Raft Consensus Implementation

This repository contains a complete implementation of the Raft consensus algorithm in Go, based on the extended Raft paper.

Features Implemented

• Part 2A: Leader Election

• Leader election and heartbeats via AppendEntries RPCs

• Election timeouts and term management

• Automatic re-election on leader failure

• Part 2B: Log Replication

• Log entry replication via AppendEntries RPCs

• Log consistency checks

• Commitment of entries across the cluster

• Part 2C: Persistence

• Persistent state management

• Crash recovery

• State machine snapshots

• Part 2D: Log Compaction

• Log trimming with snapshots

• InstallSnapshot RPC implementation

• Snapshot state transfer between peers

Project Structure

.
├── config.go     # Test configuration and utilities
├── persister.go  # Persistence layer
├── raft.go       # Core Raft implementation
├── util.go       # Helper functions
└── pkg/
	├── labgob/   # Encoding utilities
	└── labrpc/   # RPC framework

Usage

The Raft implementation supports the following interface:

// Create a new Raft server instance
rf := Make(peers, me, persister, applyCh)

// Start agreement on a new log entry
rf.Start(command interface{}) (index, term, isleader)

// Get current term and leader status
rf.GetState() (term, isLeader)

Testing

Run the full test suite with:

go test -race

Or test specific parts:

go test -run 2A -race # Leader election
go test -run 2B -race # Log replication
go test -run 2C -race # Persistence
go test -run 2D -race # Log compaction

The implementation passes all test cases while maintaining reasonable performance metrics (typical full test suite runtime is under 8 minutes).