This is a simple tool to save kubernetes resource costs. It will scale down deployments and statefulsets to 0 replicas at a scheduled time every day. It will also scale up the deployments and statefulsets to the original number of replicas at a scheduled time every day/days of a week. Primarily built to seemlessly tackle spinning down dev environments at night and weekends.
Hibernate simply uses scheduled cronjobs consisting of a go binary which interacts with the kubernetes api to edit the replica count for deployments and statefulsets.
When a resource is scaled down to 0, it will store its previous number of replicas in a ConfigMap which subsequent wake calls will be able to read. This is to ensure that the resource is scaled up to the previous number of replicas.
Furthermore, since this is a cronjob based solution, there is no need to run a long running controller process to watch for changes. This is a light weight solution to scale down dev infrastructure costs.
If you do not want to use scheduled cronjobs but would rather use this to automate mass disabling of resources via a single config file, you may opt in to use the binary instead, as the standalone binary can be used as long as you have a valid kubeconfig file.
This builds the go binary
git clone git@github.com:Performl/hibernate.git
cd hibernate
make binTODO# prerequisites
kubectl create namespace hibernate
# default manifests
kubectl apply -f https://raw.githubusercontent.com/Performl/hibernate/master/manifests/main.yaml# make sure hibernate namespace exists
kubectl create namespace hibernate# to use the binary to sleep resources manually using your local kubernetes config run:
./hibernate --mode=local --action=sleep
# to run in a kubernetes cluster run:
./hibernate --action=sleep
# to wake resources
./hibernate --mode=local --action=sleep
or
./hibernate --action=wake