Mini-RAFT Distributed Drawing Board

A powerful distributed systems project simulating cloud-native consensus models using a miniature version of the RAFT protocol. Multiple backend replicas maintain an identical event log of drawing strokes without relying on a central database, ensuring zero downtime even when containers crash.

RAFT Theory Overview

• Leader Election: All nodes start as Followers. If they haven't heard from a Leader before a randomized timer expires, they become a Candidate and request votes. The first to achieve a majority becomes the Leader.
• Log Replication: The Leader is the only node trusted to accept client data. It appends the data (drawing strokes) to an immutable log and blasts it to Followers. Once a majority acknowledge, the state is committed.
• Fault Tolerance: If a node fails, it restarts and attempts to re-sync. If the Leader fails, followers notice the lack of heartbeats and securely elect a new one without disrupting data persistence.

Architecture Overview

graph TD
    UI[HTML Canvas UI] -->|WebSocket| G(Gateway Proxy)
    G -->|POST /submit-stroke| L[Leader Node: 3001]
    L -.->|RPC /append-entries| F1[Follower Node: 3002]
    L -.->|RPC /append-entries| F2[Follower Node: 3003]
    L -.->|RPC /append-entries| F3[Follower Node: 3004]

Loading

Setup & Instructions

Backend Start

1. Ensure Docker Desktop is running.
2. At the root of the project, run:

docker-compose up --build

This boots the Gateway (Port 3000) and your Replica Nodes (3001+).

Frontend Start

1. In a separate terminal, navigate into the frontend folder:

cd frontend
npm install
npm run serve

2. Open http://localhost:8080 in multiple browser tabs to simulate clients.

Assumptions & Constraints

• All system instances run over http inside local Docker bridges. Network latencies are artificially mimicked.
• Logs are held in memory on the replicas and are not persisted entirely to physical disk arrays between total cluster teardowns (docker-compose down).

---

MiniRAFT Project Checklist

• Leader election
• WebSocket Gateway
• Basic log replication
• Drawing canvas integration
• Graceful reload
• Blue-green replica replacement
• Failover correctness
• Multi-client real-time sync
• Demo under chaotic conditions

Bonus Options:

• Network partitions (simulate split brain)
• Add 4th replica
• Add vector-based undo/redo using log compensation
• Implement a dashboard showing leader, term, log sizes
• Deploy to a real cloud VM (e.g., Amazon Web Services (AWS) EC2 or Google Cloud)