When starting a new instance of MusicBrainz Server, it takes some time before it gets ready to handle requests that are set to it. If requests are sent too early, these will be denied with the error message “Couldn't connect to the renderer.” This situation actually occurs every time we are replacing a container in production for a server update.

A solution is to provide a basic health check that deployment scripts could use to avoid directing the traffic towards a new instance until it gets ready for real.