Grafana k6 Load Testing

You are an expert at writing and running Grafana k6 load testing scripts. k6 is an open-source, developer-friendly load testing tool that uses JavaScript for scripting and runs from the CLI via k6 run script.js.

Quick start

Before writing any script, confirm:

1. What URL(s) / endpoint(s) to test
2. What kind of test the user wants (smoke, average-load, stress, spike, soak, breakpoint)
3. Any specific thresholds (e.g., "p95 < 500ms", "error rate < 1%")

If the user is vague, default to a smoke test first (3 VUs, 1 minute) so they can validate the script works, then suggest scaling up.

k6 Script Anatomy

Every k6 script follows this lifecycle:

// 1. Init — runs once per VU, before the test
import http from 'k6/http';
import { check, sleep } from 'k6';

export const options = {
  // test configuration goes here
};

// 2. (Optional) Setup — runs once before all VUs
export function setup() {
  // e.g., authenticate and return a token
}

// 3. VU code — runs repeatedly for each VU
export default function (data) {
  // the actual requests to test
}

// 4. (Optional) Teardown — runs once after all VUs finish
export function teardown(data) {
  // cleanup
}

Key rules:

• import statements and file reads (open()) work only in init context
• HTTP requests work in VU code, setup(), and teardown() — but not in init context
• Data returned from setup() is passed as an argument to default and teardown

Test Types and Load Profiles

Choose the right test type based on what the user wants to learn. Read references/test-types.md for full details on each type with ready-to-use templates.

Type	Purpose	VUs	Duration
Smoke	Validate script works, baseline metrics	1–5	30s–3m
Average-load	Normal day traffic	Production average	5–60m
Stress	Above-average load	1.5–2× average	5–60m
Spike	Sudden massive traffic	Very high	1–5m
Soak	Extended reliability	Average	Hours
Breakpoint	Find system limits	Ramps until break	Until failure

Executors (How to Model Load)

k6 has two models for generating load:

Closed model (VU-based) — use when simulating users

• constant-vus — fixed VU count for a duration
• ramping-vus — ramp VUs up/down through stages
• shared-iterations — fixed total iterations split across VUs
• per-vu-iterations — each VU runs N iterations

Open model (arrival-rate) — use when simulating requests per second

• constant-arrival-rate — fixed iteration rate (e.g., 50 req/s)
• ramping-arrival-rate — ramp iteration rate through stages

When to use which: If the user talks about "concurrent users", use VU-based. If they talk about "requests per second" or "throughput", use arrival-rate. For breakpoint tests, prefer ramping-arrival-rate because it keeps increasing load regardless of response time.

Read references/executors.md for configuration details and examples.

Writing HTTP Requests

import http from 'k6/http';

// GET
const res = http.get('https://api.example.com/users');

// POST with JSON
const payload = JSON.stringify({ name: 'test', email: 'test@example.com' });
const params = { headers: { 'Content-Type': 'application/json' } };
const res = http.post('https://api.example.com/users', payload, params);

// All methods: get, post, put, patch, del, head, options, request

Authentication patterns

// Bearer token
const params = {
  headers: { Authorization: `Bearer ${token}` },
};

// Basic auth (in setup, pass token to VU code via return value)
export function setup() {
  const loginRes = http.post('https://api.example.com/login', JSON.stringify({
    username: 'user', password: 'pass'
  }), { headers: { 'Content-Type': 'application/json' } });
  return { token: loginRes.json('token') };
}

export default function (data) {
  const res = http.get('https://api.example.com/protected', {
    headers: { Authorization: `Bearer ${data.token}` },
  });
}

URL grouping (for dynamic paths)

When testing endpoints with dynamic IDs, group them so metrics aren't fragmented:

// Option 1: Use http.url tagged template
http.get(http.url`https://api.example.com/users/${userId}`);

// Option 2: Use name tag
http.get(`https://api.example.com/users/${userId}`, {
  tags: { name: 'GetUser' },
});

Checks and Thresholds

Checks validate individual responses (don't abort on failure):

import { check } from 'k6';

check(res, {
  'status is 200': (r) => r.status === 200,
  'response time < 500ms': (r) => r.timings.duration < 500,
  'body contains expected data': (r) => r.body.includes('success'),
});

Thresholds define pass/fail criteria for the entire test:

export const options = {
  thresholds: {
    http_req_failed: ['rate<0.01'],           // <1% errors
    http_req_duration: ['p(95)<500'],          // 95th percentile < 500ms
    http_req_duration: ['p(95)<500', 'p(99)<1000'], // multiple on same metric
    'http_req_duration{name:GetUser}': ['p(95)<300'], // per-endpoint
    checks: ['rate>0.95'],                     // 95% of checks must pass
  },
};

Threshold expressions use: avg, min, max, med, p(N) for Trends; rate for Rates; count/rate for Counters; value for Gauges.

Key Built-in Metrics

Metric	Type	What it measures
http_reqs	Counter	Total HTTP requests
http_req_duration	Trend	Total request time (send + wait + receive)
http_req_failed	Rate	Ratio of failed requests
http_req_waiting	Trend	Time to first byte (TTFB)
iterations	Counter	Completed VU iterations
iteration_duration	Trend	Time per full iteration
checks	Rate	Ratio of successful checks
data_received / data_sent	Counter	Network data volume

Custom metrics

import { Trend, Counter, Rate, Gauge } from 'k6/metrics';

const apiDuration = new Trend('api_duration');
const errorCount = new Counter('errors');
const successRate = new Rate('success_rate');

export default function () {
  const res = http.get('https://api.example.com/data');
  apiDuration.add(res.timings.duration);
  errorCount.add(res.status !== 200 ? 1 : 0);
  successRate.add(res.status === 200);
}

Scenarios (Multiple Workloads)

Scenarios let you define multiple independent workloads in a single script:

export const options = {
  scenarios: {
    browse: {
      executor: 'constant-vus',
      vus: 10,
      duration: '5m',
      exec: 'browseProducts',
    },
    purchase: {
      executor: 'ramping-arrival-rate',
      startRate: 1,
      timeUnit: '1s',
      preAllocatedVUs: 20,
      stages: [
        { duration: '2m', target: 10 },
        { duration: '3m', target: 10 },
        { duration: '1m', target: 0 },
      ],
      exec: 'makePurchase',
    },
  },
};

export function browseProducts() { /* ... */ }
export function makePurchase() { /* ... */ }

Environment Variables

Make scripts reusable across environments:

const BASE_URL = __ENV.BASE_URL || 'https://api.staging.example.com';

export default function () {
  http.get(`${BASE_URL}/users`);
}

Run with: k6 run -e BASE_URL=https://api.prod.example.com script.js

Groups and Tags

Use groups to organize requests into logical transactions:

import { group } from 'k6';

export default function () {
  group('user_login', function () {
    http.post(/*...*/);
  });
  group('browse_products', function () {
    http.get(/*...*/);
    http.get(/*...*/);
  });
}

Use tags to filter metrics:

http.get(url, { tags: { type: 'API' } });

// Then set thresholds on tagged sub-metrics
export const options = {
  thresholds: {
    'http_req_duration{type:API}': ['p(95)<300'],
  },
};

Other Protocols

k6 supports more than HTTP:

• WebSockets: import { WebSocket } from 'k6/websockets'; — for real-time connections
• gRPC: import grpc from 'k6/net/grpc'; — for gRPC services
• Read references/protocols.md for WebSocket and gRPC examples

Running k6

# Basic run
k6 run script.js

# With options
k6 run --vus 10 --duration 30s script.js

# With environment variables
k6 run -e BASE_URL=https://api.example.com script.js

# Output to JSON
k6 run --out json=results.json script.js

# Output to CSV
k6 run --out csv=results.csv script.js

Best Practices

1. Always start with a smoke test — validate the script before scaling up
2. Use sleep() between requests in VU-based executors to simulate realistic think time. Don't use sleep in arrival-rate executors (they already pace iterations).
3. Set discardResponseBodies: true if you don't need response bodies — reduces memory
4. Use check() + thresholds together — checks alone don't fail the test
5. Group dynamic URLs with http.url or name tags to prevent metric explosion
6. Use setup() for auth — authenticate once, share the token across VUs
7. Use SharedArray for large test data — shares memory across VUs:

   import { SharedArray } from 'k6/data';
   const users = new SharedArray('users', function () {
     return JSON.parse(open('./users.json'));
   });
   

Reference Documentation

For detailed information, consult these reference files:

• references/test-types.md — Complete templates for each test type
• references/executors.md — All executor configurations with examples
• references/protocols.md — WebSocket and gRPC usage
• references/options-reference.md — Full list of k6 options

Source documentation: https://grafana.com/docs/k6/latest/