Kubernetes + locust (load testing)

the only way you can reset your dashboard of the master node for now is by rescheduling the master and starting with a clean pod. You can either do this with kubectl scale deployment/locust-master --replicas=0 and scale them back up with kubectl scale deployment/locust-master --replicas=1. But this will discard any results you already gathered on the master.

It's more a Locust problem then something that k8s can solve imo.

Issue seems to be that node once it tries to register and master if not up, wouldn't retry. Neither master does a constant communication to see i

Delete the master pod and wait for it come backup. Now it will have zero slaves.

Then delete node/worker pods so that they re-register. To delete pods with lables you can use below command

  # Delete pods and services with label name=myLabel.  kubectl delete pods,services -l name=myLabel

I think a similar issue is listed here. When you start a worker and shut it down it will still be listed in the UI as missing. I think this is because locust does not have a way for workers to inform the master that they are not ready for work. As said before deleting your master node and then restarting it will reset the UI.

In the meantime I think the developers know about the issue.