Creating AMI Artifacts with Atlas

In an immutable infrastructure workflow, it's important to version and store full images (artifacts) to be deployed. This section covers storing AWS AMI images in Atlas to be queried and used later.

Note the actual AMI does not get stored in Atlas. Atlas simply keeps the AMI ID as a reference to the target image. Tools like Terraform can then use this in a deploy.


If you run Packer in Atlas, the following will happen after a push:

  1. Atlas will run packer build against your template in our infrastructure. This spins up an AWS instance in your account and provisions it with any specified provisioners
  2. Packer stops the instance and stores the result as an AMI in AWS under your account. This then returns an ID (the artifact) that it passes to the Atlas post-processor
  3. The Atlas post-processor creates and uploads the new artifact version with the ID in Atlas of the type amazon.image for use later


Below is a complete example Packer template that starts an AWS instance.

  "push": {
    "name": "acmeinc/frontend"
  "provisioners": [],
  "builders": [
      "type": "amazon-ebs",
      "access_key": "",
      "secret_key": "",
      "region": "us-east-1",
      "source_ami": "ami-2ccc7a44",
      "instance_type": "c3.large",
      "ssh_username": "ubuntu",
      "ami_name": "Atlas Example {{ timestamp }}"
  "post-processors": [
      "type": "atlas",
      "artifact": "acmeinc/web-server",
      "artifact_type": "amazon.image"