MongoDB survives region outage via 2+2+1 multi‑cluster Kubernetes pattern
The CNCF guide shows how to run MongoDB across three Kubernetes clusters using a 2+2+1 replica set, so five voting members are split across sites. This layout lets the database elect a new primary even if an entire region fails, enabling zero‑downtime disaster recovery and live migrations. It uses the Percona Operator and open‑source tooling.