Google Cloud Deployment Manager

What do you like best?

A flexible infrastructure-as-code solution for Google Cloud: like Terraform or AWS CloudFormation

Supports most of the popular Google Cloud Platform services and resource types (VMs, Cloud SQL, Storage buckets, Pub/Sub, Cloud Functions etc.)

Simple declarative syntax that claims to support all forms of underlying APIs (as supported by the particular cloud service)

Ability to externalize variable or sensitive parameters as template inputs

Ability to extract outputs from the template for dealing with generated/updated resources

Custom resource support, and fully customized template processing via inline Python and Jinja2 snippets

Accessible via the CORS-friendly REST API and gcloud CLI

What do you dislike?

There is no rollback mechanism! An absolute necessity for such a resource management utility; but there is no mention of rollback in the API or docs.

Lack of proper documentation on end-to-end usage of the API; even for the usual flow (preview, submit, monitor)

Deployment Manager internally converts the template to API calls, hence there is no way to find if a particular operation is supported until it gets invoked at runtime (e.g. unlike CloudFormation which has extensive documentation on supported operations and modes for each resource type)

There are cases where a deployment can end up in an inconsistent state, especially if the previous one has failed.

Recommendations to others considering the product:

There are certain cases where a deployment can go into an inconsistent state, especially after the previous one has somehow failed.

Beware; there are no automatic rollbacks - contrary to what you might have expected!

If you are struggling to find a proper call sequence for invoking the Deployment Manager via its API, check out something like https://dzone.com/articles/deploying-your-stuff-with-google-cloud-deployment

What problems are you solving with the product? What benefits have you realized?

We use Deployment Manager for deploying Google Cloud Platform based serverless projects in our Sigma cloud IDE. It offers a unified user experience (we similarly use CloudFormation for AWS-based Sigma projects), and the deployments are quite fast and fairly reliable.