The CKAD Game Plan: Weights, Order, Verification

Why this matters

Developer exams punish perfectionism harder than admin ones: a beautiful Deployment that took twenty-five minutes fails the exam it would ace untimed, because three medium tasks went untouched. The weights tell you the exchange rate between minutes and points. Spending by weight is the highest-leverage decision in the room.

Concepts

The published weights shape the budget: Application Design and Build (20%), Application Deployment (20%), Application Observability and Maintenance (15%), Application Environment, Configuration and Security (25%), Services and Networking (20%). Configuration is the heaviest single domain, so ConfigMaps, Secrets, security contexts and resource boundaries deserve the deepest fluency, not just familiarity.

Harvest pass first: every task solvable in under five minutes gets solved immediately, regardless of weight. This banks points, warms up the keyboard, and surveys the whole exam while fresh. Then weight order for the rest, heaviest domain first, with partial attempts everywhere: a half-configured NetworkPolicy-adjacent task still earns more than a perfected trivial one.

Stuck rule, no exceptions: fifteen minutes without progress means flag and move. CKAD tasks chain (a Service question assumes the Deployment from two tasks ago exists), so a block early compounds; returning later with context from sibling tasks dissolves most blocks.

Verification is type-matched. Created a workload: get pods plus logs. Exposed it: get svc,endpoints plus a curl from a temp pod. Configured env or volumes: exec env plus exec ls the mount. Fixed a probe: describe the events. Never trust the apply output; the grader reads state, so you read state first.

Worked example

Minute budget for the standard spread:

minutes 0-10:   skim all tasks, harvest the trivial ones
minutes 10-75:  configuration + deployment domains (45% of weight)
minutes 75-100: services + design + observability
minutes 100-105: remaining flags, partials everywhere
minutes 105-120: verification sweep only, zero new tasks

The sweep is one read per task in task order. A service pointing at the wrong port selector is a zero that a ten-second endpoint check rescues.

Exam notes

Namespaces and contexts switch between tasks: confirm both before every task, not just the first. Image tags and registry paths come from the question text; copying them exactly beats memory every time. And the free retake covers this exam too, so attempt everything: a guessed field on a heavy task beats a skipped task with certainty.