Use describe output, events, logs, exec, port forwarding, ephemeral containers, node diagnostics, and network tests to isolate failures. Work through Pending Pods, image-pull errors, crash loops, DNS failures, and unreachable Services.