arazzo: 1.0.1
info:
  title: Databricks Provision Cluster Then Create Job On It
  summary: Create a cluster, wait until RUNNING, then create a job bound to that cluster.
  description: >-
    Stands up dedicated compute and a job to use it in one flow by creating a
    Databricks cluster, polling until it reaches the RUNNING state, and then
    creating a job whose task targets that cluster via existing_cluster_id. The
    cluster_id produced by the create call flows into the job's task definition.
    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: databricksApi
  url: ../openapi/databricks-openapi.yml
  type: openapi
workflows:
- workflowId: provision-cluster-and-create-job
  summary: Create a cluster, wait for RUNNING, then create a job on it.
  description: >-
    Provisions a cluster, polls until it is RUNNING, then creates a notebook job
    whose task runs on the new cluster.
  inputs:
    type: object
    required:
    - cluster_name
    - spark_version
    - node_type_id
    - num_workers
    - job_name
    - task_key
    - notebook_path
    properties:
      cluster_name:
        type: string
        description: Name for the new cluster.
      spark_version:
        type: string
        description: The Spark runtime version for the cluster.
      node_type_id:
        type: string
        description: The node type for the cluster.
      num_workers:
        type: integer
        description: The number of worker nodes.
      job_name:
        type: string
        description: The name for the new job.
      task_key:
        type: string
        description: The unique task key within the job.
      notebook_path:
        type: string
        description: The workspace path of the notebook the task runs.
  steps:
  - stepId: createCluster
    description: >-
      Create the cluster that the job task will run on. Returns the cluster_id
      reused by the job task.
    operationId: createCluster
    requestBody:
      contentType: application/json
      payload:
        cluster_name: $inputs.cluster_name
        spark_version: $inputs.spark_version
        node_type_id: $inputs.node_type_id
        num_workers: $inputs.num_workers
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      clusterId: $response.body#/cluster_id
  - stepId: pollClusterState
    description: >-
      Read the cluster status and inspect the life cycle state. Loop back while
      PENDING; continue once it is RUNNING.
    operationId: getCluster
    parameters:
    - name: cluster_id
      in: query
      value: $steps.createCluster.outputs.clusterId
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      state: $response.body#/state
    onSuccess:
    - name: stillPending
      type: goto
      stepId: pollClusterState
      criteria:
      - context: $response.body
        condition: $.state == "PENDING"
        type: jsonpath
    - name: running
      type: goto
      stepId: createJob
      criteria:
      - context: $response.body
        condition: $.state == "RUNNING"
        type: jsonpath
  - stepId: createJob
    description: >-
      Create a job whose single notebook task runs on the newly provisioned
      cluster via existing_cluster_id.
    operationId: createJob
    requestBody:
      contentType: application/json
      payload:
        name: $inputs.job_name
        tasks:
        - task_key: $inputs.task_key
          existing_cluster_id: $steps.createCluster.outputs.clusterId
          notebook_task:
            notebook_path: $inputs.notebook_path
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      jobId: $response.body#/job_id
  outputs:
    clusterId: $steps.createCluster.outputs.clusterId
    jobId: $steps.createJob.outputs.jobId