This reminds me of Kubernetes, where all cluster state is neatly structured and placed in a replicated data store (etcd) that is the source of truth for operation, with the right parts immutable (e.g. volumes).

The first thing people do is check in textual representations of those things in version control and operate on that instead.