LoadForge
BUILDINGSelf-hosted API load testing with real-time performance metrics and AI-assisted bottleneck diagnosis.
Overview
LoadForge brings enterprise-grade load testing to a self-hosted environment. It orchestrates k6 load generators via distributed BullMQ workers, capturing live telemetry and storing it in PostgreSQL. After a run, an integrated AI layer analyzes the time-series data to diagnose bottlenecks, identifying whether the issue lies in connection pooling, DB locks, or application logic.
Architecture
Why I built this
Every load testing tool I tried was either too heavy — enterprise software requiring days of setup — or too light — a shell script that hits a URL and returns average latency. I wanted something that you could self-host in minutes and that gives you real, production-quality signal about what breaks under load and why.
Engineering Challenges
- Subprocess orchestration and preventing zombie k6 processes on worker crash
- Managing massive influxes of time-series data during high-throughput tests
- Queue recovery and state management during Redis node restarts
What I learned
Spawning native subprocesses requires robust defensive programming around signal handling and stream piping.
AI evaluation of time-series data requires summarizing metrics into statistical percentiles before prompting, rather than sending raw points.
Stack
TypeScript
Next.jsFastify
RedisBullMQ
PostgreSQLDrizzlek6
Docker