Raft Consensus

An implementation of the Raft distributed consensus protocol in Rust, built by working through the original paper. A learning project for understanding distributed systems from first principles.

The simulation above runs a live 5-node Raft cluster compiled to WASM. Click any node to kill or revive it. Use Partition to introduce a network split between two nodes. Type a command and hit Submit to send it to the current leader — watch it replicate across the cluster.

What's Implemented

• Leader election — randomized timeouts, vote counting, term-based ordering
• Log replication — AppendEntries RPC with conflict resolution
• Safety guarantees — majority voting, log matching, commit rules
• Async runtime — tokio-based event loop with timer management
• In-memory network — message routing for testing cluster behavior

Architecture

Uses an action/effect pattern: event handlers return Vec<Action> describing state changes, which are then executed. This separation makes the protocol logic testable independent of side effects. Core types map directly to the paper's Figure 2 specification.

# Creates a 3-node cluster, one node wins election within ~200-300ms
cargo run

Tech Stack

• Rust — core implementation
• Tokio — async runtime