Azure Databricks Create a Job and Run It to Completion

arazzo: 1.0.1
info:
  title: Azure Databricks Create a Job and Run It to Completion
  summary: Create a notebook job, trigger a run, and poll until TERMINATED.
  description: >-
    Defines a new notebook job and executes it once end to end. The workflow
    creates the job from a single notebook task running on an existing cluster,
    triggers a run, then polls the run get endpoint until the run life cycle
    state is TERMINATED, branching on the result state to a success or failure
    end. Every step spells out its request inline so the flow can be read and
    executed without opening the underlying OpenAPI description.
  version: 1.0.0
sourceDescriptions:
- name: jobsApi
  url: ../openapi/microsoft-azure-databricks-jobs-api-openapi.yml
  type: openapi
workflows:
- workflowId: create-job-and-run
  summary: Create a notebook job, run it now, and wait for the run to finish.
  description: >-
    Creates a job with a single notebook task, triggers runJobNow, then polls
    getJobRun until life_cycle_state is TERMINATED, branching on result_state.
  inputs:
    type: object
    required:
    - token
    - jobName
    - taskKey
    - notebookPath
    - existingClusterId
    properties:
      token:
        type: string
        description: Databricks personal access token for the Authorization header.
      jobName:
        type: string
        description: Name for the new job.
      taskKey:
        type: string
        description: Unique key identifying the single task within the job.
      notebookPath:
        type: string
        description: Absolute workspace path of the notebook to run.
      existingClusterId:
        type: string
        description: Id of an existing all-purpose cluster to run the task on.
  steps:
  - stepId: createJob
    description: >-
      Create a job containing a single notebook task bound to an existing
      cluster. The canonical job_id is returned.
    operationId: createJob
    parameters:
    - name: Authorization
      in: header
      value: Bearer $inputs.token
    requestBody:
      contentType: application/json
      payload:
        name: $inputs.jobName
        tasks:
        - task_key: $inputs.taskKey
          existing_cluster_id: $inputs.existingClusterId
          notebook_task:
            notebook_path: $inputs.notebookPath
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      jobId: $response.body#/job_id
  - stepId: runJobNow
    description: >-
      Trigger an immediate run of the newly created job and capture the run_id
      for polling.
    operationId: runJobNow
    parameters:
    - name: Authorization
      in: header
      value: Bearer $inputs.token
    requestBody:
      contentType: application/json
      payload:
        job_id: $steps.createJob.outputs.jobId
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      runId: $response.body#/run_id
  - stepId: pollRun
    description: >-
      Retrieve the run state. Repeat until the run life_cycle_state is
      TERMINATED, then branch on the result_state.
    operationId: getJobRun
    parameters:
    - name: Authorization
      in: header
      value: Bearer $inputs.token
    - name: run_id
      in: query
      value: $steps.runJobNow.outputs.runId
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      lifeCycleState: $response.body#/state/life_cycle_state
      resultState: $response.body#/state/result_state
    onSuccess:
    - name: succeeded
      type: goto
      stepId: fetchOutput
      criteria:
      - context: $response.body
        condition: $.state.life_cycle_state == "TERMINATED" && $.state.result_state == "SUCCESS"
        type: jsonpath
    - name: finishedNotSuccess
      type: end
      criteria:
      - context: $response.body
        condition: $.state.life_cycle_state == "TERMINATED" || $.state.life_cycle_state == "INTERNAL_ERROR"
        type: jsonpath
    - name: stillRunning
      type: goto
      stepId: pollRun
      criteria:
      - context: $response.body
        condition: $.state.life_cycle_state == "PENDING" || $.state.life_cycle_state == "RUNNING" || $.state.life_cycle_state == "TERMINATING"
        type: jsonpath
  - stepId: fetchOutput
    description: >-
      Retrieve the notebook output for the successful run, capturing the value
      passed to dbutils.notebook.exit().
    operationId: getJobRunOutput
    parameters:
    - name: Authorization
      in: header
      value: Bearer $inputs.token
    - name: run_id
      in: query
      value: $steps.runJobNow.outputs.runId
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      notebookResult: $response.body#/notebook_output/result
  outputs:
    jobId: $steps.createJob.outputs.jobId
    runId: $steps.runJobNow.outputs.runId
    resultState: $steps.pollRun.outputs.resultState
    notebookResult: $steps.fetchOutput.outputs.notebookResult