Running clusters in highly dynamic contexts means that we often run stuff we don't really need. Wasting money is stupid and we really want to make sure that we only run stuff when we need to. This little piece of code just scale down ASGs based on day and time to allow us to have clusters that are available only during working hours.
The options should be self explanatory, if not please submit a bug report. There are however different options to deploy:
- run it on a master node: in this way we will need to pass the ASG name as parameter (we never want to scale down to 0 the master!)
- run it on a worker node: in this case we can use the autodiscovery
An example deployment file is the following:
apiVersion: extensions/v1beta1
kind: Deployment
metadata:
namespace: kube-system
name: downscaler
labels:
k8s-app: downscaler
spec:
template:
metadata:
annotations:
scheduler.alpha.kubernetes.io/critical-pod: ""
labels:
k8s-app: downscaler
spec:
# run on each master node
nodeSelector:
node-role.kubernetes.io/master: ""
tolerations:
- effect: NoSchedule
key: node-role.kubernetes.io/master
- key: CriticalAddonsOnly
operator: Exists
hostNetwork: true
containers:
- name: downscaler
image: x0rg/kube-nodes-downscaler:v0.0.5
imagePullPolicy: Always
args:
- --start=7
- --end=17
- --asg-name=ASGN_NAME
resources:
requests:
memory: 20Mi
cpu: 10m
limits:
memory: 20Mi
cpu: 100m
Images are provided on Docker Hub. Images with the latest tag are not published on purpose to be able to easily figure out what is running in production and not rely on a tag that gets rewritten. You can find what is the latest imaget tag by looking at this page.
A Helm chart is available. You can install it using:
helm repo add raffo-kube-nodes-downscaler https://raw.githubusercontent.com/Raffo/kube-nodes-downscaler/master/charts/
helm install --name kube-nodes-downscaler raffo-kube-nodes-downscaler/kube-nodes-downscaler