Scaling Kanban Prototype

This project is a prototype of a scaling architecture for a collaborative Kanban web app with live updates via websockets.

Why?

I did this to gain more experience with technologies I was curious about:

• Docker compose
• Rust backend
• Websockets
• Pub/Sub
• Distributing load between multiple backends

I did not create unit tests as this is just a small prototype

How to run

0. Install docker.
1. Clone the repository.
2. Run the run-demo.sh bash file. This starts all services including 2 backend instances, edit the file to use a higher number of backend instances (change backend=2 to backend=6 for example).
3. Open http://localhost:3500 and select a board.
4. Open the board in multiple tabs and make changes to the board. The changes should be synchronized between all tabs.
5. The backend logs (docker compose logs backend -f) should indicate that multiple backends are serving clients.

Architecture

I wanted to create a solution that would support editing of the same kanban boards by multiple users across multiple backend instances.

Reverse proxy & load balancer: Traefik

Was really easy to set up: The configuration is completely inside the docker compose file. Traefik also reacts to up and down scaling without restarts and the configuration can be inspected via a frontend on http://localhost:8080.

Backend: Axum (Rust)

Seemed like a good alternative to Actix (it is developed by the Tokio team).

Frontend: React (TypeScript)

I created this frontend 2 years ago and reused it for this prototype (it still uses react-beautiful-dnd).

Message Service (Pub/Sub): Nats.io

Nats claims to be easy to set up (it was) and I just wanted to try it out.

Database: MongoDB

I decided to use MongoDB as it is easy to set up and to adjust the data layout.

This is how it works:

1. Load balancer distributes incoming requests for the backend to the available backend instances
2. Frontend establishes Websocket connection to a backend instance for the current board. The frontend will then receive updates to the board made by other users via this connection.
3. User changes to the board are sent from frontend to a backend instance via PUT calls
4. Backend instance modifies the boards document in the MongoDB collection, increments a version field and sends updated board in JSON response to frontend.
5. Backend instance then publishes a message to Pub/Sub containing the new JSON representation of the board
6. Pub/Sub distributes the updated kanban board JSON representation to all subscribed backend instances
7. Subscribed backend instances notify connected frontends about change to the kanban board via the Websocket connections of step 2
8. All notified frontends update their board state if the version of the received board is newer