Implementation and Configuration
This domain tests whether you can move from an approved design to a repeatable, auditable running state. Expect questions on infrastructure as code, environments, secrets, deployment strategy, and configuration control.
Core rules
- Version-control everything that defines an environment: application code, configuration, policy, and schemas. Manual console changes are exceptions with a ticket and a backport.
- Idempotency means applying the same configuration twice yields the same state, not duplicates. Declarative tools converge; imperative scripts drift.
- Environments are tiers such as development, test, staging, and production, with promotion gates. Parity reduces 'works in test' surprises.
- Never store secrets in repositories or images. Use a secrets manager with short-lived credentials and rotation.
- Deployment strategies include rolling, blue/green, and canary. Define success metrics and automatic rollback triggers before deploy.
- A configuration baseline records the approved state. Drift detection compares actual to baseline.
- Require peer review and automated checks: lint, policy-as-code, unit tests, and integration tests before merge.
- Change records tie commit, ticket, approver, and deploy time for audit.
Worked example
A script appends a user to /etc/app/users. Running it twice creates duplicate entries. The declarative fix is a desired-state file that lists the exact user set, and a configuration tool that enforces it. Reasoning: the first run changes state; the second run converges. Acceptance test: run the tool twice, compare checksums and user count. Expect no change and no duplicates.
Traps
- Storing secrets in Git and assuming deleting the file removes history.
- Hotfixing production directly and losing the change from source control.
- Allowing environment drift until staging no longer predicts production.
- Running a canary without defined success metrics or automatic rollback.
- Treating 'it deployed' as success instead of verifying user-facing behavior.
Key points
- Version-control code, configuration, policy, and schemas together.
- Idempotent, declarative configuration converges; imperative scripts drift.
- Secrets belong in a secrets manager, never in repositories or images.
- Canary and blue/green deployments need success metrics and rollback triggers.
- Configuration baselines and drift detection keep environments aligned.