Database Engineering: Replication & High Availability

Replication keeps databases available when servers fail. Single-leader, multi-leader, and leaderless each trade consistency differently — this post explains all three clearly.

Replication lag creates bugs that only appear under load. Read-your-own-writes, monotonic reads, and consistent prefix reads are the guarantees that prevent them.

Automatic failover sounds simple — consensus makes it hard. This post explains leader election, Raft vs Paxos trade-offs, and what high availability means in distributed databases.