Orchestrator Plugin for Backstage

The Orchestrator for Backstage is a mechanism designed to facilitate the implementation and execution of developer self-service flows. It serves as a vital component that enhances and augments the existing scaffolder functionality of Backstage with a more flexible and powerful set of features including long-running and asynchronous flows.

The orchestrator works harmoniously with other Backstage components such as the Software Catalog, permissions, and plugins as well as others. By leveraging its capabilities, organizations can orchestrate and coordinate developer self-service flows effectively.

Context

The Backstage Orchestrator plugin aims to provide a better option to Scaffolder, based on workflows to have a more flexible and powerful tool that addresses the need by streamlining and automating processes, allowing developers to focus more on coding and innovation.

The orchestrator relies on SonataFlow, a powerful tool for building cloud-native workflow applications.

The main idea is to keep the same user experience for users, leveraging the UI components, input forms, and flow that Scaffolder provides, this way it should be straightforward for users and transparent no matter whether using Templates or Workflows, both can live together being compatible with integration points.

The orchestrator controls the flow orchestrating operations/tasks that may be executed in any external service including Scaffolder Actions, this way it is possible to leverage any existing Action hence Software Templates can be easily migrated to workflows opening the door to extend them to more complex use cases.

Capabilities

Advanced core capabilities

• Stateful/long-lived
• Branching and parallelism
• Error management and compensation
• Event-driven supporting CloudEvents
• Audit logging
• Sub-flows
• Choreography
• Timer/timeout control
• Built-in powerful expression evaluation with JQ
• Low Code/No code
• Cloud-native architecture Kubernetes/OpenShift with Operator support
• OpenAPI / REST built-in integration etc.

Client-side tooling

• Orchestration visualization / graphical editor
• Integration with service catalog/actions
• GitHub integration
• Form generation
• Runtime monitoring of instances
• Dashboards
• Potential custom integrations (user interaction, notifications, etc.)

For administrators

Installation

The Orchestrator plugin is composed of the following packages:

• @janus-idp/backstage-plugin-orchestrator-backend package connects the Backstage server to the Orchestrator. For setup process, see Backend Setup
• @janus-idp/backstage-plugin-orchestrator package contains frontend components for the Orchestrator plugin. For setup process, see Frontend Setup
• @janus-idp/backstage-plugin-orchestrator-common package contains shared code between the Orchestrator plugin packages.

Prerequisites for running the plugins locally in development mode

• Docker up and running

Setting up the Orchestrator as a dynamic plugin in a Helm deployment

Please follow this link for instructions: https://github.com/janus-idp/backstage-showcase/blob/main/showcase-docs/dynamic-plugins.md#helm-deployment

Setting up the configuration for the Orchestrator plugin

The following configuration is required for the Orchestrator plugin to work properly:

backend:
  csp:
    frame-ancestors: ['http://localhost:3000', 'http://localhost:7007']
    script-src: ["'self'", "'unsafe-inline'", "'unsafe-eval'"]
    script-src-elem: ["'self'", "'unsafe-inline'", "'unsafe-eval'"]
    connect-src: ["'self'", 'http:', 'https:', 'data:']
orchestrator:
  sonataFlowService:
    baseUrl: http://localhost
    port: 8899
    autoStart: true
    workflowsSource:
      gitRepositoryUrl: https://github.com/parodos-dev/backstage-orchestrator-workflows
      localPath: /tmp/orchestrator/repository
  dataIndexService:
    url: http://localhost:8899

• When interacting with an existing SonataFlow infrastructure, the sonataFlowService config section must be entirely omitted and the dataIndexService.url must point to the existing Data Index Service.

For more information about the configuration options, including other optional properties, see the config.d.ts file.

Setting up the Orchestrator backend package

1. Install the Orchestrator backend plugin using the following command:

   yarn workspace backend add @janus-idp/backstage-plugin-orchestrator-backend

2. Add the following code to the packages/backend/src/index.ts file:

   const backend = createBackend();
   
   /* highlight-add-next-line */
   backend.add(import('@janus-idp/backstage-plugin-orchestrator-backend'));
   
   backend.start();

Setting up the Orchestrator frontend package

1. Install the Orchestrator frontend plugin using the following command:

   yarn workspace app add @janus-idp/backstage-plugin-orchestrator

2. Add a route to the OrchestratorPage and the customized template card component to Backstage App (packages/app/src/App.tsx):

   /* highlight-add-next-line */
   import { OrchestratorPage } from '@janus-idp/backstage-plugin-orchestrator';
   
   const routes = (
     <FlatRoutes>
       {/* ... */}
       {/* highlight-add-next-line */}
       <Route path="/orchestrator" element={<OrchestratorPage />} />
     </FlatRoutes>
   );

3. Add the Orchestrator to Backstage sidebar (packages/app/src/components/Root/Root.tsx):

   /* highlight-add-next-line */
   import { OrchestratorIcon } from '@janus-idp/backstage-plugin-orchestrator';
   
   export const Root = ({ children }: PropsWithChildren<{}>) => (
     <SidebarPage>
       <Sidebar>
         <SidebarGroup label="Menu" icon={<MenuIcon />}>
           {/* ... */}
           {/* highlight-add-start */}
           <SidebarItem
             icon={OrchestratorIcon}
             to="orchestrator"
             text="Orchestrator"
           />
           {/* highlight-add-end */}
         </SidebarGroup>
         {/* ... */}
       </Sidebar>
       {children}
     </SidebarPage>
   );

For users

Using the Orchestrator plugin in Backstage

The Orchestrator plugin enhances the Backstage with the execution of developer self-service flows. It provides a graphical editor to visualize workflow definitions, and a dashboard to monitor the execution of the workflows.

Refer to the Quick start to install the Orchestrator using the helm chart and execute a sample workflow through the Red Hat Developer Hub orchestrator plugin UI.

OpenAPI

The plugin provides OpenAPI v2 endpoints definition to facilitate communication between the frontend and backend. This approach minimizes the data that needs to be sent to the frontend, provides flexibility and avoids dependencies on changes in the CNCF serverless specification. It also allows for a seamless transition if there's a need to replace the backend implementation.

In addition, by leveraging on OpenAPI spec, it is possible to generate clients and create CI steps.

OpenAPI specification file is available in orchestrator-common.
OpenAPI specification documentation is available here

NOTE:
While the OpenAPI specification is available in the Orchestrator plugin, the UI currently does not rely on this spec.
We plan to incorporate v2 endpoints into the UI in the near future.

orchestrator-common

The typescript schema is generated in auto-generated folder from openapi.yaml specification file.

orchestrator-backend

The orchestrator backend can use the generated schema to validate the HTTP requests and responses.

NOTE: Temporary the validation has been disabled. It will be enabled when the orchestrator frontend will switch to the use of v2 endpoints only.

Development instruction

Checkout the backstage-plugin

git clone [email protected]:janus-idp/backstage-plugins.git

If you need to change the OpenAPI spec, edit the openapi.yaml according to your needs and then execute from the project root folder:

yarn --cwd plugins/orchestrator-common openapi

This command updates the auto-generated files and the auto-generated docs.

NOTE: Do not manually edit auto-generated files

If you add a new component in the spec, then you need to export the generated typescript object here. For example, if you define

components:
  schemas:
    Person:
      type: object
      properties:
        name:
          type: string
        surname:
          type: string

then

export type Person = components['schemas']['Person'];

When defining a new endpoint, you have to define the operationId. That id is the one that you can use to implement the endpoint logic.

For example, let's assume you add

paths:
  /names:
    get:
      operationId: getNames
      description: Get a list of names
      responses:
        '200':
          description: Success
          content:
            application/json:
              schema:
               type: array
                items:
                  $ref: '#/components/schemas/Person'

Then you can implement the endpoint in router.ts referring the operationId getNames:

api.register('getNames', async (_c, _req, res: express.Response, next) => {
  // YOUR LOGIC HERE
  const result: Person[] = [
    { name: 'John', surname: 'Snow' },
    { name: 'John', surname: 'Black' },
  ];

  res.status(200).json(result);
});