捆绑包配置示例

项目
08/01/2024

本文提供了捆绑包功能和常见捆绑包用例的示例配置。

提示

本文中的部分示例以及其他示例可以在捆绑包示例存储库中找到。

使用无服务器计算的作业

Databricks 资产捆绑包支持在无服务器计算上运行的作业。若要对此进行配置，可以省略作业的 clusters 设置，也可以指定环境，如以下示例所示。

# A serverless job (no cluster definition)
resources:
  jobs:
    serverless_job_no_cluster:
      name: serverless_job_no_cluster

      email_notifications:
        on_failure:
          - someone@example.com

      tasks:
        - task_key: notebook_task
          notebook_task:
            notebook_path: ../src/notebook.ipynb

# A serverless job (environment spec)
resources:
  jobs:
    serverless_job_environment:
      name: serverless_job_environment

      tasks:
        - task_key: task
          spark_python_task:
            python_file: ../src/main.py

          # The key that references an environment spec in a job.
          environment_key: default

      # A list of task execution environment specifications that can be referenced by tasks of this job.
      environments:
        - environment_key: default

          # Full documentation of this spec can be found at:
          # https://docs.databricks.com/api/workspace/jobs/create#environments-spec
          spec:
            client: "1"
            dependencies:
              - cowsay

使用无服务器计算的管道

Databricks 资产捆绑包支持在无服务器计算上运行的管道。若要对此进行配置，请将管道 serverless 设置设置为 true。以下示例配置定义了一个在无服务器计算上运行的管道，以及一个每小时触发管道刷新的作业。

# A pipeline that runs on serverless compute
resources:
  pipelines:
    my_pipeline:
      name: my_pipeline
      target: ${bundle.environment}
      serverless: true
      catalog: users
      libraries:
        - notebook:
            path: ../src/my_pipeline.ipynb

      configuration:
        bundle.sourcePath: /Workspace/${workspace.file_path}/src

# This defines a job to refresh a pipeline that is triggered every hour
resources:
  jobs:
    my_job:
      name: my_job

      # Run this job once an hour.
      trigger:
        periodic:
          interval: 1
          unit: HOURS

      email_notifications:
        on_failure:
          - someone@example.com

      tasks:
        - task_key: refresh_pipeline
          pipeline_task:
            pipeline_id: ${resources.pipelines.my_pipeline.id}

包含 SQL 笔记本的作业

以下示例配置定义了一个包含 SQL 笔记本的作业。

resources:
  jobs:
    job_with_sql_notebook:
      name: Job to demonstrate using a SQL notebook with a SQL warehouse

      tasks:
        - task_key: notebook
          notebook_task:
            notebook_path: ./select.sql
            warehouse_id: 799f096837fzzzz4

使用多个 wheel 文件的作业

以下示例配置定义了一个包含一个作业的捆绑包，该作业使用了多个 *.whl 文件。

# job.yml
resources:
  jobs:
    example_job:
      name: "Example with multiple wheels"
      tasks:
        - task_key: task

          spark_python_task:
            python_file: ../src/call_wheel.py

          libraries:
            - whl: ../my_custom_wheel1/dist/*.whl
            - whl: ../my_custom_wheel2/dist/*.whl

          new_cluster:
            node_type_id: i3.xlarge
            num_workers: 0
            spark_version: 14.3.x-scala2.12
            spark_conf:
                "spark.databricks.cluster.profile": "singleNode"
                "spark.master": "local[*, 4]"
            custom_tags:
                "ResourceClass": "SingleNode"

# databricks.yml
bundle:
  name: job_with_multiple_wheels

include:
  - ./resources/job.yml

workspace:
  host: https://myworkspace.cloud.databricks.com

artifacts:
  my_custom_wheel1:
    type: whl
    build: poetry build
    path: ./my_custom_wheel1

  my_custom_wheel2:
    type: whl
    build: poetry build
    path: ./my_custom_wheel2

targets:
  dev:
    default: true
    mode: development

使用 requirements.txt 文件的作业

以下示例配置定义了一个使用 requirements.txt 文件的作业。

resources:
  jobs:
    job_with_requirements_txt:
      name: Example job that uses a requirements.txt file

      tasks:
        - task_key: task
          job_cluster_key: default
          spark_python_task:
            python_file: ../src/main.py
          libraries:
            - requirements: /Workspace/${workspace.file_path}/requirements.txt

将 JAR 文件上传到 Unity Catalog 的捆绑包

可以将 Unity Catalog 卷指定为项目路径，以便将 JAR 文件和 wheel 文件等所有项目上传到 Unity Catalog 卷。以下示例捆绑包将 JAR 文件上传到 Unity Catalog。有关 artifact_path 映射的信息，请参阅 artifact_path。

bundle:
  name: jar-bundle

workspace:
  host: https://myworkspace.cloud.databricks.com
  artifact_path: /Volumes/main/default/my_volume

artifacts:
  my_java_code:
    path: ./sample-java
    build: "javac PrintArgs.java && jar cvfm PrintArgs.jar META-INF/MANIFEST.MF PrintArgs.class"
    files:
      - source: ./sample-java/PrintArgs.jar

resources:
  jobs:
    jar_job:
      name: "Spark Jar Job"
      tasks:
        - task_key: SparkJarTask
          new_cluster:
            num_workers: 1
            spark_version: "14.3.x-scala2.12"
            node_type_id: "i3.xlarge"
          spark_jar_task:
            main_class_name: PrintArgs
          libraries:
            - jar: ./sample-java/PrintArgs.jar

通过