Celery is brittle. If you stop a celery service, odds are, it will either not stop at all or leave stray child processes running.

It is not transactional, at least not with a MongoDB back-end. If you stop a celery service, any tasks that were in progress may or may not complete, but none of them will be picked up again when you resume.

No prioritization of queues or tasks. It is possible to set up separate services that handle different queues, but that's a PITA and there's still no guaranteed order of processing.

Celery is a PITA and over-complicated for the problem it solves.

Do you know why pyres relies on itty? Seems like kind of an odd dependency...

The pyres web interface (where you can view job status) is built using itty.

edit: looks like they're using flask now, I don't see itty being used anywhere. They also split the web interface (resweb) into a separate project, so pyres doesn't depend on a web framework now.

Ah ok, thanks. I guess the documentation is a little outdated then. Thanks for the suggestion, I'll check pyres out when I get to that point in my project.