Continuous Integration and Load Testing

Continuous integration (CI) is a widely accepted way to speed up your software development process, by merging code frequently and testing through automation. Many teams run automated functional test suites as part of their build and deployment process.

It’s also a great idea to include load and performance testing in your continuous integration pipeline! Automating load tests with every build or deployment can help you detect performance regressions quickly, and also uncover certain kinds of bugs that might only surface when the system is under load.


Every one of your Loadster scenarios has a unique trigger code or URL that you can use to remotely launch the test. You can find this trigger for a scenario by expanding the dropdown on the Launch Test button.

Obtaining a trigger code or URL for remotely running a scenario
Obtaining a trigger code or URL for remotely running a scenario

This trigger code is unique to the scenario, and does not require separate authentication apart from being long and random and difficult to guess, so be sure to keep it safe as you would any other secret credential.

Running Tests with the Loadster CLI

The Loadster CLI is the preferred way to launch Loadster tests from your CI pipeline, build server, or cron job. It’s available as a self-contained native binary for Windows, Mac, and Linux, and makes running and observing load tests a breeze.

Download the Loadster CLI for Windows, Mac, or Linux

To install it, simply download and extract the archive, and place the executable anywhere on your filesystem.

Starting a Test With the Loadster CLI

To run a test with the Loadster CLI in your PATH, you’ll need the scenario’s trigger code as shown above. You can then start a test with loadster start <trigger-code> as shown here:

$ loadster start WsHDupkZEYkSD7bT

Test launched! To view it in your browser:

Fetch the current test status as JSON at any time:

The CLI can also produce output in JSON format rather than human-friendly text, if you use the --json switch:

$ loadster run WsHDupkZEYkSD7bT --json

  "message": "Test launched!",
  "reportUrl": "",
  "statusUrl": ""

Including a Label

Sometimes it’s helpful to tie a load test back to a specific build number or configuration. You can do this with the --label switch.

$ loadster run WsHDupkZEYkSD7bT --label=ci-build-238-beta

Afterwards, this label will show up on the test report and in your Activity Feed, so you can easily reference load test results back to the relevant build.

Blocking and Observing Test Status

To make the CLI block until the test finishes instead of exiting immediately after starting the test, use the run command instead.

$ loadster run WsHDupkZEYkSD7bT                   # text output
$ loadster run WsHDupkZEYkSD7bT --json            # json output

When you use the run command (instead of the start command), the CLI will keep watching the test progress, and print useful stats every few seconds. For example:

 - runningUsers: 175.00
 - responseTimeAverage: 0.25
 - responseTimeP90: 0.71
 - uploadThroughputBytesPerSecond: 2437.50
 - downloadThroughputBytesPerSecond: 190337.25
 - pagesPerSecond: 11.08
 - hitsPerSecond: 32.17
 - totalPages: 2660.00
 - totalHits: 11320.00
 - totalErrors: 1.00
 - totalIterations: 156.00

When the load test finishes, the CLI will exit.


The Loadster CLI also supports assertions, so you can evaluate high-level test result metrics against pass-fail criteria. This is useful for automatically failing a build if the performance test results aren’t inline with expectations.

$ loadster run WsHDupkZEYkSD7bT --assert 'totalErrors == 0' --assert 'avgHitsPerSecond >= 24.5'

The example above makes two assertions: the total number of errors in the test must be equal to zero, and the average hits per second throughout the test must be more than 24.5. Assertions are not required, but you can use pretty much as many as you want.

Assertions work on the following metrics:

  • maxUsers
  • avgBytesPerSecond
  • maxBytesPerSecond
  • avgPagesPerSecond
  • maxPagesPerSecond
  • avgHitsPerSecond
  • maxHitsPerSecond
  • totalBytesTransferred
  • totalPages
  • totalHits
  • totalErrors
  • totalIterations

The following operators are supported in assertions:

  • ==
  • <=
  • >=
  • <
  • >

At the end of the test, a breakdown of which assertions passed/failed is printed. If any assertions failed, the process exits with a non-zero status code.

Running Tests with Your HTTP Client

If you don’t want to use the Loadster CLI or it’s unavailable on your platform, you can also run and observe tests with your HTTP client of choice. For this example, we’ll use curl.

Starting a Test

Making a POST to your scenario’s trigger will cause it to start the test:

$ curl -X POST
  "message": "Test launched!",
  "reportUrl": "",

The JSON response includes three things:

  • message - A human-readable message, suitable for showing in a simple test runner UI or something.
  • reportUrl - A URL to the test report, where you can watch live as it runs, or view the final report once finished.
  • statusUrl - A URL for fetching the current test status (read on).

If you want to keep an eye on the test as it runs, or cause your CI process to wait for the test to finish, you or your CI script will want to make note of these URLs.

Including a Label

When starting a test with your own HTTP client, you can specify a label by including a label parameter when you POST to the trigger.

$ curl -X POST

Afterwards, this label will show up on the test report and in your Activity Feed, so you can easily reference load test results back to the relevant build.

Getting Test Status

Immediately after launching a test, you can query its current high-level status at any time by hitting the statusUrl that was returned by the trigger.

$ curl
   "hitsPerSecond" : 24.6666666666667,
   "downloadThroughputBytesPerSecond" : 7520,
   "totalErrors" : 1754,
   "pagesPerSecond" : 24.6666666666667,
   "failed" : false,
   "errorsByType" : {
      "HTTP 404: Not Found" : 1755
   "engines" : [
         "id" : "a6ebd079-f4a2-4137-a23e-642dd2b27f10",
         "online" : true,
         "name" : "N. America - Virginia - 1",
         "utilization" : 11.4286739959778,
         "region" : "north-america-1"
   "downloadBytes" : 554880,
   "canceled" : false,
   "uploadBytes" : 455220,
   "responseTimeAverage" : 0.0390824742268041,
   "elapsedTime" : 103997,
   "populations" : [
         "runningUsers" : 25,
         "peakUsers" : 25,
         "name" : "Population",
         "id" : "f9f88cea-3226-4265-9c0a-94855687a0cb"
   "totalIterations" : 1754,
   "downloadThroughputBitsPerSecond" : 60160,
   "responseTimeP80" : 0.065,
   "running" : true,
   "finished" : false,
   "uploadThroughputBytesPerSecond" : 6171,
   "responseTimeP90" : 0.0682,
   "totalPages" : 1754,
   "started" : true,
   "totalHits" : 1754,
   "uploadThroughputBitsPerSecond" : 49368

This status includes lots of interesting things to tell you how the test is running. The started, running, and finished booleans tell what stage it is in its lifecycle. If you need your continuous integration process to wait for the test to complete, you could poll this status and wait until finished is true.

The other fields contain high-level metrics in your test, and they will change as the test runs. They contain a current snapshot of throughput, response times, errors, etc.

These numbers may occasionally be interesting to humans, but you could make your build script look at them too. You could even configure your CI build to pass or fail depending on the outcome of the test. For example, you could fail your build if there are errors (totalErrors > 0) or too many slow responses (responseTimeP90 > 1.5).

Tracking Test Results

Finished tests will show up in the Activity Feed of your dashboard, alongside tests that were started manually by you and others on your team.

Automated tests show up in the Activity Feed
Automated tests show up in the Activity Feed

Next Steps

There are about a million ways to do continuous integration. We haven’t tried all of them. If you have feature requests or questions, please get in touch with and let’s work together on it!