传感器的气流EMR执行步骤

Question

我在气流中制作了以下DAG，在其中执行一组EMRStep以运行管道。

default_args = {
    'owner': 'airflow',
    'depends_on_past': False,
    'start_date': datetime(2017, 07, 20, 10, 00),
    'email': ['airflow@airflow.com'],
    'email_on_failure': False,
    'email_on_retry': False,
    'retries': 5,
    'retry_delay': timedelta(minutes=2),
}

dag = DAG('dag_import_match_hourly',
      default_args=default_args,
      description='Fancy Description',
      schedule_interval=timedelta(hours=1),
      dagrun_timeout=timedelta(hours=2))

try:
    merge_s3_match_step = EmrAddStepsOperator(
        task_id='merge_s3_match_step',
        job_flow_id=cluster_id,
        aws_conn_id='aws_default',
        steps=create_step('Merge S3 Match'),
        dag=dag
    )

    mapreduce_step = EmrAddStepsOperator(
        task_id='mapreduce_match_step',
        job_flow_id=cluster_id,
        aws_conn_id='aws_default',
        steps=create_step('MapReduce Match Hourly'),
        dag=dag
    )

    merge_hdfs_step = EmrAddStepsOperator(
        task_id='merge_hdfs_step',
        job_flow_id=cluster_id,
        aws_conn_id='aws_default',
        steps=create_step('Merge HDFS Match Hourly'),
        dag=dag
    )

    ## Sensors
    check_merge_s3 = EmrStepSensor(
        task_id='watch_merge_s3',
        job_flow_id=cluster_id,
        step_id="{{ task_instance.xcom_pull('merge_s3_match_step', key='return_value')[0] }}",
        aws_conn_id='aws_default',
        dag=dag
    )

    check_mapreduce = EmrStepSensor(
        task_id='watch_mapreduce',
        job_flow_id=cluster_id,
        step_id="{{ task_instance.xcom_pull('mapreduce_match_step', key='return_value')[0] }}",
        aws_conn_id='aws_default',
        dag=dag
    )

    check_merge_hdfs = EmrStepSensor(
        task_id='watch_merge_hdfs',
        job_flow_id=cluster_id,
        step_id="{{ task_instance.xcom_pull('merge_hdfs_step', key='return_value')[0] }}",
        aws_conn_id='aws_default',
        dag=dag
    )

    mapreduce_step.set_upstream(merge_s3_match_step)
    merge_s3_match_step.set_downstream(check_merge_s3)

    mapreduce_step.set_downstream(check_mapreduce)

    merge_hdfs_step.set_upstream(mapreduce_step)
    merge_hdfs_step.set_downstream(check_merge_hdfs)

except AirflowException as ae:
    print ae.message

DAG可以正常工作，但是我希望使用传感器来确保仅当EMR作业已正确完成时才执行下一步。我尝试了几件事，但都没有用。上面的代码无法正常工作。有人知道如何使用EMRSensorStep实现我的目标吗？

Answer 1

您的EmrStepSensor任务似乎需要设置正确的依赖关系，例如check_mapreduce，如果要等待check_mapreduce完成，则下一步应该为merge_hdfs_step.set_upstream(check_mapreduce)或check_mapreduce.set_downstream(merge_hdfs_step)。因此它将是TaskA >> SensorA >> TaskB >> SensorB >> TaskC >> SensorC，请尝试使用这种方式来设置依赖项