sqlalchemy-celery-beat

A Scheduler Based Sqlalchemy for Celery.

NOTE: This project was originally developed by AngelLiang to use sqlalchemy as the database scheduler for Flask or FastAPI, like django-celery-beat for django. I am trying to continue on his work and maintain a working solution.

Prerequisites

• Python 3
• celery >= 5.0
• sqlalchemy >= 1.4

First you must install celery and sqlalchemy, and celery should be >=5.0

$ pip install sqlalchemy celery

Installing

Install from PyPi:

$ pip install sqlalchemy-celery-beat

Install from source by cloning this repository:

$ git clone git@github.com:farahats9/sqlalchemy-celery-beat.git
$ cd sqlalchemy-celery-beat
$ python setup.py install

Usage

After you have installed sqlalchemy_celery_beat, you can easily start with following steps:

This is a demo for exmaple, you can check the code in examples directory

1. start celery worker

   $ celery -A tasks worker -l info
   

2. start the celery beat with DatabaseScheduler as scheduler:

   $ celery -A tasks beat -S sqlalchemy_celery_beat.schedulers:DatabaseScheduler -l info
   

   you can also use the shorthand argument -S sqlalchemy

Description

After the celery beat is started, by default it create a sqlite database(schedule.db) in current folder. You can use SQLiteStudio.exe to inspect it.

Sample from the PeriodicTask model's table

When you want to update scheduler, you can update the data in schedule.db. But sqlalchemy_celery_beat don't update the scheduler immediately. Then you shoule be change the first column's last_update field in the celery_periodic_task_changed to now datetime. Finally the celery beat will update scheduler at next wake-up time.

Database Configuration

You can configure sqlalchemy db uri when you configure the celery, example as:

from celery import Celery

celery = Celery('tasks')

beat_dburi = 'sqlite:///schedule.db'

celery.conf.update(
    {
        'beat_dburi': beat_dburi,
        'beat_schema': None  # you can make the scheduler tables under different schema (tested for postgresql, not available in sqlite)
    }
)

Also, you can use MySQL or PostgreSQL.

# MySQL: `pip install mysql-connector`
beat_dburi = 'mysql+mysqlconnector://root:root@127.0.0.1:3306/celery-schedule'

# PostgreSQL: `pip install psycopg2`
beat_dburi = 'postgresql+psycopg2://postgres:postgres@127.0.0.1:5432/celery-schedule'

Passing arguments to SQLAlchemy engine creation

You can pass arguments using the beat_engine_options keyword in the config dictionary, for example let's make the engine use echo=True to show verbose ouptut:

celery.conf.update(
    {
        'beat_dburi': beat_dburi,
        'beat_engine_options': {
            'echo': True
        },
        ...
    }
)

You can use this to pass any options required by your DB driver, for more information about what options you can use check the SQLAlchemy docs.

Example Code 1

View examples/base/tasks.py for details.

Run Worker in console 1

$ cd examples/base

# Celery < 5.0
$ celery worker -A tasks:celery -l info

# Celery >= 5.0
$ celery -A tasks:celery worker -l info

Run Beat in console 2

$ cd examples/base

# Celery < 5.0
$ celery beat -A tasks:celery -S tasks:DatabaseScheduler -l info

# Celery >= 5.0
$ celery -A tasks:celery beat -S tasks:DatabaseScheduler -l info

Example Code 2

Example creating interval-based periodic task

To create a periodic task executing at an interval you must first create the interval object:

>>> from sqlalchemy_celery_beat.models import PeriodicTask, IntervalSchedule, Period
>>> from sqlalchemy_celery_beat.session import SessionManager
>>> from celeryconfig import beat_dburi
>>> session_manager = SessionManager()
>>> session = session_manager.session_factory(beat_dburi)

# executes every 10 seconds.
>>> schedule = session.query(IntervalSchedule).filter_by(every=10, period=Period.SECONDS).first()
>>> if not schedule:
...     schedule = IntervalSchedule(every=10, period=Period.SECONDS)
...     session.add(schedule)
...     session.commit()

That's all the fields you need: a period type and the frequency.

You can choose between a specific set of periods:

• Period.DAYS
• Period.HOURS
• Period.MINUTES
• Period.SECONDS
• Period.MICROSECONDS

note:

If you have multiple periodic tasks executing every 10 seconds,
then they should all point to the same schedule object.

Now that we have defined the schedule object, we can create the periodic task entry:

    >>> task = PeriodicTask(
    ...     schedule_model=schedule,            # we created this above.
    ...     name='Importing contacts',          # simply describes this periodic task.
    ...     task='proj.tasks.import_contacts',  # name of task.
    ... )
    >>> session.add(task)
    >>> session.commit()

Note that this is a very basic example, you can also specify the arguments and keyword arguments used to execute the task, the queue to send it to[*], and set an expiry time.

Here's an example specifying the arguments, note how JSON serialization is required:

>>> import json
>>> from datetime import datetime, timedelta, UTC

>>> periodic_task = PeriodicTask(
...     schedule_model=schedule,                  # we created this above.
...     name='Importing contacts',          # simply describes this periodic task.
...     task='proj.tasks.import_contacts',  # name of task.
...     args=json.dumps(['arg1', 'arg2']),
...     kwargs=json.dumps({
...        'be_careful': True,
...     }),
...     expires=datetime.now(UTC) + timedelta(seconds=30)
... )
... session.add(periodic_task)
... session.commit()

Example creating crontab-based periodic task

A crontab schedule has the fields: minute, hour, day_of_week, day_of_month and month_of_year, so if you want the equivalent of a 30 * * * * (execute every 30 minutes) crontab entry you specify:

>>> from sqlalchemy_celery_beat.models import PeriodicTask, CrontabSchedule
>>> schedule = CrontabSchedule(
...     minute='30',
...     hour='*',
...     day_of_week='*',
...     day_of_month='*',
...     month_of_year='*',
...     timezone='UTC',
... )

The crontab schedule is linked to a specific timezone using the 'timezone' input parameter.

Then to create a periodic task using this schedule, use the same approach as the interval-based periodic task earlier in this document, the schedule_model is a generic foreign-key implementation which makes things very easy and efficient:

>>> periodic_task = PeriodicTask(
...     schedule_model=schedule,
...     name='Importing contacts',
...     task='proj.tasks.import_contacts',
... )

What the previous code actually do is this:

>>> periodic_task = PeriodicTask(
...     schedule_id=schedule.id,
...     discriminator=schedule.discriminator,
...     name='Importing contacts',
...     task='proj.tasks.import_contacts',
... )

So when you can use discriminator + schedule_id or use the convenient property schedule_model and it will populate them for you behind the scenes.

Temporarily disable a periodic task

You can use the enabled flag to temporarily disable a periodic task:

>>> periodic_task.enabled = False
>>> session.add(periodic_task)
>>> session.commit()

If you are using a bulk operation to update or delete multiple tasks at the same time, the changes won't be noticed by the scheduler until you do PeriodicTaskChanged.update_changed() or .update_from_session()

example:

from sqlalchemy_celery_beat.models import PeriodicTaskChanged
from sqlalchemy_celery_beat.session import SessionManager, session_cleanup

session_manager = SessionManager()
session = session_manager.session_factory(beat_dburi)

with session_cleanup(session):
    stmt = update(PeriodicTask).where(PeriodicTask.name == 'task-123').values(enabled=False)

    session.execute(stmt)  # using execute causes no orm event to fire, changes are in the database but the schduler has no idea
    session.commit()

    PeriodicTaskChanged.update_from_session(session)
    # now scheduler reloads the tasks and all is good

This is not needed when you are updating a specific object using session.add(task) because it will trigger the after_update, after_delete or after_insert events.

Example running periodic tasks

The periodic tasks still need 'workers' to execute them. So make sure the default Celery package is installed. (If not installed, please follow the installation instructions here: https://github.com/celery/celery)

Both the worker and beat services need to be running at the same time.

1. Start a Celery worker service (specify your project name):

   $ celery -A [project-name] worker --loglevel=info
   

2. As a separate process, start the beat service (specify the scheduler):

   $ celery -A [project-name] beat -l info --scheduler sqlalchemy_celery_beat.schedulers:DatabaseScheduler
   

Working on adding the following features

• ✅ Add ClockedSchedule model
• ✅ Implement a generic foreign key
• ✅ More robust attribute validation on models
• ✅ Add Tests
• Add more examples
• Support for Async drivers like asyncpg and psycopg3 async mode
• Use Alembic migrations

Any help with the tasks above or feedback is appreciated 🙂

Acknowledgments

• django-celery-beat
• celerybeatredis
• celery
• SQLAlchemy