Kubernetes (down)scaling: combining autoscalers for minimal resource allocations
You can reduce costs by turning off workloads when you don’t need them. For example, you might want to avoid running any workload in a dev cluster during the evening and reduce the number of worker nodes. Or you could run a web service only where there are actual requests. In all of those cases, you need to balance the need to scale the number of replicas and worker nodes.
In this lab, you will learn:
- How you can scale your workloads with the horizontal autoscaler.
- How the cluster autoscaler works (e.g. the autoscaler does not look at memory and CPU).
- How to combine the horizontal and cluster autoscaler to expand and shrink your cluster nodes efficiently.