🧭 Raft Consensus in Rust (with Fault Injection Simulator)

A from-scratch implementation of the Raft consensus algorithm in Rust, built with a deterministic multi-node simulation environment supporting network delays, message drops, and partitions.

This project focuses on correctness under failures, demonstrating how distributed systems maintain consistency despite unreliable networks.

---

✨ Highlights

• 🧠 Leader election (RequestVote RPC)
• 🔁 Log replication (AppendEntries RPC)
• 🧾 Structured log entries with command abstraction
• 🌐 Simulated distributed network using threads + channels
• ⚡ Fault injection:
  • Message drops
  • Random network delays
  • Network partitions
• 🧪 Designed for testing distributed system behavior under chaos

---

🏗️ Architecture

Each node runs as an independent thread with:

• Local Raft state (term, role, log)
• Message inbox (channel-based)
• Timers for election + heartbeat

A central Network layer simulates real-world failures:

Nodes (threads) ↓ Message passing (mpsc channels) ↓ Network layer ├── random delay ├── message drop └── partition isolation

---

📡 Raft RPCs Implemented

RequestVote

• Election initiation
• Term comparison
• Vote granting logic

AppendEntries

• Heartbeats
• Log replication
• Leader commit propagation

---

🧪 Fault Injection (Key Feature)

The simulation includes a configurable chaos layer:

• drop_rate: probability of message loss
• min_delay_ms / max_delay_ms: network latency
• partition map: isolates nodes into groups

This allows testing scenarios like:

• Leader isolation
• Split-brain elections
• Delayed replication
• Recovery after partition healing

---

▶️ Running the Simulation

cargo run

You will observe:
 - Leader elections
 - Term changes
 - Log replication
 - Behavior under network instability

🧠 Design Goals
 - Understand Raft beyond theory
 - Explore behavior under realistic failures
 - Build intuition for:
   - safety vs liveness
   - timing sensitivity
   - distributed coordination

📈 Roadmap
v0.1 (current)
- Leader election
- Basic log replication
- Fault-injection network

v0.2
- Stable commit index handling
- More robust election timers

v0.3
- Log consistency guarantees
- Safety invariants testing

v0.4
- Persistence (WAL / disk-backed logs)
- Crash + recovery

v0.5
- Metrics + benchmarking (latency, throughput)
- Stress testing with randomized failures

🔗 Relation to My Work

This project complements my research in:
- Byzantine Fault Tolerance (pMVBA)
- Distributed systems correctness
- Fault-tolerant protocol design

While Raft targets crash fault tolerance (CFT), my research extends into Byzantine environments, bridging theory and real-world systems.

💡 Future Extensions
- Snapshotting / log compaction
- gRPC-based real network (replace simulation)
- Integration with a KV store (state machine)
- Comparison with BFT protocols

🛠️ Tech Stack
- Rust (concurrency, safety)
- std::thread + mpsc channels
- Custom simulation framework

📌 Why This Project

Distributed systems don't fail in ideal conditions —
they fail under partitions, delays, and partial failures.

This project is built to explore exactly that.