Implementation, Configuration, and Operations
What this section covers
This domain rewards process discipline. Understand infrastructure as code: declarative definitions, version control, peer review, and idempotency - applying the same configuration twice yields the same state. Know how configuration drift accumulates and why immutable/golden-image replacement often beats in-place patching. Study environment promotion (dev → test → prod), CI/CD pipeline stages, and change management with a documented rollback path. On the operations side, master observability: metrics for trends and thresholds, logs for detail, traces for request flow - and alert on symptoms users feel, not on every internal counter. Include backup scheduling plus restore testing, patch windows, capacity monitoring, and runbooks that a tired on-call engineer can follow at 3 a.m.
Traps
- "Verify first": confirm current state before changing anything.
- Order-of-operations questions - backup/snapshot before the destructive step.
- Scripts that are not idempotent and fail or duplicate on re-run.
- Choosing "replace the server" when the question stresses minimal disruption to a live service.
- Confusing a rollback (revert the change) with a roll-forward (fix forward quickly).
Worked example
A deployment causes errors after 5% of traffic shifts. Correct sequence: pause the rollout, verify the symptom and blast radius in metrics/logs, roll back to the last known-good version, then diagnose in a non-production environment. Rule of thumb: when a change correlates with a new failure and a fast revert exists, revert first and investigate second.