Airflow DAG Loop - 如何使每次迭代顺序而不是并行

Question

我有一个像这样的 Apache Airflow DAG：

DAG_NAME='my_dag'
sections = ["0", "1", "2", "3"]

with DAG(DAG_NAME, default_args=default_args, schedule_interval=None) as dag:

        for s in sections:
            a = DummyOperator(task_id=f"section_{s}_start")
            b = SubDagOperator(task_id=f"init_{s}_subdag",subdag=init_section(DAG_NAME,f"init_{s}_subdag", default_args))
            c = SubDagOperator(task_id=f"process_{s}_subdag", subdag=process_section(DAG_NAME,f"process_{s}_subdag", default_args))
            d = SubDagOperator(task_id=f"update_{s}_subdag", subdag=update_section(DAG_NAME,f"update_{s}_subdag", default_args))
            e = DummyOperator(task_id=f"section_{s}_end")
            a>>b>>c>>d>>e

这段代码会像这样呈现我的任务

如何使任务顺序为：

section_0_start>>init_0_subdag>>process_0_subdag>>update_0_subdag>>section_0_end section_0_end>>section_1_start section_1_start>>init_1_subdag>>process_1_subdag>>update_1_subdag>>section_1_end

.....

依此类推，从第 0 节开始，以第 3 节任务结束

谢谢

Answer 1

像这样修改for-loop：

    previous_e = None
    for s in sections:
        a = ...
        ...
        e = ...
        if previous_e:
            previous_e >> a
        a>>b>>c>>d>>e
        previous_e = e

Airflow DAG Loop - 如何使每次迭代顺序而不是并行

Airflow DAG Loop - How to make each iteration sequential instead of parallel

directed-acyclic-graphs

airflow