AWS Docker deployment

1FROM python:3.12-slim
2WORKDIR /app
3COPY requirements.txt .
4RUN pip install --no-cache-dir -r requirements.txt
5COPY . .
6CMD ["python", "app.py"]

1docker build -t myapp:latest .
2aws ecr create-repository --repository-name myapp
3aws ecr get-login-password --region us-east-1 \
4| docker login --username AWS --password-stdin 123456789012.dkr.ecr.us-east-1.amazonaws.com docker tag myapp:latest 123456789012.dkr.ecr.us-east-1.amazonaws.com/myapp:latest docker push 123456789012.dkr.ecr.us-east-1.amazonaws.com/myapp:latest ``` At that point AWS can pull the image during deployment. ## Deploy With ECS Fargate In ECS, you normally define a task, then run it as a service. The task definition describes the container image, CPU and memory allocation, port mapping, environment variables, and log configuration. Minimal task definition excerpt: ```json { "family": "myapp", "networkMode": "awsvpc", "requiresCompatibilities": ["FARGATE"], "cpu": "256", "memory": "512", "containerDefinitions": [ { "name": "myapp", "image": "123456789012.dkr.ecr.us-east-1.amazonaws.com/myapp:latest", "essential": true, "portMappings": [ { "containerPort": 8080, "protocol": "tcp" } ], "logConfiguration": { "logDriver": "awslogs", "options": { "awslogs-group": "/ecs/myapp", "awslogs-region": "us-east-1", "awslogs-stream-prefix": "ecs" } } } ] } ``` You would then attach the ECS service to a VPC, subnets, and usually an Application Load Balancer for incoming traffic. ## Treat AWS Deployment as an Ops Problem, Not Just a Docker Problem A successful deployment is usually decided by the surrounding AWS configuration: - IAM roles for pulling images and accessing AWS resources - security groups for inbound and outbound traffic - CloudWatch logs for debugging startup failures - Secrets Manager or Parameter Store for configuration - health checks so bad tasks are replaced automatically This is why “my Docker app works locally” does not guarantee a good AWS deployment. Local Docker proves the image builds and starts. AWS still needs the network and identity pieces to be correct. ## A Useful Local-to-AWS Progression A practical deployment flow is: 1. run the container locally with the same `CMD` used in production 2. push the image to ECR 3. create an ECS task definition 4. run one task manually and inspect logs 5. create an ECS service and attach a load balancer if it is a web app 6. add autoscaling only after the basic service is stable That sequence catches configuration errors earlier than trying to automate everything at once. ## When EKS Makes Sense EKS is appropriate when you already operate Kubernetes or need Kubernetes-native features such as custom controllers, Helm-based deployment workflows, or portability across clusters. If you are not already in that world, using EKS only to run one or two Docker services usually adds complexity without real benefit. The container image is the same either way; the difference is the control plane and operational model around it. ## Common Pitfalls A common mistake is focusing on the Dockerfile while ignoring IAM. If the task cannot pull from ECR or read secrets, the deployment fails even though the image is valid. Another issue is using ECS Fargate without checking memory and CPU limits. Containers that fit locally may be killed quickly in AWS if the task size is too small. The third common problem is skipping logs and health checks. Without CloudWatch logs and load balancer health checks, diagnosing startup failures becomes slow and guess-heavy. ## Summary - On AWS, Docker deployment usually means ECR plus ECS or EKS, not just uploading an image. - ECS with Fargate is often the simplest managed path for standard services. - Push images to ECR, then deploy them through task definitions and services. - IAM, networking, secrets, and logging are as important as the container itself. - Stabilize one running task first, then add load balancing and autoscaling.