pgreplicaproxy

README

pgreplicaproxy is a load-balancing proxy for PostgreSQL connections, intended to simplify the deployment of one or more read-replica database servers.

This is currently a proof-of-concept application; it is not production ready.

pgreplicaproxy is configured with a list of IP addresses that represent all the PostgreSQL nodes in your database cluster. It monitors all the configured servers to determine whether they are available, and whether they are a master or a replica. It then receives client database connections, and based upon the database name that the client attempts to connect to, pgreplicaproxy will proxy the connection to the online master server or an online read-replica server. If the database name ends in _replica, then a replica connection will be used, and the _replica suffix will be removed.

There are a few major issues that prevent pgreplicaproxy from being generally useful today:

• When a backend server changes state (master -> replica, replica -> master, etc.) it should close the current network connections to that server so that they're not mistakenly running against the wrong server type. (issue #3)

• Lack of testing under high load and concurrency situations.

Installing

1. Install Go 1.1 for your OS (http://golang.org/doc/install).

2. Set your GOPATH environment variable. If you're not familiar with Go, this is the root of where you'd like to store projects and compiled binaries.

3. Execute go get github.com/replicon/pgreplicaproxy.

4. A compiled pgreplicaproxy binary should now be in $GOPATH/bin.

Using

1. Copy example.cfg into pgreplicaproxy.cfg, and edit it as desired.

2. Run pgreplicaproxy in the same directory as pgreplicaproxy.cfg.

3. Connect to the host/port provided on the "listen" line of the config file, with the appropriate username, password, and database name to use one of the backend connections from the .cfg file.

4. Try connecting again, and append _replica to the database name to connect to a replica instead (if one is up and available).