Important: This documentation is about an older version. It's relevant only to the release noted, many of the features and functions have been updated or replaced. Please view the current version.

Documentation Grafana k6 Using k6 Workaround to calculate iteration_duration

Workaround to calculate iteration_duration

A common requested case is to track the iteration_duration metric without including time spent for setup and teardown functions. This feature is not yet available but a threshold on iteration_duration or any pre-existing metrics can be used as a workaround.

It’s based on the concept of creating thresholds for sub-metrics created by tags for the required scope and setting the criteria that always pass. It works with any enabled tags that already works with threshold, for example:

• iteration_duration{scenario:default} generates a sub-metric collecting samples only for the default scenario’s iteration. scenario:default is used because that’s the internal k6 scenario name when we haven’t specified options. scenarios` explicitly and are just using the execution shortcuts instead.
• iteration_duration{group:::setup} or iteration_duration{group:::teardown} create sub-metrics collecting the duration only for setup and teardown. k6 implicitly creates groups for setup and teardown, and :: is the group separator.
• http_req_duration{scenario:default} can be useful as well for isolating executed long-running requests.

JavaScript

Copy

JavaScript Copy

import { sleep } from 'k6';
import http from 'k6/http';

export const options = {
  vus: 1,
  iterations: 1,
  thresholds: {
    'iteration_duration{scenario:default}': [`max>=0`],
    'iteration_duration{group:::setup}': [`max>=0`],
    'iteration_duration{group:::teardown}': [`max>=0`],
    'http_req_duration{scenario:default}': [`max>=0`],
  },
};

export function setup() {
  http.get('https://httpbin.test.k6.io/delay/5');
}

export default function () {
  http.get('http://test.k6.io/?where=default');
  sleep(0.5);
}

export function teardown() {
  http.get('https://httpbin.test.k6.io/delay/3');
  sleep(5);
}

Dedicated sub-metrics have been generated collecting samples only for the scope defined by thresholds:

bash

Copy

bash Copy

http_req_duration..............: avg=1.48s    min=101.95ms med=148.4ms  max=5.22s    p(90)=4.21s    p(95)=4.71s
  { expected_response:true }...: avg=1.48s    min=101.95ms med=148.4ms  max=5.22s    p(90)=4.21s    p(95)=4.71s
✓ { scenario:default }.........: avg=148.4ms  min=103.1ms  med=148.4ms  max=193.7ms  p(90)=184.64ms p(95)=189.17ms

iteration_duration.............: avg=5.51s    min=1.61s    med=6.13s    max=8.81s    p(90)=8.27s    p(95)=8.54s
✓ { group:::setup }............: avg=6.13s    min=6.13s    med=6.13s    max=6.13s    p(90)=6.13s    p(95)=6.13s
✓ { group:::teardown }.........: avg=8.81s    min=8.81s    med=8.81s    max=8.81s    p(90)=8.81s    p(95)=8.81s
✓ { scenario:default }.........: avg=1.61s    min=1.61s    med=1.61s    max=1.61s    p(90)=1.61s    p(95)=1.61s