构建您自己的流水线组件

通常,在运行组件时,您不仅会看到指向正在启动的组件作业的链接,而且还会看到底层云资源(如 Vertex 批量预测作业或 Dataflow 作业)的链接。

gcp_resource proto 是一个特殊参数,您可以在组件中使用该参数,以便 Google Cloud 控制台在 Vertex AI Pipelines 控制台中提供资源日志和状态的自定义视图。

输出 gcp_resource 参数

使用基于容器的组件

首先,您需要在组件中定义 gcp_resource 参数,如下面的 component.py 文件示例所示:

Python

如需了解如何安装或更新 Vertex AI SDK for Python,请参阅安装 Vertex AI SDK for Python。 如需了解详情,请参阅 Python API 参考文档

# Copyright 2023 The Kubeflow Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from typing import List

from google_cloud_pipeline_components import _image
from google_cloud_pipeline_components import _placeholders
from kfp.dsl import container_component
from kfp.dsl import ContainerSpec
from kfp.dsl import OutputPath


@container_component
def dataflow_python(
    python_module_path: str,
    temp_location: str,
    gcp_resources: OutputPath(str),
    location: str = 'us-central1',
    requirements_file_path: str = '',
    args: List[str] = [],
    project: str = _placeholders.PROJECT_ID_PLACEHOLDER,
):
  # fmt: off
  """Launch a self-executing Beam Python file on Google Cloud using the
  Dataflow Runner.

  Args:
      location: Location of the Dataflow job. If not set, defaults to `'us-central1'`.
      python_module_path: The GCS path to the Python file to run.
      temp_location: A GCS path for Dataflow to stage temporary job files created during the execution of the pipeline.
      requirements_file_path: The GCS path to the pip requirements file.
      args: The list of args to pass to the Python file. Can include additional parameters for the Dataflow Runner.
      project: Project to create the Dataflow job. Defaults to the project in which the PipelineJob is run.

  Returns:
      gcp_resources: Serialized gcp_resources proto tracking the Dataflow job. For more details, see https://github.com/kubeflow/pipelines/blob/master/components/google-cloud/google_cloud_pipeline_components/proto/README.md.
  """
  # fmt: on
  return ContainerSpec(
      image=_image.GCPC_IMAGE_TAG,
      command=[
          'python3',
          '-u',
          '-m',
          'google_cloud_pipeline_components.container.v1.dataflow.dataflow_launcher',
      ],
      args=[
          '--project',
          project,
          '--location',
          location,
          '--python_module_path',
          python_module_path,
          '--temp_location',
          temp_location,
          '--requirements_file_path',
          requirements_file_path,
          '--args',
          args,
          '--gcp_resources',
          gcp_resources,
      ],
  )

接下来,在容器中安装 Google Cloud 流水线组件软件包:

pip install --upgrade google-cloud-pipeline-components

接下来,在 Python 代码中,将资源定义为 gcp_resource 参数:

Python

如需了解如何安装或更新 Vertex AI SDK for Python,请参阅安装 Vertex AI SDK for Python。 如需了解详情,请参阅 Python API 参考文档

from google_cloud_pipeline_components.proto.gcp_resources_pb2 import GcpResources
from google.protobuf.json_format import MessageToJson

dataflow_resources = GcpResources()
dr = dataflow_resources.resources.add()
dr.resource_type='DataflowJob'
dr.resource_uri='https://dataflow.googleapis.com/v1b3/projects/[your-project]/locations/us-east1/jobs/[dataflow-job-id]'

with open(gcp_resources, 'w') as f:
    f.write(MessageToJson(dataflow_resources))

使用 Python 组件

或者,您可以返回 gcp_resources 输出参数,就像返回任何字符串输出参数一样:

@dsl.component(
    base_image='python:3.9',
    packages_to_install=['google-cloud-pipeline-components==2.16.1'],
)
def launch_dataflow_component(project: str, location:str) -> NamedTuple("Outputs",  [("gcp_resources", str)]):
  # Launch the dataflow job
  dataflow_job_id = [dataflow-id]
  dataflow_resources = GcpResources()
  dr = dataflow_resources.resources.add()
  dr.resource_type='DataflowJob'
  dr.resource_uri=f'https://dataflow.googleapis.com/v1b3/projects/{project}/locations/{location}/jobs/{dataflow_job_id}'
  gcp_resources=MessageToJson(dataflow_resources)
  return gcp_resources

支持的 resource_type

您可以将 resource_type 设置为任意字符串,但只有以下类型在 Google Cloud 控制台中具有链接:

  • BatchPredictionJob
  • BigQueryJob
  • CustomJob
  • DataflowJob
  • HyperparameterTuningJob

编写组件以取消底层资源

取消流水线作业后,Google Cloud 底层资源默认会继续运行。它们不会自动取消。要更改此行为,您应该将 SIGTERM 处理程序附加到流水线作业。建议在长时间运行作业的轮询循环之前执行此操作。

多个 Google Cloud 流水线组件已实现取消操作,包括:

  • 批量预测作业
  • BigQuery ML 作业
  • 自定义作业
  • Dataproc Serverless 批处理作业
  • 超参数调节作业

如需了解详情(包括说明如何附加 SIGTERM 处理程序的代码示例),请参阅以下 GitHub 链接:

实现 SIGTERM 处理程序时,请考虑以下事项:

  • 取消传播仅在组件运行几分钟后才生效。这通常是因为需要在调用 Python 信号处理程序之前处理后台启动任务。
  • 某些 Google Cloud 资源可能未实现取消操作。例如,创建或删除 Vertex AI 端点或模型可能会创建一个长时间运行的操作,以通过其 REST API 接受取消请求,但不会实现取消操作本身。