k6 Load Test Config Builder

k6 is an open-source load testing tool built for developer workflows. It uses JavaScript/TypeScript for test scripts, integrates with CI/CD pipelines, and outputs structured metrics (p95, p99 latency, error rate, throughput). Unlike JMeter, k6 has a low resource footprint — a single k6 instance can simulate thousands of VUs. It's cloud-friendly and integrates natively with Grafana dashboards.

Smoke test: Minimal load (1-2 VUs) to verify the test script works. Load test: Expected normal load to validate baseline performance. Stress test: Gradually increase beyond normal capacity to find the breaking point. Spike test: Sudden massive increase to simulate flash crowd events. Soak test: Sustained normal load for extended periods (hours) to detect memory leaks and resource exhaustion.

A Virtual User (VU) in k6 represents a simulated user executing your test script in a continuous loop. Each VU runs independently and maintains its own state (cookies, session). The number of concurrent VUs is the primary load parameter. Note: VUs are not the same as requests per second — the RPS depends on iteration duration which includes response time and think time.

Common latency SLO targets: API endpoints: p95 < 200ms, p99 < 500ms. Database queries: p95 < 50ms, p99 < 100ms. Page loads: p95 < 2s, p99 < 4s. For k6 thresholds, typically: http_req_duration p95 < 500, error_rate < 0.01 (1%). These are starting points — adjust based on your specific SLAs and user expectations.

Key metrics to analyze: (1) http_req_duration — latency percentiles (p50, p95, p99). (2) http_req_failed — error rate, should stay below 1% under normal load. (3) iterations — total completed test iterations. (4) vus — concurrent virtual users at peak. Look for the knee point where latency spikes and errors appear — that is your system's capacity limit. Compare p99 against your SLA targets.