Replication Overview

To ensure data durability and availability, Oracle NoSQL Database uses a single-master replication strategy. Using a single machine to perform write operations, Oracle NoSQL Database then broadcasts those operations to multiple read-only replicas.

The Concepts Guide describes a shard as a collection of replication nodes, associated with a single master node and multiple replicas. Your store contains multiple shards, and your data is spread evenly across all of the shards that your store uses.

When you perform a write operation in your store, Oracle NoSQL Database completes the write operation on the master node in use by the shard containing your data. The master node performs this write according to whatever durability guarantees are in place at the time. If you set a strong durability guarantee, the master requires the participation of some or all of the replicas in the shard to complete the write operation.

If the master node of the shard becomes unavailable for any reason, the replica nodes in primary zones hold an election to determine which of the remaining replication nodes should take over as the master node. The replication node with the most up-to-date data wins the election.

The election is decided based on a simple majority vote. This means that a majority of the nodes in the shard in primary zones must be available to participate in the election to select a new master.