Raft Consensus Algorithm (Go)

An implementation of the Raft consensus algorithm in Go, including leader election, log replication, persistence, and snapshot-based log compaction.

This project was implemented following the MIT 6.824 Distributed Systems Raft labs.

Reference

• MIT 6.824: https://pdos.csail.mit.edu/6.824/
• Raft Paper: https://raft.github.io/raft.pdf
• Visualization: https://raft.github.io/

Raft Protocol Overview

Highlights

• Full Raft implementation: leader election, log replication, and safety checks.
• Durable state via persistence and snapshotting for log compaction.
• Replicated key/value store on top of Raft.
• Sharded key/value store with a shard controller service.
• MapReduce lab code and harness for integration testing.

Project Layout

• src/raft: Raft core (election, append entries, persistence, snapshots).
• src/kvraft: Replicated key/value server built on Raft.
• src/shardctrler: Shard controller for reconfiguration.
• src/shardkv: Sharded key/value store built on Raft.
• src/labrpc: Simulated RPC/network layer used by the labs.
• src/labgob: Lab-specific encoder/decoder utilities.
• src/porcupine: Linearizability checker used by tests.
• src/mr, src/mrapps, src/main: MapReduce implementation and test drivers.

Getting Started

This repo is organized as the MIT 6.5840 module (src/go.mod). From the repo root:

cd src
go test ./raft

Running Tests

Run tests by package (from src/):

go test ./raft
go test ./kvraft
go test ./shardctrler
go test ./shardkv

Run everything:

go test ./...

MapReduce Playground

The MapReduce harness lives in src/main. A quick smoke test:

cd src/main
bash test-mr.sh

Notes

• This code targets the MIT 6.824 lab environment and is not production-ready.
• The RPC layer and persistence are provided by lab utilities to focus on Raft logic.