我正在尝试自定义我的PythonOperator并将其放在$ AIRFLOW_HOME / plugins下,如下所示:
class MyPythonOperator(PythonOperator):
def my_callable(param1, param2, param3):
# do something
@apply_defaults
def __init__(self, task_id, *args, **kwargs):
super(MyPythonOperator, self).__init__(
task_id=task_id,
python_callable = self.my_callable,
provide_context = True,
*args, **kwargs)
然后,我定义一个气流dag代码,这非常简单,只需执行两个任务即可:
args = {
'owner': 'airflow',
'start_date': airflow.utils.dates.days_ago(2),
}
dag = DAG(
dag_id='example_workflow',
default_args=args,
schedule_interval='0 0 * * *',
dagrun_timeout=timedelta(minutes=60),
)
task1 = MyPythonOperator(
task_id='task1',
params={'param1': 'param1_value',
'param2': 'param2_value',
'param3': 'param3_value'},
dag=dag
)
task2 = MyPythonOperator(
task_id='task2',
params={'param1': 'param1_value',
'param2': 'param2_value',
'param2': 'param3_value'},
dag=dag
)
task1 >> task2
但是我运行dag python代码后,得到错误消息:
$ python example_airflow_code.py
[2019-05-15 19:51:10,338] {init.py:51} INFO - Using executor SequentialExecutor
usage: example_airflow_code.py [-h]
{list_tasks,backfill,test,run,pause,unpause,list_dag_runs}
...
example_airflow_code.py: error: too few arguments
我尝试了一些调试,并在此行插入一个断点:
super(MyPythonOperator, self).init(
我在调用超级构造函数之前发现,self.dag和self.dag_id的值异常,值为:
str: Traceback (most recent call last):
File "/Applications/Eclipse.app/Contents/Eclipse/plugins/org.python.pydev.core_6.4.4.201807281807/pysrc/_pydevd_bundle/pydevd_resolver.py", line 166, in _getPyDictionary
attr = getattr(var, n)
File "/Users/zhuangxy/anaconda2/lib/python2.7/site-packages/airflow/models/init.py", line 2399, in dag_id
return 'adhoc_' + self.owner
AttributeError: 'MyPythonOperator' object has no attribute 'owner'
任何人都知道这个例子有什么问题吗? 非常感谢你!
答案 0 :(得分:0)
我最近也遇到了这个。看来您缺少自定义PythonOperator上的context
参数。
更改方法定义,使其看起来像这样:
def my_callable(param1, param2, param3, **context):
# do something
失败的原因是您在操作员中提供的provide_context=True
标志。由于某种原因,可调用的python正在您的参数中寻找它。