Podcalypse is a simple k8s tool that kills pods at a given constant rate, provided they match a label.
This is useful when you want to check that your system (e.g. your load balancer) is configured to tolerate a given level of disruption (usually caused by rolling upgrades).
It's not meant to be a generic chaos-monkey tool, but rather a more focused stress-testing tool.
This project currently has no releases; you need to build it yourself. The easiest way is to use
the ko project and do:
KO_DOCKER_REPO=your/docker/registry/image/name ko apply --strict -f deploy.yaml
this will compile the Go code in this repo, push a docker image, splice that image name into the deploy.yaml file and finally apply it to your current k8s cluster context.
Podcalypse will periodically kill all running pods that have the followin label:
labels:
mkm.pub/podcalipse: "true"
just apply this to the pods you want to test and run the podcalypse controller.
The main knob to tune is the PODCALYPSE_RATE environment variable which controls how often (per second) will a pod be killed.