Database replication maintains copies of data across multiple servers. Synchronous replication guarantees consistency but impacts write latency. Asynchronous replication improves performance but risks data loss during failures. Understanding these tradeoffs guides architecture decisions matching availability requirements.
Replication Topologies
Primary-replica topology directs writes to a single primary with replicas handling reads. Multi-primary enables writes to multiple nodes with conflict resolution. Chain replication reduces primary load by cascading changes. Choose topologies matching write patterns and consistency requirements.
- Primary-replica suits most applications with clear write/read separation
- Use synchronous replication for zero-data-loss requirements accepting latency impact
- Async replication improves write performance with acceptable lag
- Implement connection poolers routing reads to replicas transparently
- Test failover procedures regularly to ensure smooth recovery
Failover Strategies
Automatic failover promotes replicas when primaries fail. Manual failover provides control but delays recovery. Proxy layers like PgBouncer or ProxySQL route connections to healthy nodes. Monitor replication lag to ensure replicas are suitable failover candidates.