GKE cluster autoscaler vs Autoscaling in Managed instance groups

Question

I am using Google Container Engine . Now I want auto scaling functionality in my cluster . As per documentation GKE autoscaler is in beta release . I can also enable autoscaling in instance group that is managing cluster nodes . Cluster autoscaler add/remove nodes so that all scheduled pods have a place to run where instance group add/remove nodes based on different policies like average cpu utilization . I think by adjusting pods CPU limit and target CPU utilization for pods in Kubernetes autoscaler , Managed Instance Group autoscaling can also be used to resize GKE cluster . So my question is what should I use ?

Answer 1

Short answer - don't use GCE MIG autoscaling feature. It will just not work properly with your cluster.

See details in this FAQ: https://github.com/kubernetes/autoscaler/blob/master/cluster-autoscaler/FAQ.md#should-i-use-a-cpu-usage-based-node-autoscaler-with-kubernetes

(read the question linked above and 2 next ones)