在Airflow中,我们可以使用on_failure_callback参数在任务失败时执行自定义代码。然而,如果您使用的是airflow-python包,建议使用它提供的异常包装器,以获取更好的异常处理和通知能力。
具体来说,该包装器可以捕获Airflow任务中出现的异常,并将其包装为标准的Python异常格式。然后可以使用通知工具(如电子邮件、Slack等)将此异常发送给相关人员以便追踪和解决问题。
以下是一个示例,演示如何在DAG中使用此异常包装器:
from airflow import DAG
from airflow.operators.python_operator import PythonOperator
from airflow.models.baseoperator import BaseOperator
from airflow.utils.decorators import apply_defaults
from airflow.exceptions import AirflowException
class ExceptionWrapperOperator(BaseOperator):
"""
A simple operator to wrap PythonOperators with exception handling for
on_failure_callback. By default it will raise the original exception after
calling on_failure_callback hook.
"""
template_fields = ('op', 'on_failure_callback')
@apply_defaults
def __init__(self, op, on_failure_callback, raise_exception=True, *args, **kwargs):
super().__init__(*args, **kwargs)
self.op = op
self.on_failure_callback = on_failure_callback
self.raise_exception = raise_exception
def execute(self, context):
try:
return self.op.execute(context)
except AirflowException as e:
self.log.error(f'Airflow exception encountered: {e}')
if self.on_failure_callback:
self.log.info(f'Executing on_failure_callback: {self.on_failure_callback}')
self.on_failure_callback(context)
if self.raise_exception:
raise e
def my_on_failure_callback(context):
# TODO: send notification
print("task failed, do something!")
dag = DAG(dag_id='my_dag', default_args={
'start_date': datetime(2022, 1, 1)
})
def my_task():
# something interesting..
1/0
with