A collection of cluster reliability tools built for Kubernetes
Governor is a collection of tools for improving the stability of the large Kubernetes clusters as a single Docker image.
Two common problems observed in large Kubernetes clusters are:
- Node failure due to underlying cloud provider issues.
- Pods being stuck in "Terminating" state and unable to be cleaned up.
node-reaper provides the capability for worker nodes to be force terminated so that replacement ones come up. pod-reaper does a force termination of pods stuck in Terminating state for a certain amount of time.
Assuming an AWS-hosted running kubernetes cluster:
kubectl create namespace governor
# Using a CronJob
kubectl apply -n governor -f https://raw.githubusercontent.com/keikoproj/governor/master/examples/node-reaper.yaml
kubectl apply -n governor -f https://raw.githubusercontent.com/keikoproj/governor/master/examples/pod-reaper.yaml| Package | Description | Docs |
|---|---|---|
| node-reaper | terminates nodes in scaling groups | node-reaper |
| pod-reaper | force terminates stuck pods | pod-reaper |
Please see CHANGELOG.md.
Please see CONTRIBUTING.md.
Please see DEVELOPER.md.