Deployment & Infra
Deployment capstone: free-recall review
Retrieval beats re-reading. Each prompt asks you to reconstruct a contract that connects two or more stages of the deployment chain. Say or write a full answer from memory first, then open the model answer and compare.
Reconstruct the spine of the whole track without looking back: why a release is one composed object, and what enforces the contract at each seam between stages.
- 01Explain why a release is one composed object rather than seven independent steps, and where its failures live.
- 02Why can a perfectly correct rolling update still cause an outage, and what single piece of configuration prevents it? Contrast readiness and liveness while you are at it.
- 03Define expand-contract and explain why it is what makes the rollout strategy and a database migration safely composable.
- 04Even with probes and safe migrations, a cutover can sever in-flight requests. Explain the race and the two-sided fix involving L7 and SIGTERM.
- 05Why must secrets be injected at deploy/runtime rather than baked into the image, and what is the catch with a Kubernetes Secret?
- 06What is configuration drift, why is it dangerous even when the service is up, and how does IaC close the loop with the rest of the chain?
If you could reconstruct each answer from memory, you hold the track’s spine: a release is one composed object, and you engineer the seams — an immutable digest fixes the artifact, the readiness probe makes “ready” mean “serving,” expand-contract keeps the rollout and migration N-1 compatible, L7 draining plus SIGTERM handling protect in-flight requests, runtime secret injection keeps credentials rotatable and off the image, and IaC as the single source of truth kills drift. The links were never the hard part; the contracts between them are.