Organiza tus páginas con colecciones
Guarda y categoriza el contenido según tus preferencias.
Escribe un componente para mostrar un vínculo de la consola de Google Cloud
Es común que cuando ejecutas un componente, no solo desees ver el vínculo al trabajo de componente que se inicia, sino también el vínculo a los recursos subyacentes de la nube, como los trabajos de predicción por lotes de Vertex o los trabajos de Dataflow.
El proto gcp_resource es un parámetro especial que puedes usar en el componente para permitir que la consola de Google Cloud proporcione una vista personalizada de los registros y el estado del recurso en la consola de Vertex AI Pipelines.
Obtén el resultado del parámetro gcp_resource.
Usa un componente basado en contenedores
Primero, deberás definir el parámetro gcp_resource en el componente, como se muestra en el siguiente archivo component.py de ejemplo:
# 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.fromtypingimportListfromgoogle_cloud_pipeline_componentsimport_imagefromgoogle_cloud_pipeline_componentsimport_placeholdersfromkfp.dslimportcontainer_componentfromkfp.dslimportContainerSpecfromkfp.dslimportOutputPath@container_componentdefdataflow_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: onreturnContainerSpec(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,],)
A continuación, dentro del contenedor, instala el paquete de componentes de canalización Google Cloud :
Puedes configurar resource_type para que sea una cadena arbitraria, pero solo los siguientes tipos tienen vínculos en la consola de Google Cloud :
BatchPredictionJob
BigQueryJob
CustomJob
DataflowJob
HyperparameterTuningJob
Escribe un componente para cancelar los recursos subyacentes
Cuando se cancela un trabajo de canalización, el comportamiento predeterminado es que los recursos Google Cloud subyacentes se sigan ejecutando. No se cancelan de forma automática. Para cambiar este comportamiento, debes adjuntar un controlador SIGTERM al trabajo de canalización. Un buen lugar para hacerlo es justo antes de un bucle de sondeo para un trabajo que podría ejecutarse durante mucho tiempo.
La cancelación se implementó en varios Google Cloud componentes de la canalización, incluidos los siguientes:
Trabajo de predicción por lotes
Trabajo de BigQuery ML
Trabajo personalizado
Trabajo por lotes sin servidores de Dataproc
Trabajos de ajuste de hiperparámetros
Para obtener más información, incluido un código de muestra que indica cómo adjuntar un controlador de SIGTERM, consulta los siguientes vínculos de GitHub:
Ten en cuenta lo siguiente cuando implementes tu controlador de SIGTERM:
La propagación de cancelación funciona solo después de que el componente haya estado en ejecución durante unos minutos. Por lo general, esto se debe a las tareas de inicio en segundo plano que deben procesarse antes de que se llame a los controladores de señales de Python.
Es posible que algunos Google Cloud recursos no tengan implementada la cancelación. Por ejemplo, crear o borrar un extremo o modelo de Vertex AI podría crear una operación de larga duración que acepte una solicitud de cancelación a través de su API de REST, pero no implementa la operación de cancelación en sí.
[[["Fácil de comprender","easyToUnderstand","thumb-up"],["Resolvió mi problema","solvedMyProblem","thumb-up"],["Otro","otherUp","thumb-up"]],[["Difícil de entender","hardToUnderstand","thumb-down"],["Información o código de muestra incorrectos","incorrectInformationOrSampleCode","thumb-down"],["Faltan la información o los ejemplos que necesito","missingTheInformationSamplesINeed","thumb-down"],["Problema de traducción","translationIssue","thumb-down"],["Otro","otherDown","thumb-down"]],["Última actualización: 2025-09-04 (UTC)"],[],[],null,["# Build your own pipeline components\n\n| To learn more,\n| run the \"Custom training workflow with prebuilt Pipeline Components and custom components\" notebook in one of the following\n| environments:\n|\n| [Open in Colab](https://colab.research.google.com/github/GoogleCloudPlatform/vertex-ai-samples/blob/main/notebooks/official/pipelines/google_cloud_pipeline_components_model_train_upload_deploy.ipynb)\n|\n|\n| \\|\n|\n| [Open in Colab Enterprise](https://console.cloud.google.com/vertex-ai/colab/import/https%3A%2F%2Fraw.githubusercontent.com%2FGoogleCloudPlatform%2Fvertex-ai-samples%2Fmain%2Fnotebooks%2Fofficial%2Fpipelines%2Fgoogle_cloud_pipeline_components_model_train_upload_deploy.ipynb)\n|\n|\n| \\|\n|\n| [Open\n| in Vertex AI Workbench](https://console.cloud.google.com/vertex-ai/workbench/deploy-notebook?download_url=https%3A%2F%2Fraw.githubusercontent.com%2FGoogleCloudPlatform%2Fvertex-ai-samples%2Fmain%2Fnotebooks%2Fofficial%2Fpipelines%2Fgoogle_cloud_pipeline_components_model_train_upload_deploy.ipynb)\n|\n|\n| \\|\n|\n| [View on GitHub](https://github.com/GoogleCloudPlatform/vertex-ai-samples/blob/main/notebooks/official/pipelines/google_cloud_pipeline_components_model_train_upload_deploy.ipynb)\n\nWrite a component to show a Google Cloud console link\n-----------------------------------------------------\n\nIt's common that when running a component, you want to not only see the link to the component job being launched, but also the link to the underlying cloud resources, such as the Vertex batch prediction jobs or dataflow jobs.\n\nThe [`gcp_resource` proto](https://github.com/kubeflow/pipelines/tree/master/components/google-cloud/google_cloud_pipeline_components/proto) is a special parameter that you can use in your component to enable the Google Cloud console to provide a customized view of the resource's logs and status in the Vertex AI Pipelines console.\n\n### Output the `gcp_resource` parameter\n\n#### Using a container-based component\n\nFirst, you'll need to define the `gcp_resource` parameter in your component as shown in the following example `component.py` file: \n\n### Python\n\nTo learn how to install or update the Vertex AI SDK for Python, see [Install the Vertex AI SDK for Python](/vertex-ai/docs/start/use-vertex-ai-python-sdk).\n\nFor more information, see the\n[Python API reference documentation](/python/docs/reference/aiplatform/latest).\n\n # Copyright 2023 The Kubeflow Authors. All Rights Reserved.\n #\n # Licensed under the Apache License, Version 2.0 (the \"License\");\n # you may not use this file except in compliance with the License.\n # You may obtain a copy of the License at\n #\n # http://www.apache.org/licenses/LICENSE-2.0\n #\n # Unless required by applicable law or agreed to in writing, software\n # distributed under the License is distributed on an \"AS IS\" BASIS,\n # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n # See the License for the specific language governing permissions and\n # limitations under the License.\n from typing import List\n\n from google_cloud_pipeline_components import _image\n from google_cloud_pipeline_components import _placeholders\n from kfp.dsl import container_component\n from kfp.dsl import ContainerSpec\n from kfp.dsl import OutputPath\n\n\n @container_component\n def dataflow_python(\n python_module_path: str,\n temp_location: str,\n gcp_resources: OutputPath(str),\n location: str = 'us-central1',\n requirements_file_path: str = '',\n args: List[str] = [],\n project: str = _placeholders.PROJECT_ID_PLACEHOLDER,\n ):\n # fmt: off\n \"\"\"Launch a self-executing Beam Python file on Google Cloud using the\n Dataflow Runner.\n\n Args:\n location: Location of the Dataflow job. If not set, defaults to `'us-central1'`.\n python_module_path: The GCS path to the Python file to run.\n temp_location: A GCS path for Dataflow to stage temporary job files created during the execution of the pipeline.\n requirements_file_path: The GCS path to the pip requirements file.\n args: The list of args to pass to the Python file. Can include additional parameters for the Dataflow Runner.\n project: Project to create the Dataflow job. Defaults to the project in which the PipelineJob is run.\n\n Returns:\n 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.\n \"\"\"\n # fmt: on\n return ContainerSpec(\n image=_image.GCPC_IMAGE_TAG,\n command=[\n 'python3',\n '-u',\n '-m',\n 'google_cloud_pipeline_components.container.v1.dataflow.dataflow_launcher',\n ],\n args=[\n '--project',\n project,\n '--location',\n location,\n '--python_module_path',\n python_module_path,\n '--temp_location',\n temp_location,\n '--requirements_file_path',\n requirements_file_path,\n '--args',\n args,\n '--gcp_resources',\n gcp_resources,\n ],\n )\n\n\u003cbr /\u003e\n\nNext, inside the container, install the Google Cloud Pipeline Components package: \n\n pip install --upgrade google-cloud-pipeline-components\n\nNext, in the Python code, define the resource as a `gcp_resource` parameter: \n\n### Python\n\nTo learn how to install or update the Vertex AI SDK for Python, see [Install the Vertex AI SDK for Python](/vertex-ai/docs/start/use-vertex-ai-python-sdk).\n\nFor more information, see the\n[Python API reference documentation](/python/docs/reference/aiplatform/latest).\n\n from google_cloud_pipeline_components.proto.gcp_resources_pb2 import GcpResources\n from google.protobuf.json_format import MessageToJson\n\n dataflow_resources = GcpResources()\n dr = dataflow_resources.resources.add()\n dr.resource_type='DataflowJob'\n dr.resource_uri='https://dataflow.googleapis.com/v1b3/projects/[your-project]/locations/us-east1/jobs/[dataflow-job-id]'\n\n with open(gcp_resources, 'w') as f:\n f.write(MessageToJson(dataflow_resources))\n\n\u003cbr /\u003e\n\n#### Using a Python component\n\nAlternatively, you can return the `gcp_resources` output parameter as you would any string output parameter: \n\n @dsl.component(\n base_image='python:3.9',\n packages_to_install=['google-cloud-pipeline-components==2.19.0'],\n )\n def launch_dataflow_component(project: str, location:str) -\u003e NamedTuple(\"Outputs\", [(\"gcp_resources\", str)]):\n # Launch the dataflow job\n dataflow_job_id = [dataflow-id]\n dataflow_resources = GcpResources()\n dr = dataflow_resources.resources.add()\n dr.resource_type='DataflowJob'\n dr.resource_uri=f'https://dataflow.googleapis.com/v1b3/projects/{project}/locations/{location}/jobs/{dataflow_job_id}'\n gcp_resources=MessageToJson(dataflow_resources)\n return gcp_resources\n\n#### Supported `resource_type` values\n\nYou can set the `resource_type` to be an arbitrary string, but only the following types have links in the Google Cloud console:\n\n- BatchPredictionJob\n- BigQueryJob\n- CustomJob\n- DataflowJob\n- HyperparameterTuningJob\n\nWrite a component to cancel the underlying resources\n----------------------------------------------------\n\nWhen a pipeline job is canceled, the default behavior is for the underlying Google Cloud resources to keep running. They are not canceled automatically. To change this behavior, you should attach a [SIGTERM](https://docs.python.org/3/library/signal.html#signal.SIGTERM) handler to the pipeline job. A good place to do this is just before a polling loop for a job that could run for a long time.\n\nCancellation has been implemented on several Google Cloud Pipeline Components, including:\n\n- Batch prediction job\n- BigQuery ML job\n- Custom job\n- Dataproc Serverless batch job\n- Hyperparameter tuning job\n\nFor more information, including sample code that shows how to attach a SIGTERM handler, see the following GitHub links:\n\n- \u003chttps://github.com/kubeflow/pipelines/blob/google-cloud-pipeline-components-2.19.0/components/google-cloud/google_cloud_pipeline_components/container/utils/execution_context.py\u003e\n- \u003chttps://github.com/kubeflow/pipelines/blob/google-cloud-pipeline-components-2.19.0/components/google-cloud/google_cloud_pipeline_components/container/v1/gcp_launcher/job_remote_runner.py#L124\u003e\n\nConsider the following when implementing your SIGTERM handler:\n\n- Cancellation propagation works only after the component has been running for a few minutes. This is typically due to background startup tasks that need to be [processed](https://docs.python.org/3/library/signal.html#execution-of-python-signal-handlers) before the Python signal handlers are called.\n- Some Google Cloud resources might not have cancellation implemented. For example, creating or deleting a Vertex AI Endpoint or Model could create a long-running operation that accepts a cancellation request through its REST API, but doesn't implement the cancellation operation itself."]]