performance-testing

Load testing and stress testing find API breaking points by running controlled virtual user ramps with k6 and Locust. You configure staged ramp-ups to measure capacity, identify bottlenecks, and validate performance thresholds like p95 < 500ms.

Spike testing validates system behavior during sudden traffic bursts, while soak testing runs long-duration steady-state loads to uncover memory leaks. Both are controlled performance tests that measure response times, error rates, and throughput under sustained or extreme stress.

Configure k6 thresholds for p95 latency and error rate validation by setting performance goals like p95 < 500ms and error rate < 1% in your script. The test fails automatically if these custom metrics exceed the defined performance thresholds.

You can run automated performance testing with k6 in GitHub Actions CI pipelines by exporting JSON or InfluxDB results. Configurable thresholds gate CI validation, allowing automated checks to block deployments if response times or error rates fail.

Locust supports custom metrics collection for bottleneck analysis by tracking trends, counters, and rates during your load test. This metrics collection validates throughput and response times to help detect system bottlenecks under concurrent user load.

You should not use stress testing against production environments without isolation because finding breaking points generates extreme concurrent user load. Run controlled load, spike, and soak tests against staging environments to safely validate capacity and avoid incidents.