"Pilots have every landing automated. Every single one. But young pilots are still made to land planes manually. Because when automation fails, someone needs to actually know how to fly."
____ ____ _
| __ ) _ _ __ _| __ ) _ _ ___| |_ ___ _ __
| _ \| | | |/ _` | _ \| | | / __| __/ _ \ '__|
| |_) | |_| | (_| | |_) | |_| \__ \ || __/ |
|____/ \__,_|\__, |____/ \__,_|___/\__\___|_|
|___/
We learned debugging by breaking things. By staring at a bug for 3 hours and finally getting it. By writing terrible code, shipping it, and cleaning up the mess. That cycle of fail, understand, improve — that's not inefficiency. That's the learning mechanism.
AI shortcuts all of it.
A fresher once asked: "How will I learn if everything gets done in minutes?" — and he's right. We're so busy making our teams faster that we might be forgetting to make them stronger.
BugBuster is a no-AI flight simulator for engineers. It spins up a production-like microservices stack (Node.js, Python, Java, Go, nginx) with real observability (Grafana, Jaeger, Prometheus, Loki), injects a real bug, and makes you find and fix it yourself. No copilot. No autocomplete. Just you, the logs, the traces, and the metrics.
Because if you don't understand what's happening under the hood, you can't steer the AI that does it for you.
Grab the latest release for your platform:
# macOS (Apple Silicon)
curl -L https://github.com/anshulsao/bugbuster/releases/latest/download/bugbuster-darwin-arm64 -o bugbuster
chmod +x bugbuster
sudo mv bugbuster /usr/local/bin/
# macOS (Intel)
curl -L https://github.com/anshulsao/bugbuster/releases/latest/download/bugbuster-darwin-amd64 -o bugbuster
chmod +x bugbuster
sudo mv bugbuster /usr/local/bin/
# Linux (x86_64)
curl -L https://github.com/anshulsao/bugbuster/releases/latest/download/bugbuster-linux-amd64 -o bugbuster
chmod +x bugbuster
sudo mv bugbuster /usr/local/bin/
# Linux (ARM64)
curl -L https://github.com/anshulsao/bugbuster/releases/latest/download/bugbuster-linux-arm64 -o bugbuster
chmod +x bugbuster
sudo mv bugbuster /usr/local/bin/Requires Go 1.22+.
git clone https://github.com/anshulsao/bugbuster.git
cd bugbuster
go build -o bugbuster ./cmd/bugbuster
sudo mv bugbuster /usr/local/bin/BugBuster runs services in Docker. You need:
- Docker (v20+) with Docker Compose v2
- ~4 GB free RAM (for all containers)
Verify Docker is running:
docker compose version
# Docker Compose version v2.x.x# Launch the interactive TUI
bugbusterThis opens a terminal UI where you can browse scenarios, start environments, test APIs, get hints, and submit your root cause analysis — all from one screen.
# List available scenarios
bugbuster list
# Start a scenario
bugbuster start ghost-latency
# Get a hint (costs points)
bugbuster hint
# Check session status
bugbuster status
# Submit your RCA
bugbuster submit
# Tear down when done
bugbuster stop
# View your scores
bugbuster leaderboard You run bugbuster start <scenario>
│
▼
┌─────────────────────────────┐
│ Docker Compose brings up: │
│ │
│ nginx ─► order (Node.js) │
│ ─► catalog (Python) │
│ ─► payment (Java) │
│ ─► notifier (Go) │
│ │
│ + Redis, RabbitMQ, DBs │
│ + Grafana, Jaeger, Prom │
│ │
│ A bug is injected via │
│ compose override + env vars│
└─────────────────────────────┘
│
▼
You investigate using real
observability tools:
Grafana → localhost:3000
Jaeger → localhost:16686
Prometheus→ localhost:9091
│
▼
bugbuster submit
→ checks your RCA category
→ runs validation scripts
→ scores your performance
Once a scenario is running, these tools are available:
| Tool | URL | Credentials |
|---|---|---|
| Grafana | http://localhost:3000 | admin / bugbuster |
| Jaeger | http://localhost:16686 | — |
| Prometheus | http://localhost:9091 | — |
| Loki (via Grafana) | http://localhost:3000 | — |
| RabbitMQ | http://localhost:15672 | bugbuster / bugbuster |
| MailHog | http://localhost:8025 | — |
| API Gateway | http://localhost:8888 | — |
Dashboard:
1-7 Quick-select action
j/k Navigate action list
enter Select action
ctrl+c Quit (tears down containers)
Submit screen:
j/k Scroll results
r Back to dashboard
esc Back
Hints screen:
enter Reveal next hint
esc Back
| Scenario | Level | Time | What You'll Learn |
|---|---|---|---|
| Ghost Latency | Medium | 25 min | Resource saturation, USE method, thread pool exhaustion |
More scenarios coming soon.
If you have Claude Code (claude) or Gemini CLI (gemini) installed, BugBuster will automatically use it to evaluate your RCA submission and give personalized feedback after you solve a scenario.
MIT