Stop 10 · Performance TestingTopics · 4/8

k6 / Gatling (Overview)

In short

Some teams want performance tests that look like normal code: written in an editor, reviewed in pull requests, run in CI. That's where k6 and Gatling shine.

Modern, code-based alternatives to JMeter — k6 uses JavaScript, Gatling uses Scala — often preferred by teams wanting performance tests written and version-controlled like regular code, integrated into CI/CD.

The purple blob jogs on a treadmill reading a plan from a notebook clipped to the handrail.

Think of it like this

A workout plan written in a notebook

Instead of fiddling with the treadmill's buttons each time, you write the plan in a notebook: 5 minutes warm-up, 20 minutes run, 5 minutes cool down. Anyone can read it, change it and repeat it. k6 and Gatling scripts are that notebook.

Watch it happen

GUI-first vs code-first tools

JMeter

Build plans in a GUI
Saved as large XML files
Huge plugin ecosystem

k6 / Gatling

Plain scripts in JavaScript or Scala
Easy to diff and review in Git
Thresholds fail CI builds directly

All three can load test the same app. Pick code-first if your team lives in Git and CI. Pick JMeter if testers prefer a GUI.

The big ideas

What you need to know

k6 uses JavaScript

Scripts are JavaScript, run by a fast Go-based engine. Maintained by Grafana Labs.

Gatling uses the JVM

Gatling scripts started in Scala, and newer versions also support Java and Kotlin.

Thresholds as pass or fail

Write 'p95 under 500 ms' in the script, and the run fails if it's broken.

Permission first

These tools can generate serious traffic. Point them only at systems you're allowed to test.

See it in code

Walk through the code

1import http from "k6/http";2import { check, sleep } from "k6";3export const options = {4  vus: 20,5  duration: "1m",6  thresholds: {7    http_req_duration: ["p(95)<500"],8    http_req_failed: ["rate<0.01"],9  },10};11 12export default function () {13  const res = http.get("https://test.k6.io/");14  check(res, { "status is 200": (r) => r.status === 200 });15  sleep(1);16}

1/320 virtual users loop for one minute.

Save as smoke.js and run k6 run smoke.js. test.k6.io is a public site Grafana provides for practice.

Save yourself the pain

Common mistakes

Oops

Removing sleep() so each virtual user hammers the server non-stop.

Do this instead

Add think time. Real users pause, and without it 20 users behave like hundreds.

Oops

Treating check() failures as failing the run.

Do this instead

Checks are recorded but don't fail the run alone. Use thresholds to set pass or fail.

Your mission

Run k6 in five minutes

~25 min

Quick check

What makes a k6 run fail automatically in CI?

Go deeper

1Explore k6's documentation and write a simple test script.
2Compare the code-based approach of k6/Gatling to JMeter's GUI-based test plan approach.
3Learn why code-based tools tend to fit more naturally into a CI/CD pipeline (Phase 11).