Hackathon

Here is a raw, honest project story that reflects the real-world struggle of moving from a simple app to a complex Docker environment. No fluff—just the reality of what happened.

Inspiration We wanted to see if we could take a standard Flask app and turn it into a "production-grade" system. Everyone says "just use Docker," but we wanted to actually experience the transition from a local script to a load-balanced fleet of containers.

What it does It’s a URL management API built with Flask and Peewee ORM. On the surface, it handles URL data, but under the hood, it’s designed to be a distributed system with an Nginx load balancer sitting in front of three identical "clones" of the app, all connected to a single PostgreSQL database.

How we built it We used a "Silver Tier" architecture:

Backend: Flask for the API logic.

Database: PostgreSQL for persistent storage.

ORM: Peewee to manage our models (User, Url, Event).

Infrastructure: Docker Compose to orchestrate five different containers (1 Nginx, 3 Web Clones, 1 DB).

Server: Gunicorn to handle multiple workers in production.

Challenges we ran into This is where things got real. We faced a 100% failure rate on our first major load tests.

The Docker "Hang": Our terminal kept freezing because the containers were stuck in an install loop. All three clones tried to download dependencies at the same time, choking the system.

502 Bad Gateway: Nginx was up, but it couldn't talk to the app. The app was crashing on start because it was trying to query the database before the database was actually ready to accept connections.

The Blank Slate: Even when the containers finally stayed up, the API returned errors because the database tables didn't exist. We realized that in Docker, you have to manually trigger the creation of your schema; it doesn't just happen by magic.

Accomplishments that we're proud of We’re proud of the fact that we didn't just give up when the screen stayed black and the terminal hung. We successfully diagnosed the "race condition" between our database and our web workers. We managed to get a multi-container environment configured correctly, even if the road there was full of 502 errors.

What we learned We learned that Infrastructure is the hardest part of software. Writing the Python code took an hour; getting the Docker networking and database synchronization to work took all day. We learned the importance of "pre-baking" dependencies into Docker images so they don't have to download every time the app starts.

What's next for Hackathon Next, we’re going for the Gold Tier. We are moving away from local Docker and deploying straight to the DigitalOcean App Platform. We’re taking everything we learned about database initialization and Gunicorn workers to build a cloud-native version that can actually survive a 200-user load test without hanging.

Built With

• sweat