Bully Algorithm — Distributed Coordinator Election

Simulasi Bully Algorithm menggunakan 5 Docker node Go yang berkomunikasi satu sama lain via HTTP, lengkap dengan web dashboard real-time.

Apa itu Bully Algorithm?

Bully Algorithm (Garcia-Molina, 1982) adalah protokol election klasik di sistem terdistribusi. Aturannya sederhana:

Node dengan ID tertinggi selalu menang ("membully") node lain untuk jadi koordinator.

Proses Election:

1. Node X mendeteksi coordinator mati
2. Node X broadcast pesan ELECTION ke semua node ber-ID > X
3. Jika ada yang balas OK  → X kalah, menunggu COORDINATOR
4. Jika tidak ada yang balas → X menang, broadcast COORDINATOR ke semua
5. Semua node update coordinator mereka

Pesan yang digunakan:

Pesan	Dari → Ke	Arti
ELECTION	Node X → semua node ID > X	"Saya mau jadi leader, ada yang lebih tinggi?"
OK	Node Y → X (jika Y > X)	"Saya lebih tinggi, kamu kalah"
COORDINATOR	Pemenang → semua	"Saya koordinator baru"
HEARTBEAT	Coordinator → semua	"Saya masih hidup"

---

Arsitektur

┌──────────────────────────────────────────────────────┐
│                  Docker Network (bully-net)           │
│                                                      │
│  ┌─────────┐  ┌─────────┐  ┌─────────┐              │
│  │ NODE-1  │  │ NODE-2  │  │ NODE-3  │              │
│  │ :8080   │  │ :8080   │  │ :8080   │              │
│  └────┬────┘  └────┬────┘  └────┬────┘              │
│       │             │             │                   │
│       └─────────────┼─────────────┘                   │
│                     │ HTTP                            │
│  ┌─────────┐  ┌─────┴───┐                            │
│  │ NODE-4  │  │ NODE-5  │ ← highest ID = leader      │
│  │ :8080   │  │ :8080   │                            │
│  └─────────┘  └─────────┘                            │
│                                                      │
│  ┌──────────────────────┐                            │
│  │ DASHBOARD  :3000     │ ← monitor & control        │
│  └──────────────────────┘                            │
└──────────────────────────────────────────────────────┘

---

Cara Jalankan

Prasyarat

• Docker & Docker Compose terinstall

1. Build & Start

docker compose up -d --build

2. Buka Dashboard

open http://localhost:3000

3. Jalankan Skenario

Crash coordinator:

./scripts/bully.sh crash-coordinator
# Tunggu ~6 detik → election otomatis terjadi

Crash node tertentu:

./scripts/bully.sh crash 5   # crash NODE-5
./scripts/bully.sh recover 5 # recover NODE-5

Trigger election manual:

./scripts/bully.sh elect 2   # mulai election dari NODE-2

Skenario cascade failure:

./scripts/bully.sh scenario-cascade
# Crash node 5 → election → crash node 4 → election → recover 5 → 5 bully balik jadi leader

Stress test:

./scripts/bully.sh scenario-split

---

Struktur File

bully-algorithm/
├── docker-compose.yml          # 5 node + 1 dashboard
├── node/
│   ├── main.go                 # Bully Algorithm implementation
│   ├── go.mod
│   └── Dockerfile
├── dashboard/
│   ├── main.go                 # Dashboard proxy server
│   ├── go.mod
│   ├── Dockerfile
│   └── static/
│       └── index.html          # Real-time web dashboard
└── scripts/
    └── bully.sh                # Helper commands

---

API Node (port 8080 tiap node)

Endpoint	Method	Deskripsi
GET /status	GET	Status lengkap node
POST /message	POST	Terima pesan election/heartbeat
POST /crash	POST	Simulasi crash
POST /recover	POST	Simulasi recovery
POST /election	POST	Trigger election manual

---

Implementasi Detail

Timeout & Timing

• Heartbeat interval: 2 detik (coordinator → semua node)
• Leader dead timeout: 6 detik tanpa heartbeat → trigger election
• Election response timeout: 2 detik per node
• Wait for COORDINATOR: 5 detik setelah dapat OK

Startup Election

Saat node pertama kali start, mereka menunggu dengan jitter berdasarkan ID:

• NODE-5 menunggu paling sedikit → menang pertama kali
• NODE-1 menunggu paling lama → menerima hasil election

Recovery

Saat node crash recovery, ia langsung memulai election. Jika ID-nya adalah yang tertinggi, ia akan mem-bully node lain dan merebut kembali posisi coordinator.

---

Contoh Flow (dari logs)

[15:04:01.000] NODE-3 | 💔 Coordinator NODE-5 appears DEAD — starting election
[15:04:01.010] NODE-3 | 🗳️  Starting ELECTION (term 2)
[15:04:01.020] NODE-3 | 📨 Sent ELECTION to NODE-4, NODE-5
[15:04:01.100] NODE-4 | 📨 Received ELECTION from NODE-3 (term 2)
[15:04:01.110] NODE-4 | ✅ Sent OK to NODE-3 (I have higher ID)
[15:04:01.120] NODE-4 | 🗳️  Starting ELECTION (term 2)
[15:04:01.200] NODE-3 | 📬 Got OK from NODE-4 — stepping down
[15:04:03.200] NODE-4 | ⚠️  No response from NODE-5 (may be down)
[15:04:03.210] NODE-4 | 👑 No higher nodes alive — I win ELECTION (term 2)
[15:04:03.220] NODE-4 | 📢 Broadcasting COORDINATOR (term 2)
[15:04:03.300] NODE-1 | 👑 NODE-4 is the new COORDINATOR (term 2)
[15:04:03.310] NODE-2 | 👑 NODE-4 is the new COORDINATOR (term 2)
[15:04:03.320] NODE-3 | 👑 NODE-4 is the new COORDINATOR (term 2)