[tracker1]

Well, for example kubernetes/coreos create their own complexities, of which docker is only part (you can even roll your own, but this takes time/effort), that doesn't even handle CI/CD parts, or manual intervention steps. There's operations overhead for these solutions.

If you are a mid/large sized organization it becomes easier to take different approaches... if your single ops guy on a 5-man dev+ops team is overwelmed, and your other main ops person is tied up with coding/architecture then even flushing out such a system is difficult. We're using wildcard DNS to make routing from nginx more straight forward, and deploying each service to 3 dokku-alt servers... which is easier than some of the dynamic provisioning and other pieces behind say coreos/kubernetes.

For that matter, we're keeping our semi-microservices in check, only having a few of them for user-accessed data, and a handful of others for handling work queues.