We are using Docker instead of building AMIs for each deploy because the building of a new AMI can be quite slow. Also, the ability to include the Dockerfile inside the projects provides a simple, clean way to declare how the application should be built.
Each new deploy goes onto a fresh EC2 instance. We never deploy to the same instance twice. If you want to refresh config variables, you can restart the instance and the new config will be picked up.
In the demo, we showed Github's Releases features as the way to serve code to our instances. This could very well be a build service like docker.io. The thing to note here is that we set a RELEASE_URL on our EC2 instance to indicate to our setup script the location of the code and Dockerfile.
We rely on the AWS console to manage our deploys. There are many libraries for popular languages that will help build scripts for common deployment tasks. It should be quite simple to build a campfire of hipchat bot to automate the deployment task.
Bookmarks on this page are managed by an admin user. 