Replication

Leader–follower

One primary takes writes; replicas copy its log and serve reads. Scales reads, not writes.

Replication lag

Async replicas are behind by ms–seconds. A user may not see their own write → read-your-writes from primary for that user.

Sync vs async

Sync: no data loss on failover but slower writes. Async: fast, but recent writes can be lost.

Failover

Promote a replica when the primary dies. Risk: split brain (two primaries) — needs fencing/consensus.

Multi-leader / leaderless

Writes in several places (multi-region, Dynamo-style). Must resolve conflicts.