# Multi-Region Architecture: Active-Passive vs Active-Active "What happens if an entire region goes down?" is a classic senior-level follow-up. Here is a compact way to answer it. ## Why go multi-region - Survive a full region outage (power, network, a bad config push). - Lower latency by serving users from a nearby region. - Data residency rules that require data to stay in a country. ## Active-passive One region serves all traffic; a standby region receives replicated data and takes over on failure. - Simple: all writes go to one place, so no write conflicts. - Replication is usually asynchronous, so failover can lose the last few seconds of writes (your RPO). - Failover takes time: detect the outage, promote the replica database, shift DNS or global load-balancer traffic (your RTO). Practice it regularly, or it won't work when needed. ## Active-active Every region serves reads and writes. - Best latency and no idle capacity. - Concurrent writes to the same data in two regions can conflict. Options: - **Partition ownership**: each user or tenant has a home region that owns their writes; other regions forward writes there. - **Conflict-free data types** or last-writer-wins for data where that is acceptable. - **Globally consistent databases** (Spanner, CockroachDB) that use consensus across regions, paying cross-region latency on writes. ## Supporting pieces - Global traffic routing: GeoDNS or anycast load balancers with health checks. - Stateless services deployed identically everywhere; config and secrets replicated. - Caches per region, warmed before failover. - Avoid global single points of failure such as one auth service or one control plane. ## Interview answer shape State the RPO and RTO you're targeting, pick active-passive unless latency or uptime goals demand more, and explain how you'd handle write conflicts if you choose active-active. Disclosure: I'm with Linkjob AI (https://www.linkjob.ai/), an AI assistant for job interviews.