load-testing

Define an options object with stages for ramp-up, steady state, and ramp-down, then write a default function that makes HTTP requests and validates responses with checks. Add thresholds like p(95)<500 for latency and rate<0.01 for failures to enforce performance criteria.

Load tests verify behavior under expected traffic, stress tests find breaking points by gradually increasing load, spike tests check resilience to sudden traffic surges, and soak tests run normal load for hours to detect memory leaks and degradation.

Install k6 via the official apt repository in a workflow step, run your test script with k6 run, and upload the JSON results as an artifact. Thresholds with abortOnFail cause non-zero exit codes, failing the pipeline automatically on regressions.

Common baselines are p95 response time under 500ms, p99 under 1 second, and error rate below 1%. You can also set per-endpoint thresholds using tags, such as stricter limits for search endpoints versus login.

k6 load testing does not apply to unit tests, functional correctness tests, or UI rendering performance. Use Lighthouse for frontend rendering metrics and dedicated security scanners for network vulnerability assessment instead.