CQRS

Command Query Responsibility Segregation splits write models optimized for business rules from read models optimized for queries — different schemas, stores, or scaling. Updates propagate via events or sync pipelines; temporary lag between write and read is expected.

Trade-off: read performance and tailored views versus synchronization complexity and stale reads.

On interviews: when CQRS is overkill for a CRUD service; design a read model for a search page fed by order events.

Common pitfalls: CQRS everywhere without read pain; no strategy for read model rebuild; dual writes without outbox.

Checklist:

• Justify separate read store with concrete query needs.
• Project read models idempotently from events.
• Plan full rebuild from event log.
• Document staleness for each read endpoint.