Automate VM Deployment

You are building a cloud-native system that will be deployed across several cloud datacenters, probably across more than one cloud vendor, or across public and private clouds.

How can you effectively manage deploying the VM’s that a container or PaaS solution (such as Cloud Foundry, or Kubernetes) run on in a multi-vendor, multi-cloud environment?

Managing base images for multiple cloud vendors by hand is error prone as the various base images can diverge. When relying on a single cloud vendor, maintaining a base image (in AWS speak, an “AMI”) is still workable; with multiple cloud vendors and the networking differences they often bring, this becomes impossible even at a small scale.

Therefore,

Use an automation tool like Chef or Puppet to configure machines; in fact, you should prefer to use cloud vendor-independent orchestration tools such as TerraForm to help up bringing them up.

Automation is not unique to this set of patterns, but it becomes extra important when using multiple vendors. Instead of relying on vendor-dependent tooling to manage and protect networks of machines, machines need to be interconnected with an  Overlay Network, machines in different vendors’ datacenters get slightly different settings because networks and available memory/cpu sizes differ, and so on.

Even with a handful of machines, this quickly becomes very error prone when doing manually. Whereas many good DevOps practices can be postponed by teams for a while, when applying the patterns in this language, VM Automation becomes critical very quickly.