A developer recently shared how they built a 'Micro-Social' API on a $5 VPS, aiming for sub-300ms latency. After choosing Bun and ElysiaJS over Node.js for speed, their initial deployment crashed under a simulated load test using k6. The server rejected 86% of requests due to rate limiting.

The culprit was a misconfigured global rate limiter set to 100 requests per minute, essentially DDOSing the server from a single test IP. To fix this, the developer tuned the rate limiter to 2,500 req/min and increased the PostgreSQL connection pool to 80. They also used Docker to run two container replicas, utilizing both vCPUs on the VPS.

The final result was a 100% success rate with 200ms latency. This proves that proper tuning beats Kubernetes for small projects.