You may not need heavyweights like Celery or RQ. Maintaing an AMQP server might be overkill. There's a simpler, easier way to distribute work.

Redset provides simple, generic sorted sets backed by Redis that can be used to coordinate distributed systems and parcel out work. Unlike more common distribution libraries like Celery or RQ, redset avoids duplicate work for certain use-cases by maintaining a set of tasks instead of a list or queue. And it does so with a dead-simple interface that feels natural for Python.

Redset is currently used in the wild to do things like

• maintain a high-throughput work queue of streaming updates to be processed
• power a multi-producer, multi-consumer scraping architecture that won't do the same work twice
• maintain a simple, cross-process set of "seen" items that each have a TTL
• schedule non-duplicate, periodic polling of analytics on social services

• No worker daemons to run, no heavy AMQP service to monitor
• Safe for multiple producers and consumers
• Seamless, simple use with Python objects using serializers
• Zero dependencies: you provide an object that implements the redis.client.Redis interface, we don't ask any questions.
• Simple, easy-to-read implementation
• Mimics Python's native set interface
• Battle-tested
• Python 3 compatible

import json
import redis

from redset import TimeSortedSet

r = redis.Redis()
ss = TimeSortedSet(r, 'important_json_biz', serializer=json)

ss.add({'foo': 'bar1'})
ss.add({'foo': 'bar2'})

ss.add({'foo': 'bar3'})
ss.add({'foo': 'bar3'})

len(ss)
# 3


# ...some other process A

ss.peek()
# {'foo': 'bar1'}

ss.pop()
# {'foo': 'bar1'}


# ...meanwhile in process B (at exactly same time as A's pop)

ss.take(2)
# [{'foo': 'bar2'}, {'foo': 'bar3'}]

Here

This software was developed at Percolate, where we use it for all sorts of things that involve maintaining synchronized sets of things across process boundaries. A common use-case is to use redset for coordinating time-sensitive tasks where duplicate requests may be generated.

Redset is unopinionated about how consumers look or behave. Want to have a plain 'ol Python consumer managed by supervisor? Fine. Want to be able to pop off items from within a celery job? Great. Redset has no say in where or how it is used: mechanism, not policy.

redset.SortedSet and its subclasses can be instantiated with a few paramters that are notable.

Since Redis only stores primitive numbers and strings, handling serialization and deserialization is a key part of making redset set usage simple in Python.

A serializer instance can be passed (which adheres to the redset.interfaces.Serializer interface, though it need not subclass it) to automatically handle packing and unpacking items managed with redset.

A callable that specifies how to generate a score for items being added can also be passed to SortedSet's constructor as scorer. This callable takes one argument, which is the item object (i.e. the item before serialization) to be "scored."

• redis-py
• celery
• RQ