mms-gianni/locust-admin

GUI and API to parallelizing locust.io loadtests on kubernetes

Locust Swarmadmin

Locust is an easy to use, scriptable and scalable performance testing tool. But it is not very flexible. Specially when it comes to run multiple different configured loadtests in kubernetes cloud.

This project solves this problem by adding a additional admin interface and API to dynamically start Locust instances in your kubernetes cluster.

All configurations are stored on your kubernetes cluster.

API Swagger documentation : https://mms-gianni.github.io/swarmadmin/swagger.html

Installation on Kubernetes

The Helm installation is based on Deliveryhero's helm charts. (Version 0.20.2)

cd k8s/
helm install -f values.yaml locustswarm . -n locustswarm --create-namespace
helm upgrade -f values.yaml locustswarm .

The results are exported and stored in Prometheus. They can easily be displayed on a grafana board

https://grafana.com/grafana/dashboards/15109

Docker container environment variables

For a full list of available environment variables, please visit .env.example

KUBECONFIG_BASE64

A base64 encoded kubeconfig

export KUBECONFIG_BASE64=$(cat kubeconfig | base64)

KUBECONFIG_PATH

Fallback if KUBECONFIG_PATH is empty

Development

running a local KiND cluster

kind create cluster --config kind.yaml
kubectl apply -f https://raw.githubusercontent.com/kubernetes/ingress-nginx/main/deploy/static/provider/kind/deploy.yaml
kind export kubeconfig --name superlocust --kubeconfig ./kubeconfig

Make sure you are using the internal kubeconfig for the ENV variable to configure the helm charts

export KUBECONFIG_BASE64=$(kind get kubeconfig --name superlocust --internal | base64)

start on local node

cd client && npm run build && cd .. && npm run serve

Possible future features

• Store loadtest results in a database (MongoDB? or Prometheus? https://github.com/siimon/prom-client) or a Prometheus exporter
• Add provision support for other clouds (AWS, Google, Azure, Digital Ocean ... ) won't be implemented since it would be to complicated to maintain