Scribbly

A minimal blogging website using microservices deployed on AWS.

Architecture

The project uses a microservice architecture inside a monorepo.

Services:

• Web: Frontend built with Next.js
• Auth: Authentication service
• User: User management service
• Post: Blog post management service
• Postgres: Auth and User db
• MongoDB: Post db
• Redis: for caching and refresh tokens
• RabbitMQ: communication between services

All backend services are built using Fastify. The project uses a pnpm workspace.

Project Structure

scribbly/
|
├── .infra/                 # Terraform configs
|
├── packages/
│   └── common/             # Shared code and utilities
│
├── services/
│   ├── auth/
│   ├── post/
│   ├── user/
│   ├── web/
│   └── nginx/
│
├── compose.yml             # For local development
├── compose.prod.yml
|
├── package.json
└── pnpm-workspace.yaml

Local Development

For local development, the services are run using Docker Compose.

docker compose up --build

It uses docker images for Postgres, MongoDB, Redis, RabbitMQ and Nginx.

The application services are built using their respective Dockerfiles.

Nginx has been used as the reverse proxy.

AWS Deployment

The AWS deployment uses managed services equivalent to the local development environment:

Local	AWS
Postgres	RDS
MongoDB	MongoDB on EC2**
Redis	ElastiCache
RabbitMQ	AmazonMQ
Nginx	API Gateway

^{** The EC2 instance was used for MongoDB as DocumentDB is not available in free tier.}

The images are uploaded to ECR. The services are deployed to ECS on Fargate.

Terraform

The deployment is managed using Terraform (see ./.infra).

To spin up the deployment:

terraform apply

To take down the deployment:

terraform destroy

Make sure to run the commands from .\.infra folder.

CI/CD

Deployment is automated using GitHub Actions.

• infra.yml: creates or updates AWS infrastructure.
• deploy.yml: builds and deploys services. (Build Docker images > Push images to ECR > Redeploy ECS)

GitHub Actions authenticates with AWS using OIDC (OpenID Connect).

The project has minimal functionality as it was an exercise for learning microservices, AWS and terraform.

Note: The AWS infrastructure is usually shut down due to limited free tier and can be spun up when required.