Daemonizing Celery Beat with systemd

Photo by Jon Tyson on Unsplash

Recently, I came to a situation where I had to write a script that will run periodically and do some jobs. Previously I used cron jobs for these types of jobs. Now I am using Celery for this purpose, and yes, it’s a charm!

There are multiple ways to demonize celery workers or celerybeat. I user systemd as I use Ubuntu 16.04 and systemd is built-in.

Let’s think of a small scenario, we have to run an SQL query that will insert a new entity in every minute(that is not a real-world scenario, though I am using it just for example).

First, we make a directory named simple-celery and place a file named db_update.py inside it. Then paste the contents as below.

We can check if the scheduler is working properly by opening two terminals and provide the below two commands in them respectively

celery -A db_update beat --loglevel=info

and

celery -A db_update worker --loglevel=info

This is op for development, but, in production, we need to daemonize these to run them in background. To do so, we will follow the steps

1) We will create a /etc/default/celeryd configuration file.

2) Now, create another file for the worker /etc/systemd/system/celeryd.service with sudo privilege.

And a file for celerybeat /etc/systemd/system/celerybeat.service with sudo privilege.

3) Now, we will create log and pid directories.

sudo mkdir /var/log/celery /var/run/celery
sudo chown sajid:sajid /var/log/celery /var/run/celery 

N.B.: If you hard reboot your pc, you have to chwon again.

4) After that, we need to reload systemctl daemon. Remember that, we should reload this every time we make any change to the service definition file.

sudo systemctl daemon-reload

5) To enable the service to start at boot, we will run.

sudo systemctl enable celeryd
sudo systemctl enable celerybeat

6) we can now start the services.

sudo systemctl start celeryd
sudo systemctl start celerybeat

7) To verify that everything is ok, we can check the log files

cat /var/log/celery/beat.log
cat /var/log/celery/worker1.log

8) Now, we are all set up, and here comes the part of monitoring. There are command line options to do that, but we will not frequently use that in production. Instead, we will use a visual Real-time Celery web-monitor, Flower. Let’s install this via pip .

pip install flower

After installation, go to terminal and type

celery -A db_update flower

After that, visit http://localhost:5555 to get the visual monitor.

N.B.: carefully set the user and the group, the paths of CELERY_BIN, EnvironmentFile, WorkingDirectory, otherwise, we will not be able to run them.

References

1. http://docs.celeryproject.org/en/latest/index.html
2. https://www.linode.com/docs/development/python/task-queue-celery-rabbitmq/
3. https://medium.com/the-andela-way/timed-periodic-tasks-in-celery-58c99ecf3f80