Read the Borg paper for an idea of what k8s is meant to solve. If you have hundreds of services spread across 10s/100s of nodes, then the complexity of k8s starts to become amortized. Similarly if you have 100s of devs without a consistent way to do deployments, k8s starts to make sense.
Source: tech lead for a team managing a large on-prem and cloud k8s cluster.
Out of the couple of stacks I tried, kops was the easiest to set up. Juju and kubeadm were so-so.
https://kubernetes.io/docs/getting-started-guides/fedora/fed...
If you manage to automate deployment of the VMs or physical nodes (e.g., PXE), you've got a cherry on top.
highly agreed. I'm actually surprised that most people dont try Docker Swarmkit. It is unbelievably good for what it does.