Skip to content

Latest commit

 

History

History
49 lines (31 loc) · 2.06 KB

README.md

File metadata and controls

49 lines (31 loc) · 2.06 KB
Raw

Chaos Monkey

This is a simplistic implementation of https://github.com/Netflix/chaosmonkey.

This program runs inside a Kubernetes cluster, in a given namespace and deletes a random pod.

❯ ./pod-chaos-monkey --help
pod-chaos-monkey is a CLI tool used to kill a random pod in the given namespace. 
    
    The default namespace is "default", but it can be changed via CLI args.

Usage:
  pod-chaos-monkey [flags]

Flags:
      --label-selector string   Label selector to skip some pods from being deleted (default "metadata.name!=pod-chaos-monkey")
  -h, --help                    help for pod-chaos-monkey
  -n, --namespace string        Target namespace (default "default")

How to deploy in a cluster

In the deployment folder you may find some Kubernetes resources needed to run this program in your cluster.

  • pod-chaos-monkey.yaml contains the cronjob definition. Feel free to change the spec.schedule according to your needs. By default it runs once per minute. It is also possible to configure the target namespace and label selector. Just update the values in the args.
  • service-account.yaml contains the service account, role and role binding used for this cronjob. The service account has list and destroy pods privileges for all the namespaces in the cluster.
  • example.yaml deploys 4 replicas of a Nginx container. Used to see the PodChaosMonkey in action. They are deployed in the "workloads" namespace.

Then you just have to connect to your cluster and run:

kubectl apply -f deployment

Local development

Spin up a Kubernetes cluster using your favorite tool. For example, you can use Minikube:

minikube start

You can then deploy the changes quickly by running ./deploy.sh. Make sure to uncomment imagePullPolicy in pod-chaos-monkey.yaml in order to use the local version of the Docker image.

TODO list

  • Add support for running the program outside the cluster for easier debugging.