An unified framework to build task queue on Rabbitmq | Kafka | Redis


Keywords
task message queue MQ Pika kafka-python rabbitmq kafka redis rmq producer consumer celery
License
MIT
Install
pip install cocotask==0.2.1.1

Documentation

cocotask

Build task queue on either Rabbitmq, Kafka or Redis! Simple! Easy! and FAST!!!!

Why creating this framework?

  • Most task queues using Rabbitmq/Kafka/Redis are doing the same thing, but there is no unified wrapper. It's a waste of time to write code for different MQs if all you need is a task queue to distribute jobs
  • No need to worry about using Rabbitmq, Kafka or Redis. Just pick one and go! It's simply a few lines of config changes in future if you want to switch to a different underlying system
  • Hide all details for connecting/subscribing/publishing/etc. You can just use the same API for any of Kafka/Rabbitmq/Redis!
  • Minimize the efforts for team members to handcraft the code for exchange/queue handling
  • Team members should focus on how to handle messages
  • There lacks good mananger tool/lib to handle creating multiple consumers (although it's simple)
  • Celery is the original idea, but it doesn't support Kafka and Windows platform. It's also tightly coupled with Python alone.

Test

  1. install rabbitmq or kafka on local machine (either docker or pure rabbitmq).

  2. make sure it's running Python 3.5 above

  3. pip install pika kafka-python redis jsmin

  4. pip install cocotask

  5. now you have cocotask setup on your machine. To test, you can go to the ./test folder

    • Run: python producer_test.py (this will post a string. Code is very simple)
    • In another window, under ./test, run: cocotask ./config.json userworkers TestWorker 4 you'll see the worker starts and process 1 message we just posted.
  6. So the key part is how we use cocotask command tool. the parameters are:

    • config_path: path to config file
    • module_name: module of your customized worker class
    • class_name: the class name of your own worker class (in the example above, it's userworkers.TestWorker, so the module name is userworkers, the class name is TestWorker)
    • number of workers: the total number of customer workers
    • logginglevel(optional): python logging level INFO/DEBUG/etc.
    • modulepath(optional): the relative path to find the module you defined. Default is '.'
  7. Develop your own worker class and try

Development

Build your own worker

from cocotask import CocoBaseWorker

class TestWorker(CocoBaseWorker):

    def process(self, body):
        print(body)

Check userworkers/test_worker.py for reference.

Post a message

from cocotask import CocoProducerManager as pm
import json

with open('config.json', 'r') as f:
    config = json.load(f)

producer = pm.create_instance(config)

producer.connect()
producer.send('12345678')
producer.close()

Switch from RabbitMQ to Kafka or reverse In test/config.json

{
    "MQ_TYPE": "RMQ",  # change this to KAFKA if your underlying MQ is KAFKA

    "RMQ": {
        ...
    },

    "KAFKA": {
        ...
    },

    "REDIS": {
        ...
    }
}

We do support SASL_PLAINTEXT for kafka and simple auth in Redis as in the comments of the config file. Check their website to see how to setup the authentication

You can build your own dictionary object as configuration for sure, as long as it contains the required fields.

That's it. Simple and straightforward.