# Design Room-Level Hotel Reservations Without Double Booking Design a hotel reservation service that models individual room numbers rather than abstract room types. Users search for rooms available across a requested date range, place a short-lived hold, and confirm or cancel a booking. The design must prevent two users from confirming the same room for overlapping nights. Search results should help users avoid repeatedly selecting a room already being held, while the final booking path must remain correct even if any cache or real-time availability indicator is stale. ### Constraints & Assumptions - Treat stays as half-open date ranges: check-in is included and check-out is excluded. - A hold expires after a configurable duration if it is not confirmed. - The same request may be retried. - Search may be eventually consistent; confirmation may not double-book. ### Clarifying Questions to Ask - Can a booking change rooms during one stay? - How fresh must search and hold indicators be? - Are payment authorization and reservation confirmation one atomic business operation? - What cancellation, extension, and overbooking policies apply? ### What a Strong Answer Covers - Room-level entities and interval or room-night availability modeling - A durable, atomic contention point for holds and confirmations - Hold expiration, retries, idempotency, and payment failure handling - Fast search and real-time hints that do not become correctness authorities - Date boundaries, time zones, observability, and reconciliation ### Follow-up Questions - What happens when two users hold the same room simultaneously? - How do expired holds get reclaimed after a worker outage? - How would you search a long date range efficiently? - Can a cache ever safely decide that a room is bookable?