openskills.info
Course catalog
Course Preview

Database Replication Strategies

Database replication keeps copies of database changes on more than one system. A replication strategy decides where writes happen, how changes travel, when a commit is acknowledged, and what readers may see when a replica lags or a node fails.

itDatabases and data storage

Don't Panic: Database Replication Strategies

Database replication copies database changes to other servers or regions. It can improve availability, place reads nearer to users, isolate reporting, support migrations, or feed another system. It is not a backup. A replica can copy your most regrettable delete with admirable speed.

Think in seven stages: authority, order, transport, acknowledge, apply, route, and recover. Authority says who may accept writes. A single-primary design has one writer. Multi-primary and leaderless designs allow more write locations but need conflict prevention, rejection, or repair. Consensus systems use a voting group to keep one ordered history, often by stopping writes in a minority partition.

Order comes from a change log: write-ahead log, binary log, oplog, or replicated command log. Transport moves that history. Apply makes it queryable. These are not the same moment. A replica can receive a change before it durably stores it, and store it before readers can observe it.

The acknowledgment point is the promise. Asynchronous replication returns before a remote replica confirms the change, so remote latency stays away from normal commits but acknowledged writes may be absent after primary loss. Synchronous modes wait for a configured remote point. That can reduce loss and increase waiting, which is the sort of trade that should be named before the incident.

Routing decides what clients observe. Replica reads may be stale. Read-your-writes needs routing back to the writer, session pinning, or waiting for a replica position. Failover needs fencing so the old writer cannot keep accepting changes after promotion.

Use the Practice Reference to write the strategy record and rehearse failover. Do the Exercise to choose authority, acknowledgment, reads, monitoring, and recovery. The Cheatsheet keeps topology, quorums, lag, retention, and failure decisions close. Replication is useful because it is explicit about tradeoffs, not because copies are comforting.

Where this skill leads

Relevant careers

See how this topic contributes to broader role-level skill maps.

Sources