Review Misses Against CKAD Domains

Why this matters

Almost-right YAML feels one retry away and teaches the wrong lesson: the shape was wrong because the model was wrong. A probe miss is rarely indentation; it is the handler-effect-shape model from M05 never learned. Classification directs the drill at the model, and models transfer across tasks.

Concepts

Two poles plus time: shape misses (right model, wrong fields or nesting) drill with generation-plus-verify loops; model misses (wrong mental picture: labels select, fields merge, subjects bind) drill with explain-first recitation; time misses (skill present, clock absent) drill with caps and flags. Then weight the pile by CKAD domains: application design plus observability at 40 percent dominate any review week.

Each miss becomes one drill under five minutes: shape drills regenerate the manifest three times from scratch; model drills recite the rule then apply it once; time drills re-run the task with a visible cap. The log retires drills after two clean passes, same as the CKA lessons.

Worked example

Four misses from a developer mock, reviewed:

Miss 1 (observability, model): liveness killed a slow starter -> startupProbe model drill, recite + apply
Miss 2 (design, shape): Pod fields nested under template wrongly -> 3x generation drill from run --dry-run
Miss 3 (config, time): secret key typo ate 11 min -> verify-first rule card: describe secret before wiring
Miss 4 (services, model): selector pointed at nothing -> label-verdict drill: get pods --show-labels first
Weighted: observability + design = 3 of 4 -> next week is M04-M05 drills only

Same one-page shape as the CKA review. The poles decide the drill kind; the weights decide the week.

Exam notes

Shape misses seduce with near-misses: resist tuning fields and drill the model instead. Time misses in authoring hide as shape misses (rushed YAML looks wrong-shaped): ask whether calm hands would have written it right. And model drills need recitation aloud: silent re-reading rebuilds nothing.