Azure Databricks Provision and Wait for Cluster

arazzo: 1.0.1
info:
  title: Azure Databricks Provision and Wait for Cluster
  summary: Create a cluster and poll its state until it reaches RUNNING.
  description: >-
    Provisions a new Apache Spark cluster and waits for it to become ready.
    The workflow creates the cluster from the supplied configuration, captures
    the returned cluster_id, then polls the cluster get endpoint until its
    state transitions out of PENDING into RUNNING. If the cluster reports a
    TERMINATED or ERROR state the flow branches to a 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: clustersApi
  url: ../openapi/microsoft-azure-databricks-clusters-api-openapi.yml
  type: openapi
workflows:
- workflowId: provision-cluster
  summary: Create a Databricks cluster and wait until it is RUNNING.
  description: >-
    Creates a Spark cluster, then polls getCluster until the cluster state is
    RUNNING, branching to a failure end if it becomes TERMINATED.
  inputs:
    type: object
    required:
    - token
    - clusterName
    - sparkVersion
    - nodeTypeId
    - numWorkers
    properties:
      token:
        type: string
        description: Databricks personal access token for the Authorization header.
      clusterName:
        type: string
        description: Human-readable name for the new cluster.
      sparkVersion:
        type: string
        description: Databricks Runtime version key (from listSparkVersions).
      nodeTypeId:
        type: string
        description: Azure VM node type id for the workers and driver.
      numWorkers:
        type: integer
        description: Number of worker nodes for a fixed-size cluster.
      autoterminationMinutes:
        type: integer
        description: Idle minutes before the cluster auto-terminates.
        default: 30
  steps:
  - stepId: createCluster
    description: >-
      Create the cluster from the supplied configuration. The cluster starts
      in a PENDING state and the canonical cluster_id is returned.
    operationId: createCluster
    parameters:
    - name: Authorization
      in: header
      value: Bearer $inputs.token
    requestBody:
      contentType: application/json
      payload:
        cluster_name: $inputs.clusterName
        spark_version: $inputs.sparkVersion
        node_type_id: $inputs.nodeTypeId
        num_workers: $inputs.numWorkers
        autotermination_minutes: $inputs.autoterminationMinutes
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      clusterId: $response.body#/cluster_id
  - stepId: pollCluster
    description: >-
      Retrieve the current cluster state. Repeat this step until the cluster
      reports RUNNING; branch to failure if it reports TERMINATED.
    operationId: getCluster
    parameters:
    - name: Authorization
      in: header
      value: Bearer $inputs.token
    - name: cluster_id
      in: query
      value: $steps.createCluster.outputs.clusterId
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      state: $response.body#/state
      stateMessage: $response.body#/state_message
    onSuccess:
    - name: clusterRunning
      type: goto
      stepId: confirmRunning
      criteria:
      - context: $response.body
        condition: $.state == "RUNNING"
        type: jsonpath
    - name: clusterFailed
      type: goto
      stepId: reportFailure
      criteria:
      - context: $response.body
        condition: $.state == "TERMINATED" || $.state == "ERROR"
        type: jsonpath
    - name: stillPending
      type: goto
      stepId: pollCluster
      criteria:
      - context: $response.body
        condition: $.state == "PENDING" || $.state == "RESTARTING" || $.state == "RESIZING"
        type: jsonpath
  - stepId: confirmRunning
    description: >-
      Re-read the cluster once it is RUNNING to capture its final metadata for
      the workflow outputs.
    operationId: getCluster
    parameters:
    - name: Authorization
      in: header
      value: Bearer $inputs.token
    - name: cluster_id
      in: query
      value: $steps.createCluster.outputs.clusterId
    successCriteria:
    - condition: $statusCode == 200
    - condition: $response.body#/state == "RUNNING"
    outputs:
      clusterId: $response.body#/cluster_id
      state: $response.body#/state
    onSuccess:
    - name: done
      type: end
  - stepId: reportFailure
    description: >-
      Read the cluster one last time to capture the termination reason when
      provisioning did not reach RUNNING.
    operationId: getCluster
    parameters:
    - name: Authorization
      in: header
      value: Bearer $inputs.token
    - name: cluster_id
      in: query
      value: $steps.createCluster.outputs.clusterId
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      state: $response.body#/state
      terminationReason: $response.body#/termination_reason
  outputs:
    clusterId: $steps.createCluster.outputs.clusterId
    finalState: $steps.confirmRunning.outputs.state
    failureState: $steps.reportFailure.outputs.state