HashiCorp Cloud Platform (HCP) Packer's  new public beta puts the long-standing machine-image building tool in the cloud, and also delivers new features such as release channels and a deeper integration with Terraform.

Packer - a tool for building automated machine images which was the first product HashiCorp launched back in 2013 -  has been relaunched onto HashiCorp Cloud Platform, with a host of improvements. These are focused on storing and centrally managing image metadata to provide a more unified approach to building and deploying images with Packer and Terraform.

Organisations have largely been left on their own devices to ensure that the output of a Packer build is used appropriately when configuring infrastructure-as-code-tooling, such as in a Terraform plan - and the link between build and deployment is often hand-crafted in organisations using these tools. This can lead to confusion around when image IDs should be updated in Terraform plans, in turn leading to uncertainty and inconsistencies in which of the resultant built images are actually deployed. HCP Packer removes this ambiguity by classifying Packer builds into channels (eg. a release channel), and removing the manual overhead of matching up the correct Packer build into the correct Terraform deployment.

HCP Packer also improves continuous integration of image builds. API hooks in HCP Cloud allow organisations to, for example, automatically promote images to production channels based on the results of a test acceptance suite, thus removing more potentially manual steps from the image build process. HCP Packer also guides teams towards more automation, by requiring that building code is committed to a Git repository before building it. This takes engineers down a more automated route to building images, reducing human error and the common mistake of accidentally pushing untested images to existent channels.

Even without using this new deeper integration with Terraform, HCP Packer lets engineers track iterations of images, so they can automatically get the best published image on a channel when building downstream images. Previously, engineers would need to update image IDs manually to benefit from new releases. Chaining image builds together in this way allows for a friction-free flow of updated images, which an organisation can customise according to their local needs, for example to add required packages to run their software, or monitoring agents.

This means a team preparing images can label them according to quality and stability, allowing teams to choose which track to take and be sure that even when images are automatically updated, they are still going to work.

HCP Packer also improves the abstraction of different regions or clouds - giving for example, a consistent method of authoring images on multiple cloud providers (for example AWS, GCP, Azure and VMware) and across different regions.

HashiCorp's strategy around investing in golden-image technology reflects the continued use of virtual machine images in organisations. Whilst many organisations are now using containerisation and taking advantage of flexible image-build tagging to determine which image to use in downstream builds and deployments, it’s been difficult to replicate this functionality at a machine-image level. Adding this type of functionality to Packer allows teams to gain similar flexibility and assurance around machine image builds. Indeed, HCP Packer will also soon be able to build Docker images from the same model.

HCP Packer is now in public beta. HashiCorp have published a video showing many of the new benefits, and are inviting signups now.