production Deployment

Deploy your application to AWS with ECS Fargate

Your production application runs on AWS ECS Fargate. Resources are defined in infra/prd_resources.py.

Prerequisites

Before deploying, complete these steps:

Step	Guide	Why
Configure secrets	Secrets	API keys and database password
Configure settings	Settings	Region, subnets, image repo
Create ECR repository	Below	Store your Docker images

Deploy to AWS

Once prerequisites are complete:

ag infra up prd:aws

This creates:

• Docker image pushed to ECR
• RDS PostgreSQL instance
• ECS Cluster, Service, and Task Definition
• Application Load Balancer with Target Group
• AWS Secrets Manager entries

RDS takes about 5 minutes to provision. You can monitor progress in the AWS Console.

Verify Deployment

After deployment completes, verify everything is working:

Check the health endpoint

Get your load balancer DNS from the AWS Console (EC2 → Load Balancers), then:

curl http://[LOAD_BALANCER_DNS]/health

Expected response:

{"status": "ok", "instantiated_at": "2025-01-01T12:00:00Z"}

Check CloudWatch logs

aws logs tail /ecs/{infra_name}-prd --follow

Look for Application startup complete to confirm the app started.

Check ECS service status

aws ecs describe-services \
  --cluster {infra_name}-prd \
  --services {infra_name}-prd-service \
  --query 'services[0].{status:status,running:runningCount,desired:desiredCount}'

ECR Authentication

If you haven’t set up ECR yet, see Getting Started - AWS Setup.

ECR tokens expire after 12 hours. Re-authenticate before pushing:

aws ecr get-login-password --region us-east-1 | \
  docker login --username AWS --password-stdin \
  [ACCOUNT_ID].dkr.ecr.us-east-1.amazonaws.com

Or use the helper script:

./scripts/auth_ecr.sh

Update Deployments

After making changes to your code or configuration:

Rebuild and push image

ag infra up prd:docker

Update task definition

Required when you change: image, CPU, memory, or environment variables.

ag infra patch prd:aws:::td

Update service

Triggers a new deployment with the latest task definition:

ag infra patch prd:aws:::service

Shortcut: If you only rebuilt the image (no config changes), you can skip the task definition update and just patch the service.

Customize Your Deployment

Edit infra/prd_resources.py to customize:

Setting	Location	Default
CPU	prd_fastapi	1024 (1 vCPU)
Memory	prd_fastapi	2048 (2 GB)
Task count	ecs_service_count	1
Health check path	health_check_path	/health

Example customization:

prd_resources.py

prd_fastapi = FastApi(
    ...
    ecs_task_cpu="2048",      # 2 vCPU
    ecs_task_memory="4096",   # 4 GB
    ecs_service_count=2,      # 2 tasks for redundancy
)

After changes, update the task definition and service:

ag infra patch prd:aws:::td && ag infra patch prd:aws:::service

Stop Deployment

To remove all AWS resources:

ag infra down prd:aws

This deletes everything including the database. Back up your data first.