Celery hangs when an exception occurs inside a group of chains with redis as a broker

[emanuelmd]

Checklist

• This has already been asked to the discussion group 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 verified that the issue exists against the master branch of Celery.
• 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

• N/A

Related Issues

• None

Possible Duplicates

• I've skimmed the issues list and the internet and it seems there's definitely something going on with this kind of chain composition. It seems the issue has been dating since at least 2013 and I can reproduce it on Celery 5.0.1 & 4.7.x

Environment & Settings

• Celery 5.0.1 & 4.7.x
• I included the report in the demo repo

Steps to Reproduce

• Follow the instructions inside the demo repo

Required Dependencies

• Minimal Python Version: Tested with 3.7.0, 3.8.0 & 3.8.5
• Minimal Celery Version: 4.7.x

Python Packages

• Included in demo repo

Other Dependencies

• N/a

Minimally Reproducible Test Case

https://github.com/emanuelmd/freezelery

Expected Behavior

• Exceptions should be propagated from chains that contain groups

Actual Behavior

• The calling code hangs

Notes

Happy to help tracking this down. Let me know if you need any other information

[maybe-sybr]

This appears to be related to chords misbehaving again - possibly an edge case we missed in #6354 or something nearby? I have a more MRTC here which appears to cover the gist of your linked repository:

import atexit

import celery

app = celery.Celery("app", broker="redis://", backend="redis://")

@app.task
def nop(*_):
    pass

@app.task
def die(*_):
    raise RuntimeError

g = celery.group((nop.s() | die.s() | celery.group(nop.s())) for _ in range(3))
r = g.delay()

atexit.register(lambda: print(r.get(disable_sync_subtasks=False, timeout=1)))

You can run celery -A app worker -l DEBUG and then ^C out after the three runtime errors appear. The atexit sync get will timeout if the final element of each chain in a group (ie. exception in chord header which is magically constructed from the group with a preceding chain element).

Removing the external group so it's just a single chain allows the result object to be joined without timing out, so my gut feel is that perhaps a task ID or some completion event for the encapsulating group isn't being handled properly. Maybe the promoted chord doesn't retain some task ID which is needed to finalise the chain for an encapsulating canvas signature type?