Deployment with Docker

HashiCorp Packer

As we previously implied, our CM scripts are really not that optimal if we have to run them every time on a new machine that we add to the cluster or the cloud infrastructure in general. While we can do that, we really shouldn't, since in a perfect world the cluster nodes should be a flexible group that can spawn and kill instances depending on the usage with minimal user intervention so requiring a manual setup of each new machine is simply untenable even at the smallest cluster scales. With AMI image creation we can pre-bake a templated base system image with Ansible a single time, when the image is being made. By doing that, we can launch any new machine with this same image and our interaction with a running system would ...