Kubernetes Production Best Practices

What the checklist is

Getting a container running on Kubernetes is easy. Getting it to survive a rolling update, a node failure, a traffic spike and a security review is not, and the gap is usually a list of small settings nobody told you about. Learnk8s, a Kubernetes training company, publishes this checklist to close that gap. It is free, open source on GitHub, and needs no form or email: you read it in the browser and can adapt it for your own team's reviews.

What it covers

The checklist has 40 checks in five topics:

• Your application: logging to stdout and stderr, configuration from environment variables or files, handling SIGTERM and shutting down gracefully, exposing health signals, not storing state on local disk, handling long-lived connections, and container images that contain only what is needed with stable tags instead of :latest.
• Your Kubernetes manifests: readiness, liveness and startup probes, resource requests and limits, bounded ephemeral storage, explicit rolling update settings, tolerating old and new pods side by side, reloading ConfigMaps and Secrets, running as non-root with a read-only filesystem, PodDisruptionBudgets, spreading pods across nodes and zones, meaningful labels and supported API versions.
• Your security: Pod Security Standards per namespace, a dedicated ServiceAccount with minimal RBAC, NetworkPolicy, scanned images from a trusted registry, admission policies, workload identity for cloud access and an external secret store.
• Scaling: horizontal scaling, explicit autoscaler bounds, when vertical scaling makes sense, requests based on real usage, priority classes, clean scale-down and load testing.
• Going live: visible health, collected Kubernetes events, a rollback or roll-forward decision, knowing what happens when a pod or node dies, a troubleshooting runbook and a cost review.

Is it worth it?

Yes. It is one of the few resources that turns scattered Kubernetes documentation into a single, ordered review. Each check is short and practical, and together they read like the questions a senior platform engineer asks in a production readiness review.

Prerequisites and time

You should already know pods, deployments, services and kubectl, and have deployed something to a cluster, even a local kind cluster. Reading the whole list takes two to three hours. Applying it to a real service takes a day or two the first time and much less after.

What you can do after it

• Run a production readiness review on any Kubernetes service and write down what is missing.
• Turn the checks into admission policies or CI checks so new services start compliant.
• Build a golden path template for your platform that passes the checklist by default.

Common questions

Is it the same as the official Kubernetes docs? No. The official docs explain each feature; this list tells you which features a production workload needs and why.

Did the site move? The original learnk8s.io address now redirects to learnkube.com, the company's newer domain. The content is the same checklist.

Best for: developers shipping their first services to Kubernetes and platform teams who want a shared review standard.