Skip to main content

ADR005: Catalog Core Entities

Context​

We want to standardize on a few core entities that we are tracking in the Backstage catalog. This allows us to build specific plugins around them.

Decision​

Backstage should eventually support the following core entities:

  • Components are individual pieces of software
  • APIs are the boundaries between different components
  • Resources are physical or virtual infrastructure needed to operate a component

Catalog Core Entities

For now, we'll start by only implementing support for the Component entity in the Backstage catalog. This can later be extended to APIs, Resources and other potentially useful entities.

Component​

A component is a piece of software, for example a mobile application feature, web site, backend service or data pipeline (list not exhaustive). A component can be tracked in source control, or use some existing open source or commercial software. It can implement APIs for other components to consume. In turn it might depend on APIs implemented by other components, or resources that are attached to it at runtime.

Component entities are typically defined in YAML descriptor files next to the code of the component, and could look like this (actual schema will evolve):

apiVersion: backstage.io/v1alpha1
kind: Component
metadata:
  name: my-component-name
spec:
  type: service

API​

APIs form an abstraction that allows large software ecosystems to scale. Thus, APIs are a first class citizen in the Backstage model and the primary way to discover existing functionality in the ecosystem.

APIs are implemented by components and make their boundaries explicit. They might be defined using an RPC IDL (e.g. in Protobuf, GraphQL or similar), a data schema (e.g. in Avro, TFRecord or similar), or as code interfaces (e.g. framework APIs in Swift, Kotlin, Java, C++, TypeScript etc). In any case, APIs exposed by components need to be in a known machine-readable format so we can build further tooling and analysis on top.

APIs are typically indexed from existing definitions in source control and thus wouldn't need their own descriptor files, but would be stored in the catalog somewhat like this (actual schema will evolve):

apiVersion: backstage.io/v1alpha1
kind: API
metadata:
  name: my-component-api
spec:
  type: grpc
  definition: >
    service HelloService {
      rpc SayHello (HelloRequest) returns (HelloResponse);
    }

    message HelloRequest {
      string greeting = 1;
    }

    message HelloResponse {
      string reply = 1;
    }

Resource​

Resources are the infrastructure your software needs to operate at runtime like Bigtable databases, Pub/Sub topics, S3 buckets or CDNs. Modelling them together with components and APIs will allow us to visualize and create tooling around them in Backstage.

Resources are typically indexed from declarative definitions (e.g. Terraform, GCP Config Connector, AWS Cloud Formation) and/or inventories from cloud providers (e.g. GCP Asset Inventory) and thus wouldn't need their own descriptor files, but would be stored in the catalog somewhat like this (actual schema will evolve):

apiVersion: backstage.io/v1alpha1
kind: Resource
metadata:
  name: my-component-db
spec:
  type: gcp-spanner
  url: spanner.googleapis.com/projects/prj/instances/my-component-db/databases/my-db

Consequences​

We will continue fleshing out support for the Component entity in the Backstage catalog.