ftb-project-management/docs/production-readiness.md

Production Readiness Checklist

Use this before a production release and again after the release smoke test. Each item requires evidence, not a verbal assertion.

Area	Gate	Evidence
Backup and restore	PostgreSQL backup dry-run and server_data backup dry-run are reviewed.	pnpm backup:postgres -- --dry-run --env-file .env.production and pnpm backup:server-data -- --dry-run --env-file .env.production output saved in release notes.
Backup and restore	Restore command refuses overwrite without explicit confirmation.	pnpm restore:postgres -- --env-file .env.production --input backups/postgres/ftb_pm-postgres-20260708T120000Z.dump exits non-zero and names --confirm-overwrite.
Backup and restore	Fresh DB restore rehearsal completed before destructive restore.	Temporary database restore command and row-count comparison from docs/runbooks/migration-rollback.md.
Smoke tests	Release smoke is wired into GitHub Actions.	.github/workflows/deploy-production.yml runs scripts/smoke-test-release.mjs with --expected-version.
Smoke tests	Manual smoke can be run against the target.	pnpm deploy:smoke -- --base-url http://localhost --expected-version "$APP_VERSION" output.
Monitoring	Monitoring profile renders and can start without committed secrets.	docker compose --env-file .env.production -f docker-compose.prod.yml --profile monitoring config exits 0.
Monitoring	Required alert rules exist.	FtbPostgresDown, FtbDiskPressure, FtbSlowApiLogBurst, FtbSlowPrismaLogBurst, FtbJobFailureLogBurst, and FtbXiaobaoSummaryStale visible in Prometheus.
Audit	Domain write APIs record actor and resource scope where implemented.	API request sample or audit log sample for Product, Requirement, Project, Version, execution entities, and dictionary writes.
RBAC	Project and version operations enforce Owner, Admin, Member, Viewer boundaries where enabled.	Permission matrix test result or manual account walkthrough attached to release notes.
Consistency	AppData and relation tables are checked for migrated domains.	Count and missing-partition-key queries from docs/runbooks/appdata-retirement.md.
Performance	Slow API and slow Prisma thresholds are configured.	API_SLOW_REQUEST_MS and PRISMA_SLOW_QUERY_MS values recorded, plus Grafana slow-log panels checked.
Performance	V2.2 hot reads avoid full-table scans.	Requirement pool smoke uses productId; query plan or service test evidence attached for high-volume domains.
Post-release	Runtime version matches the release SHA.	/api/v1/health/version payload or pnpm deploy:check-runtime output.
Post-release	Product, requirement, workspace, Xiaobao, and AI config read paths respond.	pnpm deploy:smoke output attached.
Post-release	On-call rollback path is known.	docs/runbooks/migration-rollback.md link included in release notes.

Release owner signs off only after all required evidence is attached to the release notes or incident record.