为何Docker Swarm Operator在DAG运行结束时检查容器致任务失败？

环境配置 #

• 3节点Docker Swarm集群：1个管理节点 + 2个工作节点
• 管理节点配置为在2375端口暴露Docker API，该地址作为DockerSwarmOperator的docker_url参数传入
• Airflow通过Docker Compose部署在Swarm管理节点上

DAG 运行代码 #

import datetime

from airflow import DAG
from airflow.providers.docker.operators.docker_swarm import DockerSwarmOperator
from docker.types import Mount, NetworkAttachmentConfig

with DAG(
    dag_id="movie_retriever_dag",
    start_date=datetime.datetime(2025, 1, 4),
):
    extraction_container = DockerSwarmOperator(
        task_id="movie-extract_transform_load",
        image="movie-extract_transform_load-image:latest",
        command="python ./extract_transform_load.py -t \"{{ dag_run.conf['title'] }}\"",
        mount_tmp_dir=False,
        mounts=[
            Mount(
                target="/app/temp_data",
                source="/mnt/storage-server0/sda3/airflow/tmp",
                type="bind",
            ),
            Mount(
                target="/app/appdata/db.sqlite",
                source="/mnt/storage-server0/sda3/portfolio/data/db.sqlite",
                type="bind",
            ),
        ],
        auto_remove=True,
        networks=[NetworkAttachmentConfig(target="grafana_loki")],
        docker_url="tcp://192.168.0.173:2375",
    )

    extraction_container

实际运行流程 #

1. DAG启动运行
2. 成功在Swarm集群中创建服务
3. 服务调度到某工作节点运行容器，容器成功完成所有操作
4. DAG执行失败，抛出404错误

错误日志 #

[2025-01-12, 22:05:30 CET] {docker_swarm.py:205} INFO - Service status before exiting: complete
[2025-01-12, 22:05:30 CET] {taskinstance.py:3311} ERROR - Task failed with exception
Traceback (most recent call last):
  File "/home/airflow/.local/lib/python3.12/site-packages/docker/api/client.py", line 275, in _raise_for_status
    response.raise_for_status()
  File "/home/airflow/.local/lib/python3.12/site-packages/requests/models.py", line 1024, in raise_for_status
    raise HTTPError(http_error_msg, response=self)
requests.exceptions.HTTPError: 404 Client Error: Not Found for url: http://192.168.0.173:2375/v1.47/containers/6ec661d35385d58aa9e91e7b8a0e6e03fe920f8f8ba079a7cdf7cfdd12fe2e0f/json

经排查，报错中查找的容器ID6ec661d35385d58aa9e91e7b8a0e6e03fe920f8f8ba079a7cdf7cfdd12fe2e0f实际运行在Swarm工作节点上，管理节点无法直接查询到该容器信息。

核心疑问 #

• DockerSwarmOperator为什么要发起针对特定容器的HTTP请求？
• 既然是Swarm集群环境，为什么不通过检查服务的状态来判断任务是否完成，而是去查询单个容器？
• 当前Airflow的部署方式（在管理节点用Docker Compose部署）是否是导致这个问题的原因？

内容的提问来源于stack exchange，提问作者Iyad Elwy