Troubleshooting Services and Networking

Your Service is returning "connection refused." Or worse, it just hangs. nslookup inside the Pod returns NXDOMAIN. You have Pods running, a Service defined, and no idea where the breakdown is. This is the most common debugging scenario in a Kubernetes cluster, and it falls squarely in the CKA exam's Troubleshooting domain: 30% of the total score.

The problem with networking failures is that they can break at any of five distinct layers, and each layer has different symptoms and different tools. A DNS failure looks identical to a missing Service from the caller's perspective. A NetworkPolicy silently dropping traffic produces the same timeout as a misconfigured kube-proxy.