sunary / ank

Python data pipeline

ANK - Python Streaming system

Overview:

Python Streaming system, REST-API and Schedule task using queue message(rabbitMQ, zeroMQ, kafka) Each processor is a chain, put it together to make a pipeline.

Support chain model:

• 1 - 1:

  processor - processor
  

• 1 - n:

            / processor
  processor - processor
            \ processor
  

• n - 1:

  processor \
  processor - processor
  processor /
  

• n - n:

  processor - processor    processor - processor    processor - processor
  processor - processor    processor /              processor < 
  processor - processor    processor - processor    processor - processor
  

• join message:

  message1 \
  message2 - [message1, message2, message3]
  message3 /
  

• split message:

                                 / message1
  [message1, message2, message3] - message2
                                 \ message3
  

Requirements:

• Python 2.x
• Python 3.x

How to use:

• Install:

  • From github:
    • clone this project: git clone [email protected]:sunary/ank.git
    • install: python setup.py install
  • From pypi:
    • pip install ank

• Create new Service:

  $ ank -c TestService
  $ cd TestService

  then continue using below commands

• Edit app (processor.py):

  • Example:

  from ank.components.pipe_app import PipeApp
  
  class ExampleApp(PipeApp):
  
      def start(self):
          for i in range(100):
              self.chain_process(i)
  
      def process(self, message=None):
          '''
          Args:
              message: {'content': (*) 'content of message',
                        'flags': (list|tuple) 'define next process will be use'}
                            raise TypeError if you don't declare this in return of before branching-processor
                            if 'flags' == [True, True]: process both in next branching-processors
                            if 'flags' == [True, False]: process 1st processor in next branching-processors
                            if 'flags' == [False, True]: process 2nd processor in next branching-processors
                            if 'flags' == [False, False]: no processor, stop chain
                        is None: stop chain
          '''
          return message['content'] + 1

• Edit services and chains (services.yml):

  • Syntax:

  services:
    Object1:
      - class: module.name.ClassName
      - arguments: [$Object, %variable%] 
    
    AnkChain2:
      - class: ank.components.module_name.XXXApps
      - arguments: ~
      
  chains:
    - Object1
    - AnkChain2

  • Example:

  services:
    StartApp:
      class: processor.StartApp
      arguments: [$MongoClient, $Redis, '%batch_size%']
  
    Mongodb:
      class: pymongo.MongoClient
      arguments: ['%mongo_host%', '%mongo_port%']
  
    Redis:
      class: redis.client.StrictRedis
      arguments: ['%redis_host%', '%redis_port%']
  
    OtherApp:
      class: processor.OtherApp
      arguments: ~
  
    LogApp:
      class: components.log_app.LogApp
      arguments: ~
  
  chains:
    - StartApp
    - LogApp
    - OtherApp
    - [OddApp, EvenApp] # will be processed depend on key `flags`
    - OtherApp

  ANK will read top-down chains, find correspond services and get parameters from settings.yml.

• Generate and edit setting (settings.yml):

  $ ank create -c BaseApp

  • Example:

  parameters:
    mongo_host: localhost
    mongo_port: 27017
    
    redis_host: localhost
    redis_port: 6379
    
    queue_uri: 'amqp://admin:admin@localhost:5672/'
    exchange_name: InputExchange
    routing_key: ExchangeToQueue
    
    batch_size: 100

  Help you create settings template file. Just rename from _settings.yml to settings.yml and fill in values.

• Build Service (create docker image) and run:

  $ ank build
  $ docker run --entrypoint /bin/sh $docker_image_id

• Generate processor (_processor.py):

  $ ank gen_processor -fs settings.yml

• Generate setting (_setting.yml):

  $ ank gen_setting -fs _setting.yml

• Run Service:

  from ank import program_loader
  program_loader.main(file_setting='settings.yml')

  or

  $ ank run -fs settings.yml

Base Apps:

• PipeApp: Pipeline App.
• APIApp: REST-API interface using flask.
• ScheduleApp: Using crondtab-time format to set schedule.

Component Apps:

• LogApp: Log every messages.
• JoinApp: Join messages into one.
• SplitApp: Split message.
• ---Consumer: Get message from queue.
• ---Producer: Push message to queue.

TODO

• base apps: Base, API, Schedule
• supported chain: join, split
• supported consumer/producer: redis, kafka, zmq, rabbitmq
• supported callback success/fail
• generate: processor.py, settings.yml
• management/admin
• build/deploy
• sample: PipeApp, ScheduleApp, APIApp
• test

ank's Issues

build as dependency service

support success/fail callback

otherwise execute process, I need execute success/fail callback in ank/chain_process.py

support more queue message