Thanks so much, Shuoy.

We have some simple Ansible playbooks that we use to automate updating the Task Definitions for our services. The Ansible YAML is an almost direct translation of the AWS ECS TaskDefinition object.

Our "deploy" playbook first updates the task definition, and then it updates the service defintion to use the new task. That triggers the rolling update in ECS. It is really dead simple. We did it to fit in with the rest of our cluster automation workflow.

I would also note that our service.yaml files in Ansible also manage the service's ELBs and RDS Postgres instances as well.

Anything else I can answer, let me know.

If you're reusing existing infrastructure, I would say start with that. I'd also suggest that you use automation to deploy Kubernetes that someone else built. With vSphere, you have some options if Kubernetes is a fit for your organization. I think there's a Big Data Extension for vSphere that gives you a Kubernetes cluster. Just learning how to build stateless microservices and deploy them to Kubernetes is a lot to learn. If you're simultaneously learning how to build and scale Kubernetes clusters, that's difficult in its own right.

Similarly, starting up with a new cloud provider can be challenging. The nice thing about Kubernetes is that if you find that VMWare doesn't make it particularly easy to manage Kubernetes or if automating the platform yourself is too costly, you certainly have the option of migrating to a hybrid cloud approach later. I recommend looking at Google Cloud as well as AWS if you're dead set on Kubernetes (as opposed to considering other container orchestration platforms besides Kubernetes, e.g. AWS ECS).

Hope that helps.

Also, if you choose to try out Kubernetes on AWS, you can experiment with something like this from CoreOS:

https://coreos.com/kubernetes/docs/latest/kubernetes-on-aws.html