Elasticache with Python, or how to stop worrying and love HAProxy

You can have haproxy also check that the replica instance is in sync with the master via an xinetd service that checks the redis status. This way as you do maintenance and take nodes out of the balancer and add them back in you don’t end up having a window of stale out of date data