If you need any help, you are free to ping me with any queries at the email in my profile. My authority on the subject: having written a book [1] and a Udemy course [2]
[1]: https://www.packtpub.com/virtualization-and-cloud/orchestrat... [2]: https://www.udemy.com/mastering-docker/
I recently migrated a whole microservice stack of a half dozen services to OpenID connect and Kubernetes in two weeks. This is with about a year of casual familiarity and playing with Kubernetes, and the same migration to OpenID connect would have easily taken me 5 or 6 weeks to do in Amazon ECS, which is what we currently use in production.
Not to mention I can run a cluster on my three computers at home at no extra charge beyond electricity and play around for free. (See, dear, I'm not a hoarder!)
Setting up a cluster is even simpler now with tools like kops and kubeadm. Or just get one provisioned for you by Google or Red Hat with GKE or OpenShift.
I would highly recommend at the very least making it one of the solutions you try.
Yes, at least for development, setting up the entire environment must be a one-command execution. Every new developer to a team will obviously need to progressively learn the entire stack, but they should be up and running after a single VCS checkout and installation command.
I expect staging and production to be a perfect replica of the same environment developers use. As to how realistic it is to "launch the entire production cluster with a single command", with remote server provisioning, IP allocations, multiple hosts for various load-balanced pools, etc... I'll have to see when that time comes.
Terraform is designed for something close to one-command execution. You're going to have to swap in variables (e.g. AWS access keys for a particular account, domain names, IP addresses, etc.), but Terraform is designed for that. I would advise looking into Terraform Modules[0], which encapsulate this kind of work nicely.
[0]: https://www.hashicorp.com/blog/terraform-custom-providers.ht...
Beside that I want to get around HashiCorp tools, especially Vault for storing and Consul for service discovery.
The majority of information out there, including tutorials and blog articles about others' successful deployments, comes in the form of very high-level overviews. Everything I've found is an introduction to getting a basic docker instance running. There is very little useful information out there as to how to run a proper multi-host cluster.
There is core Docker. Tack on docker-machine, docker-compose, Swarm, and the dozens of 3rd-party cluster management abstractions such as Rancher - and the intensity of the headache never stops growing.
It sounds wonderful, but there is so much to learn to be able to tackle a full production stack. It's one thing to successfully launch a working cluster after hours of manual tinkering. It's a separate beast altogether to fully automate setting up a new cluster by issuing a single command, taking into account consistent configuration of: secure networking, persistent volumes with backups, deployment of container configuration and VCS codebases (ex: nginx vhosts and your code itself), etc.
My goal is to set up an entire project in such a way that there is a single suite of automation that can deploy all environments: development VM, staging, and production.