CODEWITHSUNDEEP
pythonemrairflow
spaghettifunk
spaghettifunk

Reputation: 2014

Airflow EMR execute step from Sensor

I made the following DAG in airflow where I am executing a set of EMRSteps to run my pipeline.

default_args = {
    'owner': 'airflow',
    'depends_on_past': False,
    'start_date': datetime(2017, 07, 20, 10, 00),
    'email': ['[email protected]'],
    '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

The DAG works fine but I would like to use the Sensors to make sure that I am going to execute the next step if and only if the EMR Job has been completed correctly. I tried few things but none of them are working. The code above doesn't do the job properly. Does someone know how to use the EMRSensorStep to achieve my goal ?

Upvotes: 4

Views: 6070

Answers (1)

Chengzhi
Chengzhi

Reputation: 2591

It looks like your EmrStepSensor tasks need to set correct dependencies, for example, check_mapreduce, if you want to wait for check_mapreduce to complete, the next step should be merge_hdfs_step.set_upstream(check_mapreduce) or check_mapreduce.set_downstream(merge_hdfs_step). So it would be TaskA>>SensorA>>TaskB>>SensorB>>TaskC>>SensorC, try to use this way to set up the dependencies

Upvotes: 7

Related Questions