The final important aspect of database sharding that we are going to explore in this chapter is reorganization. The purpose of allocating a large number of logical shards is to prepare for future expansion needs. If we started with 2048 shards, all of which are currently mapped to a single server, we will eventually want to move some of them elsewhere.

The easiest way to do this is to leverage PostgreSQL replication. Essentially, we will create a streaming replica for the server we want to split and drop the schemas we don't need on each server. Consider a database with two shards. Our end goal is to produce something like this:

On each server, we simply drop the schema indicated by the dashed box. This way, we still have two shards, and only the location of myapp2 has changed; its data remains unharmed.

This recipe will cover the process described here, making it easy to move shards to a new physical location.