Checklist

• I have verified that the issue exists against the master branch of Celery.
• This has already been asked to the discussions forum first.
• I have read the relevant section in the
  contribution guide
  on reporting bugs.
• I have checked the issues list
  for similar or identical bug reports.
• I have checked the pull requests list
  for existing proposed fixes.
• I have checked the commit log
  to find out if the bug was already fixed in the master branch.
• I have included all related issues and possible duplicate issues
  in this issue (If there are none, check this box anyway).

Mandatory Debugging Information

• I have included the output of celery -A proj report in the issue.
  (if you are not able to do this, then at least specify the Celery
  version affected).
• I have verified that the issue exists against the master branch of Celery.
• I have included the contents of pip freeze in the issue.
• I have included all the versions of all the external dependencies required
  to reproduce this bug.

Optional Debugging Information

• I have tried reproducing the issue on more than one Python version
  and/or implementation.
• I have tried reproducing the issue on more than one message broker and/or
  result backend.
• I have tried reproducing the issue on more than one version of the message
  broker and/or result backend.
• I have tried reproducing the issue on more than one operating system.
• I have tried reproducing the issue on more than one workers pool.
• I have tried reproducing the issue with autoscaling, retries,
  ETA/Countdown & rate limits disabled.
• I have tried reproducing the issue after downgrading
  and/or upgrading Celery and its dependencies.

Related Issues and Possible Duplicates

Related Issues

• None

Possible Duplicates

• None

Environment & Settings

Celery version: 4.4.7, but bug likely present since 4.4.3

Steps to Reproduce

When using a database result backend (our case is MySQL, SQLAlchemy, pymysql but should not depend) in conjunction with database_engine_options that contains a setting unsupported by NullPool, Celery worker may refuse to start or even throw a fatal error in certain conditions with the following message:

TypeError: Invalid argument(s) 'max_overflow' sent to create_engine(), using configuration MySQLDialect_pymysql/NullPool/Engine.  Please check that the keyword arguments are appropriate for this combination of components.

We hit this after upgrading from 4.3.0 to 4.4.7 but the bug (if it's considered one) is still present in master. I believe the cause to be 4695f8f .

What happens is, that in get_engine() we fall into the else branch which tries to instantiate NullPool with all arguments from database_engine_options except those that start with pool. However one of those is max_overflow, known to be only supported in QueuePool, which then causes the above mentioned error.

I understand the motivation behind the commit that introduced this behavior, but I'm not sure how to incorporate max_overflow (or potentially other QueuePool settings) into database_engine_options without the risk of hitting this error. It's that more dangerous, that it doesn't happen every time. The else branch is only executed (and the exception thrown) if we're not in a forked process (assuming preforking mode), which for example happens when a worker starts and receives a task that's expired. In that case, the worker obviously won't fork itself and will discard the task inline, attempting to store the result to the database by using the hardcoded NullPool.

I'm not sure whether we should specifically allow certain arguments instead of filtering out pool* ones, or rather be more tolerant in NullPool to accept **kwargs for all the unsupported ones (EDIT: not an option since NullPool is a foreign object), or whether this is even considered a bug at all. We circumvented the risk of this error by not using max_overflow but then, even using pool* options only relies on the internal implementation of create_engine() that silently filters them out - so I'm tempted to completely throw away database_engine_options. Another option is opting to always use NullPool which in prefork worker pool mode seems like a waste of resources.

Required Dependencies

• Minimal Python Version: N/A or Unknown
• Minimal Kombu Version: N/A or Unknown
• Minimal Broker Version: N/A or Unknown
• Minimal Celery Version: 4.4.3

Minimally Reproducible Test Case

• Use a database as a result backend, e.g. result_backend = "db+mysql+pymysql://celery:password@mariadb/celery_task_results"
• Configure database_engine_options = {"max_overflow": 1} for example (any setting not supported by NullPool constructor will do)
• Seed the task queue with a task that won't trigger a new worker subprocess fork, if using prefork worker pool. For example a task with an overdue expiration time.
• The worker will reject the task while avoiding forking, but then instantiate NullPool with an unsupported kwarg and raise an exception.

Expected Behavior

The worker handles all tasks without crashing.

Actual Behavior

The worker crashes on tasks that don't cause forking, provided that the conditions described above hold.