k8s-loadtest

A simple load testing framework for Kubernetes workloads using locust.

Kubernetes enables rapid deployment of server-based applications on top of elastic infrastructure that can scale with your needs. An important part of a rapid continuous deployment strategy includes proof that your application can scale and remain stable under load. This project includes a re-usable locust container image Dockerfile and a re-usable deployment configuration using Helm that will get you load-testing on your Kubernetes clusters in no time.

locust is a python based load testing tool that runs in a master/worker configuration. To setup your loadtest, you use a simple python library to define tests and then launch the master and workers with the file containing the main entry point for your test.

Please see the Locust Home Page for more information on creating and running your load tests.

This framework simplifies the use of locust in kubernetes by offering a helm deployment that will get a load test out of your git repository before running it on the generic locust container defined in this project's Dockerfile.

Prerequisites

If you're starting from scratch and don't have a Kubernetes cluster, or just want to test it locally, install minikube. Minikube runs on a VM on your laptop and is easy to install.

You will find more information on installing minikube here

Next you'll need two other tools:

1. kubectl - is the CLI tool for kubernetes and can be found here
2. helm - is the deployment tool for kubernetes. More information and download links can be found here

Quickstart

Once you've got your cluster started and have installed helm and kubectl, you can become familiar with this framework by using the supplied sample-app found in the contrib/sample-app folder.

Step 1: Setup your git ssh key, so kubernetes call pull your loadtest

kubectl create secret generic git-creds --from-file=ssh=~/.ssh/id_rsa

Step 2 Startup the sample server with helm

helm install contrib/sample-app/charts/sample-app --name sample-app

Step 3 Startup the load test with helm

helm install charts/locust-loadtest -f contrib/sample-app/locust-values.yaml --name locust

Step 4 Verify that its all running and get the URI for locust

Open up your kubernetes dashboad (on minikube you type: minikube dashboard), select the default namespace and check that you have:

• 2 sample-app pods
• 1 locust master pod
• 2 locust worker pods

Next click on Services and then find the NodePort where locust's control panel is exposed. This will be a port that is > 30000

Step 5 Start Testing:

1. Load the Locust Page. If you're using minikube, it will use the same IP as your dashboard followed by the NodePort of the locust service.
2. Setup the number of users to simulate: 100
3. Setup the Hatch Rate: 2

Now you should see the test page reporting 2 Slaves along with updating metrics for your test.

Load Testing

TODO: Describe the deployment in more detail

Write Your Own Loadtest

TODO: Describe extending the sample and writing a load test form scratch