Application metadata descriptor CRD
Switch branches/tags
Nothing to show
Clone or download
k8s-ci-robot Merge pull request #62 from crimsonfaith91/patch-2
Clarify that Application CRD and object are different
Latest commit 42782ae Aug 23, 2018

Kubernetes Applications

Kubernetes is an open-source system for automating deployment, scaling, and management of containerized applications.

The above description, from the Kubernetes homepage, is centered on containerized applications. Yet, the Kubernetes metadata, objects, and visualizations (e.g., within Dashboard) are focused on container infrastructure rather than the applications themselves.

The Application CRD (Custom Resource Definition) and Controller in this project aim to change that in a way that's interoperable between many supporting tools.

It provides:

  • The ability to describe an applications metadata (e.g., that an application like WordPress is running)
  • A point to connect the infrastructure, such as Deployments, to as a root object. This is useful for tying things together and even cleanup (i.e., garbage collection)
  • Information for supporting applications to help them query and understand the objects supporting an application
  • Application level health checks

This can be used by:

  • Application operators who want to center what they operate on applications
  • Tools, such as Helm, that center their package releases on application installations can do so in a way that's interoperable with other tools (e.g., Dashboard)
  • Dashboards that want to visualize the applications in addition to or instead of an infrastructure view


  1. Provide a standard API for creating, viewing, and managing applications in Kubernetes.
  2. Provide a CLI implementation, via kubectl, that interacts with the Application API.
  3. Provide installation status and garbage collection for applications.
  4. Provide a standard way for applications to surface a basic health check to the UIs.
  5. Provide an explicit mechanism for applications to declare dependencies on another application.
  6. Promote interoperability among ecosystem tools and UIs by creating a standard that tools MAY implement.
  7. Promote the use of common labels and annotations for Kubernetes Applications.


  1. Create a standard that all tools MUST implement.
  2. Provide a way for UIs to surface metrics from an application.

Application Objects

After creating the Application CRD, you can create Application objects. An Application object provides a way for you to aggregate individual Kubernetes components (e.g. Services, Deployments, StatefulSets, Ingresses, CRDs), and manage them as a group. It provides UIs with a resource that allows for the aggregation and display of all the components in the Application.

Field Type Description
spec.descriptor.type string The type of the application (e.g. WordPress, MySQL, Cassandra). You can have many applications of different names in the same namespace. They type field is used to indicate that they are all the same type of application.
spec.componentKinds [] GroupKind This array of GroupKinds is used to indicate the types of resources that the application is composed of. As an example an Application that has a service and a deployment would set this field to [{"group":"","kind": "Service"},{"group":"apps","kind":"StatefulSet"}]
spec.selector LabelSelector The selector is used to match resources that belong to the Application. All of the applications resources should have labels such that they match this selector. Users should use the label on all components of the Application and set the selector to match this label. For instance, {"matchLabels": [{"": "my-cool-app"}]} should be used as the selector for an Application named "my-cool-app", and each component should contain a label that matches.
spec.descriptor.version string A version indicator for the application (e.g. 5.7 for MySQL version 5.7).
spec.descriptor.description string A short, human readable textual description of the Application.
spec.descriptor.icons []ImageSpec A list of icons for an application. Icon information includes the source, size, and mime type.
spec.descriptor.maintainers []ContactData A list of the maintainers of the Application. Each maintainer has a name, email, and URL. This field is meant for the distributors of the Application to indicate their identity and contact information.
spec.descriptor.owners []ContactData A list of the operational owners of the application. This field is meant to be left empty by the distributors of application, and set by the installer to indicate who should be contacted in the event of a planned or unplanned disruption to the Application
spec.descriptor.keywords array string A list of keywords that identify the application. []InfoItem Info contains human readable key-value pairs for the Application.
spec.descriptor.links []Link Links are a list of descriptive URLs intended to be used to surface additional documentation, dashboards, etc.
spec.descriptor.notes string Notes contain human readable snippets intended as a quick start for the users of the Application. They may be plain text or CommonMark markdown.
spec.assemblyPhase string: "Pending", "Succeeded" or "Failed" The installer can set this field to indicate that the application's components are still being deployed ("Pending") or all are deployed already ("Succeeded"). When the application cannot be successfully assembled, the installer can set this field to "Failed".


This project uses the kubebuilder tool for code generation. kubebuilder provides the same code generation features (and a bit more) for Custom Resource Definitions and Extension API Servers that are provided by the Kubernetes project. In order to build the source, you need to download and install the latest release of kubebuilder per the instructions there.


The controller doesn't do much at the moment. However, if you'd like to build it you'll need to install Docker and golang 1.9 or greater. To build the controller into an image named image use the following command.

docker build -t <image> -f Dockerfile.controller .

Installing the CRD

In order to install the CRD you will either need to use kubectl or you will need to call against the Kubernetes CRD API directly. An example manifest is supplied in the hack directory. You can use the following command to install the CRD (where ```manifest`` is the manifest containing the CRD declaration).

kubectl apply -f <manifest>

Generating an Installation Manifest

When the CRD is installed as above, you need to ensure that the correct RBAC configuration is applied prior to installation. You can use kubebuilder create config to generate a manifest that is configured to create the requisite RBAC permissions, CRD, and controller StatefulSet in the supplied namespace. The command below will generate a manifest that can be applied to create all of the necessary components the image as the controller image and namespace as the namespace. Note that, if you would like to remove the controller from the configuration you can delete the generated StatefulSet from the manifest, and, while you must specify a controller image, you need can supply any string if you do not wish to install the controller when the manifest is applied (i.e. you intend to delete the StatefulSet from the generated manifest). Work is in progress to generate a controllerless configuration.

kubebuilder create config --controller-image <image>  --name <namespace>

Using the Application CRD

The application CRD can be used both via manifests and programmatically.


The docs directory contains a manifest that shows how to you can integrate the Application CRD with a WordPress deployment.

The Application object shown below declares that the Application is a WordPress installation that uses StatefulSets and Services. It also contains some other relevant metadata described above.

kind: Application
  name: "wordpress-01"
  labels: "wordpress-01" "3"
    matchLabels: "wordpress-01"
    - group: core
      kind: Service
    - group: apps
      kind: Deployment
    - group: apps
      kind: StatefulSet
  assemblyPhase: "Pending"
    version: "4.9.4"
    description: "WordPress is open source software you can use to create a beautiful website, blog, or app."
      - src: ""
        type: "image/png"
    type: "wordpress"
      - name: Kenneth Owens
      - "Kenneth Owens"
      - "cms"
      - "blog"
      - "wordpress"
      - description: About
        url: ""
      - description: Web Server Dashboard
        url: "https://metrics/internal/wordpress-01/web-app"
      - description: Mysql Dashboard
        url: "https://metrics/internal/wordpress-01/mysql"

Notice that each Service and StatefulSet is labeled such that Application's Selector matches the labels. "wordpress-01"

The additional labels on the Applications components come from the recommended application labels and annotations.

You can use the standard kubectl verbs (e.g. get, apply, create, delete, list, watch) to interact with an Application specified in a manifest.


kubebuilder creates a Kubernetes ClientSet for the Application object. You can create a new client using either a rest.Config or a rest.Interface as below.

client,err := clientset.NewForConfig(config)

client := clientset.New(ri)

Once you've created a client you can interact with Applications via the structs declared in types.go. For instance to retrieve an application you can used the code below.

app, err := client.AppV1Aplha1().Applications("my-namespace").Get("my-app",v1.GetOptions{})
if err != nil {

The other standard client operations are supported. The interface is described here.


Go to the documentation

Community, discussion, contribution, and support

Learn how to engage with the Kubernetes community on the community page.

You can reach the maintainers of this project at:

Code of conduct

Participation in the Kubernetes community is governed by the Kubernetes Code of Conduct.