Delete and retry tasks through Django Admin #92
Conversation
|
|
||
|
|
||
| def retry(modeladmin: "DBTaskResultAdmin", request, queryset): | ||
| rows = queryset.update(status=ResultStatus.NEW) |
There was a problem hiding this comment.
@RealOrangeOne I'm not sure if it's a good idea to reuse DBTaskResult rows; maybe we should be creating new rows instead. What do you think?
There was a problem hiding this comment.
For my use case (processing a bunch of documents), I find it useful to be able to retry failed tasks, until there are none left, fixing errors in the process. If we were to create new rows, then I'd have to stop the queue, retry the tasks, delete the old tasks, and then restart the queue.
There was a problem hiding this comment.
Perhaps "Reenqueue" which updates the task, and a "Duplicate" which duplicates.
There was a problem hiding this comment.
Then we shall have both! I've updated the PR as you suggest, with a small test to make sure that duplicate tasks actually work.
mgax
force-pushed
the
feature/admin-delete-and-retry
branch
from
faae1df to
a608fc1
Compare
|
@RealOrangeOne Any plans of merging this soon? We would need this. :) |
|
@Alwinator it's a draft PR with failing tests and conflicts. It's not as simple as just "merging" it I'm afraid. Once it's ready, it'll absolutely be merged. |
|
Thank you! @mgax Any plans to finish your request? :) |
|
Sure, I'd forgotten about this, I'll see about rebasing and fixing the tests. |
mgax
force-pushed
the
feature/admin-delete-and-retry
branch
from
9125545 to
dfb443e
Compare
|
|
||
| object.__setattr__(task_result, "_exception_class", exception_class) | ||
| object.__setattr__(task_result, "_traceback", self.traceback or None) | ||
| object.__setattr__(task_result, "_return_value", self.return_value) |
There was a problem hiding this comment.
When I rebased my changes, I noticed that this assertion was failing:
I'm pretty sure this used to work, and it looks like the database backend is not fetching return values. Did I find a bug? I'm happy to write a specific test for checking that return_value is set on the task result.
|
I've rebased the branch. I'm not sure why the CI is failing now. I've narrowed it down to this change: mgax@36d3d38. It looks like the failures are flaky, and I'm guessing that the test somehow causes other tests to fail. But I can't reproduce it locally; for me (MacOS 15.1.1, Python 3.13, Django 5.0) |
| ] | ||
| ) | ||
|
|
||
| def duplicate(self) -> Self: |
There was a problem hiding this comment.
Suggestion: This API feels like one better suited for the TaskResult, rather than being backend specific.
This comment has been minimized.
This comment has been minimized.
|
Sharing a quick snippet if someone is trying to reenqueue stale from django_tasks.backends.database.models import DBTaskResult, ResultStatus
DBTaskResult.objects.filter(status=ResultStatus.RUNNING).update(status=ResultStatus.NEW)Update: Bonus management command! # python manage.py reset_running_tasks
from django.core.management.base import BaseCommand
from django_tasks.backends.database.models import DBTaskResult, ResultStatus
class Command(BaseCommand):
help = 'Resets all DBTaskResult objects with status RUNNING to status NEW.'
def handle(self, *args, **options):
updated = DBTaskResult.objects.filter(status=ResultStatus.RUNNING).update(status=ResultStatus.NEW)
self.stdout.write(self.style.SUCCESS(f'Successfully reset {updated} running tasks to NEW.')) |
This comment has been minimized.
This comment has been minimized.
5 participants
I found it super useful to be able to delete and to reschedule tasks through the Django Admin.
Update – this PR now adds two admin actions:
statusof the selected tasks toNEW, so they will be executed again.NEW, with no result or started/finished timestamps.