dag-celery-core.py

 def _is_node_rdy(task, graph):
    tasks = session.query(Task).filter(Task.id.in_(list(graph.predecessors(task.id)))).all()
    for dep_task in tasks:
        if not dep_task.celery_task_uid or \
           not AsyncResult(dep_task.celery_task_uid).state == SUCCESS:
            return False
    return True

@app.task(bind=True)
def run(self, workflow_id, cur_task_id=None):
    print('Runnning Workflow {} and Task {}'.format(workflow_id, cur_task_id))
    workflow = session.query(Workflow).filter_by(id=workflow_id).one()
    graph = workflow.execution_graph

    next_task_ids = []
    if cur_task_id:
        task = session.query(Task).get(cur_task_id)
        if not _is_node_rdy(task, graph):
            return

        # do some business logic specific thing here
        # if it's CI workflow, it would be eg. running tests
        _process_task_node(task, self.request.id)

        next_task_ids = list(graph.successors(cur_task_id))
    else:
        next_task_ids = find_entry_point(graph)

    # prematurely update task's status here, because we should have it
    # when the next task is running, and since it's a recursive call
    # to run the next task, the status might not be updated automatically
    # when we run it
    self.update_state(state=SUCCESS)

    for task_id in next_task_ids:
        run.apply_async(
            args=(workflow_id, task_id,),
            queue=QUEUE_NAME
        )