Fokko commented on a change in pull request #6210: [AIRFLOW-5567] BaseReschedulePokeOperator URL: https://github.com/apache/airflow/pull/6210#discussion_r341844497
########## File path: airflow/models/base_async_operator.py ########## @@ -0,0 +1,161 @@ +# -*- coding: utf-8 -*- +# +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. + +""" +Base Asynchronous Operator for kicking off a long running +operations and polling for completion with reschedule mode. +""" + +from abc import abstractmethod +from typing import Dict, List, Optional, Union + +from airflow.models import SkipMixin, TaskReschedule +from airflow.models.xcom import XCOM_EXTERNAL_RESOURCE_ID_KEY +from airflow.sensors.base_sensor_operator import BaseSensorOperator +from airflow.utils.decorators import apply_defaults + + Review comment: I can't see why we need another `execute_phase_{start,final}`. The pre-execute can be overridden by a custom implementation, but in practice, this hasn't been done much. So the `TaskInstance._run_raw_task()` is what's being executed on the worker, and this will just actually run the whole task to the end. > The justification is that these execution phases are tightly coupled and should be retried together (and in order) as some kind of atomic unit. I really agree with that. We've seen situations when the DAG would start with creating a cluster, but this breaks the atomicity of the DAG. When the cluster is being killed (maybe because of idle timeout), the only way to retry the workflow, is by restarting the whole dag. It is still not clear to me how the hierarchy will look like. I think we should move most of the reschedule logic into the BaseOperator. Maybe we can enrich the `context` which is being passed to the execute method with reschedule logic to let the implementation of the execute step know that you're rescheduling. ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected] With regards, Apache Git Services
