This page explains how Google Kubernetes Engine dynamically allocates resources for certain add-ons.
Add-on autoscaling
GKE uses add-on resizer to scale the metrics-server add-on and the heapster add-on. The add-on resizer scales the resource requests and resource limits of its managed containers in proportion to the number of nodes in the cluster. using the following formula:
$$ R=b+m \times \max(n,16) $$
Where the resource request or limit, \(R\), is equal to the base request/limit, \(b\), plus the marginal resource requirement, \(m\), times the node count of the cluster, \(n\).
If the cluster has 16 or fewer nodes, add-on resizer will assign resources as if the cluster has 16 nodes.
