Airflow on_failure_callback

user_10626286 0
Viewed 15092

I have an Airflow DAG with two tasks:

  • read_csv
  • process_file

They work fine on their own. I purposely created a typo in a pandas Dataframe to learn how on_failure_callback works and to see if it is being triggered. It seems likes from the log that it doesn't:

Traceback (most recent call last):
  File "/usr/local/lib/python3.7/site-packages/airflow/models/taskinstance.py", line 1197, in handle_failure
    task.on_failure_callback(context)
TypeError: on_failure_callback() takes 0 positional arguments but 1 was given

Why isn't on_failure_callback working?

Here is a visual representation of the DAG:

enter image description here

Here is the code:

try:

    from datetime import timedelta
    from airflow import DAG
    from airflow.operators.python_operator import PythonOperator
    from datetime import datetime
    import pandas as pd

    # Setting up Triggers
    from airflow.utils.trigger_rule import TriggerRule

    # for Getting Variables from airlfow
    from airflow.models import Variable

    print("All Dag modules are ok ......")
except Exception as e:
    print("Error  {} ".format(e))


def read_csv(**context):
    data = [{"name":"Soumil","title":"Full Stack Software Engineer"}, { "name":"Nitin","title":"Full Stack Software Engineer"},]
    df = pd.DataFramee(data=data)

    dag_config = Variable.get("VAR1")
    print("VAR 1 is : {} ".format(dag_config))
    context['ti'].xcom_push(key='mykey', value=df)


def process_file(**context):
    instance = context.get("ti").xcom_pull(key='mykey')
    print(instance.head(2))
    return "Process complete "


def on_failure_callback(**context):
    print("Fail works  !  ")



with DAG(dag_id="invoices_dag",
         schedule_interval="@once",
         default_args={
             "owner": "airflow",
             "start_date": datetime(2020, 11, 1),
             "retries": 1,
             "retry_delay": timedelta(minutes=1),
             'on_failure_callback': on_failure_callback,
         },
         catchup=False) as dag:

    read_csv = PythonOperator(
        task_id="read_csv",
        python_callable=read_csv,
        op_kwargs={'filename': "Soumil.csv"},
        provide_context=True
    )

    process_file = PythonOperator(
        task_id="process_file",
        python_callable=process_file,
        provide_context=True
    )




read_csv >> process_file





# ====================================Notes====================================

# all_success           -> triggers when all tasks arecomplete
# one_success           -> trigger when one task is complete
# all_done              -> Trigger when all Tasks are Done
# all_failed            -> Trigger when all task Failed
# one_failed            -> one task is failed
# none_failed           -> No Task Failed

# ==============================================================================



# ============================== Executor====================================

# There are Three main  types of executor
# -> Sequential Executor  run single task in linear fashion wih no parllelism default Dev
# -> Local Exector  run each task in seperate process
# -> Celery Executor Run each worker node within multi node architecture Most scalable

# ===========================================================================
airflow
1 Answers

You need to specify one argument to your function that can receive the context this is due to how Airflow triggers on_failure_callback

def on_failure_callback(context):
    print("Fail works  !  ")

Note that with your implementation you can't tell from the message which task has failed so you might want to add to your error message the task details like:

def on_failure_callback(context):
    ti = context['task_instance']
    print(f"task {ti.task_id } failed in dag { ti.dag_id } ")
user_146244090
Related