Error handling is one of the most important features a framework can offer in order to have a great DX.
Some of the implementations can be a really bad DX 👀, or they can be really fun, like in Celery.
You see, there are two things that can go wrong when you send a task to a Celery worker:

  1. Connection issues with the broker and Message Queue.
  2. Exceptions raised on the worker.

Connection issues

The first issue can be solved by defining retry and retry_policy, say like this (kudos to this article):

from tasks.celery import app


It does not protect you against exceptions that the task raises - this is only useful for connection failures. Yet, I find this a really good first line of defense against transport error.
Of course, further info about this can be found in the Celery docs.

Task issues

This kind of errors can be solved in two ways:

  1. By calling self.retry() upon a task failure, manually.
  2. Automatic retry for known exceptions. This is useful when you know that an ephermal error might occur during task execution.
    A basic implementation of this looks something like this:
           retry_kwargs={'max_retries': 5})
    def refresh_timeline(user):
     return twitter.refresh_timeline(user)

Golden Tip 🎫 It is helpful to set CELERY_ACKS_LATE = True in your global celery app settings. This means that the messages will be acknowledged after the task has been executed, and not before, which is the default behavior of Celery.
This way, if a worker crashes, the message will still be in the queue.

This post is adapted from an answer I wrote on stackoverflow.