I am new to EKS and looking for the number of pods per node and sizes of EC2 instances for nodes, recommended by AWS in EKS for better performance and HA? I found limitations set by Kubernetes.io in here. But I want to know what AWS's schools of thoughts when we run our clusters with EKS? You may share your experience too.
This is not for polling, but to know the standard usages.
There is a hard limit to the number of pods that can be run on a particular worker instance type. This is because, by default, Amazon's VPC CNI assigns a subnet IP to each pod. This page lists how many interfaces/ips per interface a particular instance type can have. One interface is reserved for the host, so you get your answer by ( Maximum Network Interfaces ) * ( IPv4/6 Addresses per Interface ) - 1. For example, with a t2.medium, you get 17 pods.
Picking the right type, as mentioned by RickyA, is an exercise you will have to go through.