v3: enhance job manager
-
enforce task_id
on every created task -
audit complete removal of loop.create_task
-
add failure mode semantics to job manager api (default is log_exception
, others would beraise_exception
andretry
) -
OPTIONALLY add database-powered recovery/resuming capabilities to job manager (this is complex, and we will likely get a bug-ridden subset of an asyncio celery, and at this rate, we might just decide on making a separate library, for the good of the world, considering aiojobs doesn't do this)
considering the last point, this requires some discussion across backend: should we give focus to a separate job queue library that is asyncio-aware and provides what we want in one, like resuming, etc. the inspiration is from oban. this will likely be postponed to after this MR, but this is a very interesting idea IMO.
closes #124 (closed)
Edited by luna