Given how much effort you have to go through with etcd and discovery, why not use an MQ like RabbitMQ, that will take care of replication, etc as well?
First of all, because one of their explicit choices was to use HTTP. Using a persistent message bus like RabbitMQ has great advantages, but in terms of failure modes acts very differently to HTTP. For example, an upstream service failure would result in a timeout on getting a response (or ideally, a different instance of that service would respond to your message instead), rather than a more traditional TCP socket er…
Thanks for the explanation. I was looking for any dealbreakers with the MQ approach, but the concern you mentioned is valid. About pushing discovery down a layer, sure, but you'd do it the same way you'd query etcd, as you say.