Hands-On Microservices with Kubernetes

Using the horizontal pod autoscaler

We can use kubectl for autoscaling. Since the autoscaler relies on Heapster and the metrics server, we need to enable them using the minikube addons command. We have already enabled Heapster, so this should be good enough:

$ minikube addons enable metrics-server  metrics-server was successfully enabled

We must also specify a CPU request in the pod spec of the deployment:

    resources:       requests:         cpu: 100m

As you may recall, a resource request is what Kubernetes promises it can provide to the container if it is ever scheduled. This way, the horizontal pod autoscaler can ensure that it will start a new pod only if it can provide this requested minimum of CPU to the new pod.

Let's introduce some code that will ...