Containers in production, from a single host to Kubernetes

Running one container on your laptop is easy. Running twenty across three servers, with restarts, rolling updates and logs you can actually find, is a different job. This section is about that second job.

We cover container runtimes, resource limits, health and readiness probes, rolling and blue-green deploys, and how logs and metrics leave a container. We'll write about Kubernetes, and also about when you don't need it. A small team with one app may be better served by Docker Compose on a single well-monitored host, or a managed container service, than by a cluster they have to patch and upgrade themselves.

When we do cover Kubernetes deployment, we start from the failure cases: pods stuck in CrashLoopBackOff, a readiness probe that lies, a rollout that never finishes. Knowing how it breaks is how you learn to run it.