How to schedule tasks with different start_date in one dag?

[pkexcellent]

Hi, I want to schedule some tasks with different start_date in one dag, like this:

dag = DAG(dag_id = 'myjob', default_args=default_args)
#in default_args, 'start_date' is datetime(2015, 9, 1)
t1 = BashOperator(
     task_id = 't1',
     bash_command = "date",
     dag = dag) 

t2 = BashOperator(
     task_id = 't2',
     bash_command = "date",
     start_date = datetime(2015, 9, 1, 0, 30),
     dag = dag) 

t1.set_downstream(t2)

now my system date is 2015-09-19,
I use airflow scheduler -d myjob
t1 can run success, but t2 cannot be scheduled, why?
Thanks a lot!

[mistercrunch]

start_date really means "first run". Each run (aka execution_date) need to line up perfectly for dependencies to be met.

If you want to delay t2 by 30 minutes the airflow way, you'd use a TimeSensor (as say t3) and have t2 depend on t3. All these tasks would have the same start_date.

Make sense? Where do you think I should change the docs to clarify this?

[pkexcellent]

Thanks, that's really help!
And the start_date is really a confused concept for new users, "start date" and "start time" is really different.
Maybe some more description for the parameter start_date at class BaseOperator(object) in models.py is better:

:param start_date: start date for the task, the scheduler will start from this point in time. 
All the tasks in one dag should have the same start 'time'
:type start_date: datetime

[mistercrunch]

This should make it very clear:
00f3f69