@dnvgl/k6-script-from-har

k6-script-from-har

Dead simple k6 test generator.
Given a har file, this script parses and generates cleanly factored k6 test code.
[Only fetch calls]

This is an alternative to the har-to-k6 converter which in our experience created messy generated "write only" code that required a lot of clean up every time we used it.

Who this is for

We use this for workflow load testing of a backend heavy SPA with 10s or 100s of users.

We wanted to measure

1. Realistic load characteristics (real-world usage conditions)
2. Workflow completion time
   • baseline time for one virtual user
   • how far can we push virtual user count until the time is unacceptable (e.g. 20s for a multi-page workflow)
3. Infrastructure reliability
   • at how many virtual users does something in the infrastructure break (follow up - what broke, how do we harden our infra)

The record and playback approach was a good fit for us. We also wanted basic checks in place e.g. failure response status should show up red instead of green in results.

Installation

Ensure k6 is installed

npm install --save-dev @dnvgl/k6-script-from-har

note: files testTemplate.js and testCommon.js will be copied into repository root post install these should be committed with the repo

Usage

Given a .har file, run k6-script-from-har with the input har filename and output test filename.

note: we recommend filtering out *.har files in .gitignore har files can include secrets, and are very large

Example:
generate the test file from har

npx k6-script-from-har createTimeEntryWorkflow.har tests/createTimeEntryWorkflow.js

run the test

k6 run tests/createTimeEntryWorkflow.js

Har file creation

Manually

• Use a browser, open up dev tools, network tab
• Navigate and perform actions for workflow test
• Export har documentation: chrome export har

Playwright

• Given a playwright workflow test
• Toggle on recordHAR on browser object at start of test documentation: playwright record har
• This file can be piped into this script and keep your performance tests up-to-date with the workflow covered by the automated test.

Advanced usage

testCommon.js will fetch an Authorization Bearer token, and attach the header automatically on all httpRequest if the environment variable AT_AUTH_URL is set.
note: to set an environment variable in a command prompt wrap the command with double quotes if the value includes & characters.
e.g. set "AT_AUTH_URL=https://your.auth.url/oauth2/v2.0/token?username=myUserName&password=myPassword&etc_etc_etc"

the following variables are also available

• AT_VU_COUNT - defaults to 15
• AT_ITERATIONS - defaults to AT_VU_COUNT * 3
• k6 reference details here by default the testTemplate defines options to run a constant number of VUs a given number of iterations. More executor details (here)[https://grafana.com/docs/k6/latest/using-k6/scenarios/executors/]

testTemplate.js and testCommon.js are used by the script.
These files can be modified for your specific purposes e.g. hardcoded auth information, set the default iterations, etc.

Additional notes

The included testCommon.js and testTemplate.js files include helpers and reasonable defaults for problems we initially ran into when we started using k6 to performance test our single page app.

testTemplate.js

• includes an options block with a default of a fixed number of concurrent users (VUs) and a fixed number of iterations (per VU).
  • This was a good fit for us since our workflows were long running (10+ seconds), and we try testing our infrastructure with different fixed numbers of users, and ensure the infrastructure didn't fail or slow down terribly in any spots.
  • We used a fixed number of iterations (per VU) since the average times were skewed lower when iterations were forcably exited because they took beyond the grace period.
• includes handleSummary with custom junit reporter that provides duration and good visibility around call failures
  • we consume this in Azure Pipelines with the PublishTestResults@2 task, and pass testResultsFiles of *.junit.xml
• Review the contents of testTemplate.js for more details

testCommon.js

• includes commonSetup helper that delegates to _ensureLoggedIn(url)
  • _ensureLoggedIn makes a single call and expects the response to be an oauth token response, it saves the access_token to state. state is used by httpRequest to add an Authorization Bearer token when present.
  • httpRequest adds checks to each call that it return successfully (status >= 200 && status < 400) and timing < 10s
    • without these checks the performance tests could appear to succeed even though individual calls were beginning to fail, or take a long time.
  • Review the contents of testCommon.js for more details

Example code output

import { httpRequest, commonSetup, state } from '/testCommon.js'; //hack to reference the executing folder without needing relative path

const given_vus = __ENV.AT_VU_COUNT === undefined ? 15 : __ENV.AT_VU_COUNT;
const given_iterations = __ENV.AT_ITERATIONS === undefined ? given_vus * 3 : __ENV.AT_ITERATIONS;

export const options = {
	vus: given_vus,
	iterations: given_iterations,
};

export function setup() {
	commonSetup();
	return state;
}

export default function (setup_state) {
	Object.assign(state, setup_state);
	httpRequest('GET', 'https://status.k6.io/api/v2/status.json');
	httpRequest('GET', 'https://k6.io/data/jobs-positions.json');
}