Recommended Configurations

Replica sets support quite a few options, and it can be confusing determining the best configuration. Here following a few suggestions.

One data center

If you have just one data center, then the most economical setup is a three-node replica set, one of which is designated as an arbiter. The standard nodes each get their own box, and the arbiter lives on an application server.

Two data centers

With two data centers, you'll want to designate one data center as primary and the other as a backup for the very rare case where the first data center fails entirely.

From here, a three-node replica set is sufficient. You'll keep two replica set nodes in the primary data center and one node in the backup data center. You'll also want to give the node in the backup DC a priority of 0. As long as the first data center continues to function, your primary node will always live there. If you lose any one node, the replica set will still be able to elect a primary in the main data center.

With this relatively economical setup, you're protected against the failure of any one node and against the failure of any one data center. Of course, if it's the primary data center that fails, then you'll need to manually promote the node in the secondary data center to primary. But if you use write concern carefully, then you won't lose any data, the manual failover won't lead to much downtime.

See Also

• Data Center Awareness