production-readiness

Nothing ships without a checklist, a rollout plan, and a way back.

Inspection

Inventory deployables: health probes, metrics/alarms, logs, secrets wiring, limits, runbooks, rollback artifacts, on-call assignments.

Decision rules

Ship only when checklist, staged rollout, and rollback all exist; flags for behavior changes; halt criteria defined before rollout starts.

Activate when

• Preparing any production deploy or judging go/no-go.

Do NOT activate for

• CI mechanics (see tooling/ci-cd); release tagging (see git/git-release).

Procedure

1. Checklist: health/readiness probes real, metrics/alarms on, logs structured, secrets externalized, limits/quotas set, runbook written.
2. Rollout: feature flags for behavior changes, staged percentages with error-budget watches, automatic halt criteria.
3. Rollback: previous artifact one step away; drill before needed; data migrations reversible or explicitly irreversible with owner sign-off.
4. Verify: pre-launch review signed, monitors quiet in staging, rollback path tested, on-call named.

Failure modes

Always-true healthchecks; unmonitored deploys; irreversible migration without sign-off; on-call unnamed; flags without cleanup owners.

Escalation

Infra specifics → cloud/*; release tagging → git/git-release.