One of the nicest things about the Postgres-XC architecture is the fact that you can have multiple coordinators and they are all peers of one another. There is no master or slave. What you can do on one coordinator, you can do on all coordinators. The problem comes in when you want to evenly distribute load across all of the coordinators. You have several options on how to do this, but usually the way this is handled is through some sort of intermediary, like HA Proxy, routing connections to the least loaded server.
↧