k6-load-testing

Import http from k6/http, define options with VUs and duration or stages, then make requests in the default function. Add check() calls for response validation and thresholds in options for pass/fail criteria, then run with k6 run script.js.

Ramping-vus controls concurrent virtual users through stages, modeling user count. Ramping-arrival-rate controls iterations per second regardless of response time, modeling throughput. Use arrival-rate for breakpoint tests and requests-per-second targets.

Yes. k6 includes k6/websockets for standard WebSocket connections and k6/net/grpc for gRPC services including streaming. Other protocols like Kafka, SQL, and Redis are available through xk6 extensions.

Define thresholds in the options object using metric expressions like http_req_failed: ['rate<0.01'] or http_req_duration: ['p(95)<500']. k6 exits with a non-zero code when thresholds fail, which works directly in CI pipelines.

Requests with dynamic path segments like user IDs create separate metric series per URL. Group them using http.url tagged templates or a name tag so all variants aggregate under one metric.

Avoid sleep() with arrival-rate executors because the executor already paces iterations to hit the target rate. Use sleep() only in VU-based executors to simulate realistic user think time between requests.