Observability

[RealOrangeOne]

It's important to be able to interrogate your task queues, to understand how long tasks are taking, queue depth, and other helpful metrics for SREs.

Currently, this must be done in a backend-specific manner. However, perhaps it would be helpful to have some generic APIs provided as part of django-tasks to help parse the backend-specific storage into something more easily consumable?

[chrisgrande]

A generic API could be very useful, especially with building out a common Django Admin and exposing these parts if someone wanted to integrate them into their own front end.

[RealOrangeOne]

Working out which metrics to expose will be the challenge. Queue depth is an easy one, but perhaps p99s etc too.

Building the metrics is the easy bit. The hard bit is working out what to expose by default. That's why it's not in scope for the initial version.

[jribbens]

I agree with this - the lack of visibility of what celery is doing is one of its most annoying misfeatures.

[hooverdc]

We could also provide something for OpenTelemetry [0] to hook into for tracing. Or at least monkey-patch.

[0] https://opentelemetry.io/docs/zero-code/python/example/