Base kubernetes deployments allow for simple organisation of worker instance scalability. Combining the deployments with the ease of scaling the node groups within managed k8s offerings makes it simple to generate even the most demanding workloads.
The same base image can be used both for master and for worker pods. The actual workload, defined in load_test.py, can be injected into pods using a configmap. The image has been already built and made publicly available for your use. You are currently limited to using the dependencies that are already defined in the project and the tests can only be defined in load_test.py. You can build your own base image should you require other dependencies or additional files.
Before running deploying your locust workload into k8s, you should make sure that your cluster is addressable through kubectl. It is advisable to use a cloud provider managed service (EKS, AKS or GKE) for the k8s cluster, as they allow for scaling the nodegroups with minimal efforts.
Define your workload in the load_test.py and test in stand-alone mode locally. Set the CLUSTER_URL in secret.yaml, it should also contain your username and password. You can deploy the distributed setup into your kubernetes cluster using the following command:
./redeploy.shThis script will deploy a single pod with locust master and a deployment with three workers. All objects will be deployed into your current or default namespace.
You can also override the task weights in the worker-deployment.yaml should it be required. Please make sure your re-deploy the objects after changing the waits.
The following command can be used to forward the 8089 port from the master service to your localhost. This avoids the necessity of exposing your master service to the public internet:
kubectl port-forward service/master 8089:8089Please leave the terminal open for the time that you work with locust. The GUI can be accessed by navigating to http://localhost:8089.
The workers can be scaled up or down by using the following command:
kubectl scale deployment locust-worker-deployment --replicas 10The --replicas parameter specifies the number of desired worker instances. The kubernetes will automatically spread the worker pods between all available nodes in the cluster. Make sure you scale your node groups accordingly before running this command.