Summary

In this chapter, we learned how to determine the compute and network requirements of a Kubernetes cluster. We also touched upon the software requirements that come along with it, such as etcd, and how odd-numbered clusters are preferred (due to the census algorithm) as the cluster needs to achieve more than 50% of votes for consensus.

The etcd cluster can either run on the kube-apiserver or have a separate set of clusters dedicated just for etcd. When it comes to resources, 2 CPUs and 8 GB of RAM should be enough. When deciding on the storage system for etcd, opt for lower latency and higher IOPS storage such as SSD. We then jumped into sizing the kube-apiserver, which can be run alongside etcd. Given that both components can coexist, resources should be bumped to anything between 8 and 16 GB of RAM and between 2 and 4 CPUs per node.

In order to properly size the worker...