At Stark & Wayne we are constantly thinking, planning, exploring and working towards the ultimate continuous deployment pipelines for our production systems & our customer's production systems.
If you are looking for a new CI system that can help you deterministically pipeline running deployments and artifacts then I think you need to take 10 minutes to watch the demo video below for a system called Concourse:
In the video, we are using a Concourse pipeline to deploy a Redis cluster three times. The goal is to only do a deployment/upgrade/reconfiguration of the production Redis cluster if it the "deployment change" has been successfully applied already.
In the demo I have a "try anything" stage where I deploy a cluster of Redis using the latest assets/configuration coming in from upstream/other dev teams.
If that deployment and acceptance tests are ok, then I move to "pre-production". Here, the goal is to re-create "production" and perform an upgrade to the successful "try anything" configuration/packages.
If "pre-production" is successful (and conceptually it can be destroyed at this time), then we have a candidate for the next "production" deployment.
In the demo this production deployment is triggered automatically; though in many organizations a human might want to push the button. Humans need to do something right?
This demo is also a demonstration of BOSH deployments; though hopefully this is orthoganal to the demonstration of Concourse. Concourse was written by people who use BOSH so it currently comes with extensions (called "resources") that work with BOSH.
If you're interested in using Concourse to automate your own assets and production deployments then contact us to help. We think Concourse is a very exciting next-generation open source continously deployment solution.