Automated load testing with locust

As a dev-team, we wanted to know how good our new application is doing under load and how changes to the code-base or the infrastructure affects the response time.

We decided to look out for a load-testing solution that was able to create a high amount of load when we need it. Our research brought us to https://locust.io/ and it looked like a promising candidate.

We set up a k8s-cluster with terraform and threw in a controller- and a bunch of worker-deployments of locust. Locust makes it easy to record a test profile with your browser and use it with the controller later on.

When the tests are finished you can export a result in different file formats. We decided to go for .csv.

The next step was to put the results in a time series.

We created a BigQuery dataset that represents the load-test result. A CloudFunction is used to import the results into the dataset so we can use them for our review.

Last was a report made with Data-Studio that is linked to the dataset in Big-Query. This report updates every time a new result was pushed to the dataset.

Put that all together we got a pretty nice solution that let us keep an eye on the performance of our application while it is under constant development.

graph LR; A[Locust]-->B[Google-Bucket]; B-->C[BigQuery] C-->D[Data-Studio]

used tools and technologies #

google cloud platform
BigQuery
Google Data Studio
Google Kubernetes Engine
Google CloudFunctions
Locust ( https://locust.io/)
Terraform