Is definining scale-down-utilization-threshold on GKE possible?

Question

Can you even define the scale-down-utilization-threshold on GKE? I can't figure out where I would define this if it is actually possible.

Answer 1

I'm not sure if you mean scale down an HPA or scale down a VPA or scale down cluster nodes based on the cluster autoscaler, or scale down GCP instance groups or an AWS ASGs.

• A Kubernetes HPA supports it based on a metric (typically CPU utilization). It will scale down to a minimum number of replicas defined in the HPA Kubernetes resource.

• A Kubernetes VPA supports it based on requests/limits (CPU, Memory). Keep in mind that it's not supported if used together with an HPA that is also using CPU and memory as scale metrics.

• The Kubernetes cluster autoscaler also supports downscaling based on the --scale-down-utilization-threshold and --scan-interval options. It's not recommended to use this together with a GCP instance group or an AWS ASG.

• GCP Instance groups or AWS ASGs scale down is supported on the nodes provided that you are not using the cluster autoscaler and using CPU or other metrics specific to the cloud provider VM instances.

Hope it helps