I've already foreshadowed a bit of this post in my previous, when I mentioned trouble when not starting a second node. So, I tried to only start one node, and while the k8s side is fine with running on a single node, the pods that were on the not-restarted worker was less happy. It eventually declared the worker nodes as unhappy, and tried to terminate the pods. But I'm unsure if it will ever do anything about that; maybe I just didn't wait long enough. One thing I have learned is that while there are lots of cool things about containers, Docker and Kubernetes -- being blazing fast isn't the top of the list. Now, starting an instance on AWS isn't instant, but pulling down everything Kubernetes requires and installing it takes more than a few minutes. Starting even a simple pod/image also takes some time. When running, things aren't noticebly slow, so I'm not complaining about that, but cluster operations need a bit of patience. It's not bad that things take time, it's something to adjust to.