How to work correctly airflow schedule_interval - airflow

Question

I want to try to use Airflow instead of Cron.
But schedule_interval doesn't work as I expected.
I wrote the python code like below.
And in my understanding, Airflow should have ran on "2016/03/30 8:15:00" but it didn't work at that time.
If I changed it like this "'schedule_interval': timedelta(minutes = 5)", it worked correctly, I think.
The "notice_slack.sh" is just to call slack api to my channels.
# -*- coding: utf-8 -*-
from __future__ import absolute_import, unicode_literals
import os
from airflow.operators import BashOperator
from airflow.models import DAG
from datetime import datetime, timedelta
args = {
'owner': 'airflow',
'depends_on_past': False,
'start_date': datetime(2016, 3, 29, 8, 15),
}
dag = DAG(
dag_id='notice_slack',
default_args=args,
schedule_interval="#daily",
dagrun_timeout=timedelta(minutes=1))
# cmd file name
CMD = '/tmp/notice_slack.sh'
run_this = BashOperator(
task_id='run_transport', bash_command=CMD, dag=dag)
I want to run some of my scripts at specific time every day like this cron setting.
15 08 * * * bash /tmp/notice_slack.sh
I have read the document Scheduling & Triggers, and I know it's a little bit different cron.
So I attempt to arrange at "start_date" and "schedule_interval" settings.
Does anyone know what should I do ?
airflow version
INFO - Using executor LocalExecutor
v1.7.0
amazon-linux-ami/2015.09-release-notes

Answer 1

Try this:
# -*- coding: utf-8 -*-
from __future__ import absolute_import, unicode_literals
import os
from airflow.operators import BashOperator
from airflow.models import DAG
from datetime import datetime, timedelta
args = {
'owner': 'airflow',
'depends_on_past': False,
'start_date': datetime(2016, 3, 29),
}
dag = DAG(
dag_id='notice_slack',
default_args=args,
schedule_interval="15 08 * * *",
dagrun_timeout=timedelta(minutes=1))
# cmd file name
CMD = 'bash /tmp/notice_slack.sh'
run_this = BashOperator(
task_id='run_transport', bash_command=CMD, dag=dag)
start_date (datetime) – The start_date for the task, determines the execution_date for the first task instance. The best practice is to have the start_date rounded to your DAG’s schedule_interval.
schedule_interval (datetime.timedelta or dateutil.relativedelta.relativedelta or str that acts as a cron expression) – Defines how often that DAG runs, this timedelta object gets added to your latest task instance’s execution_date to figure out the next schedule.
Simply configuring the schedule_interval and bash_command as the same in your cron setting is okay.

Answer 2

Airflow will start your DAG when the 2016/03/30 8:15:00 + schedule interval (daily) is passed. So your DAG will run on 2016/03/31 8:15:00.
You can check the Airflow FAQ

Answer 3

First, your start date should be in the past -
Instead of 'start_date': datetime(2016, 3, 29, 8, 15)
Would you try 'start_date': datetime(2016, 2, 29, 8, 15)
and apply 'catchup':False to prevent backfills - unless this was something you wanted to do.
From Airflow documentation -
The Airflow scheduler triggers the task soon after the start_date + schedule_interval is passed.
The schedule interval can be supplied as a cron -
If you want to run it everyday at 8:15 AM, the expression would be - *'15 8 * * '
If you want to run it only on Oct 31st at 8:15 AM, the expression would be - *'15 8 31 10 '
To supply this, 'schedule_inteval':'15 8 * * *' in your Dag property
You can figure this out more from https://crontab.guru/
Alternatively, there are Airflow presets -
If any of these meet your requirements, it would be simply, 'schedule_interval':'#hourly'
Lastly, you can also apply the schedule as python timedelta object e.g. for 12 PM
'schedule_interval': timedelta(hours=12)

Answer 4

With the example you've given #daily will run your job after it passes midnight. You might try changing it either to timedelta(days=1) which is relative to your fixed start_date that includes 08:15.
Or you could use a cron spec for the schedule_interval='15 08 * * *' in which case any start date prior to 8:15 on the day BEFORE the day you wanted the first run would work.
Note that depends_on_past: False is already the default, and you may have confused its behavior with catchup=false in the DAG parameters, which would avoid making past runs for time between the start date and now where the DAG schedule interval would have run.