brod

brod lets you produce messages to the Kafka distributed publish/subscribe messaging service. It started as a fork of pykafka (https://github.com/dsully/pykafka), but became a total rewrite as we needed to add many features.

It's named after Max Brod, Franz Kafka's friend and supporter.

Requirements

You need to have access to your Kafka instance and be able to connect through TCP. You can obtain a copy and instructions on how to setup kafka at http://incubator.apache.org/kafka/

Installation

easy_install brod

Note: the zc.zk package has a dependency on Python Zoo Keeper bindings which are not included during it's installation. They can be installed with easy_install zc-zookeeper-static see the zc.zk documentation for more information http://pypi.python.org/pypi/zc.zk/0.5.

Usage

Sending a simple message

import brod
kafka = brod.Kafka(host='localhost')
kafka.produce('test-topic', 'Hello World')

Sending a sequence of messages

import brod
kafka = brod.Kafka(host='localhost')
kafka.produce('test-topic', ['Hello', 'World'])

Consuming messages one by one

import brod
kafka = brod.Kafka(host='localhost')
for offset, message in kafka.fetch('test-topic', offset=0):
    print message

Using a ZooKeeper-based consumer

from brod.zk import ZKConsumer

consumer = ZKConsumer('localhost:2181', 'test-consumer-group', 'test-topic', autocommit=True)

# Polls forever
for msg_set in consumer.poll(poll_interval=1):
    for offset, msg in msg_set:
        print offset, msg_set.broker_partition, msg

Nonblocking Tornado client support

import time
import tornado.ioloop
import tornado.web

from brod import LATEST_OFFSET
from brod.nonblocking import KafkaTornado

class MainHandler(tornado.web.RequestHandler):
    def initialize(self, kafka, topic):
        self.kafka = kafka
        self.topic = topic

    def post(self):
        data = self.get_argument('data')
        self.kafka.produce(self.topic, data)
    
    @tornado.web.asynchronous
    def get(self):
        brod.offsets(self.topic, LATEST_OFFSET, max_offsets=2, 
            callback=self._on_offset)

    def _on_offset(self, offsets):
        offset = offsets[-1] # Get the second to latest offset
        brod.fetch(self.topic, offset, callback=self._on_fetch)

    def _on_fetch(self, messages):
        for offset, message in messages:
            self.write('{0}: {1}'.format(offset, message))
        self.finish()


kafka = KafkaTornado()

application = tornado.web.Application([
    (r'/', MainHandler, {
        'kafka': kafka,
        'topic': 'hello-world'
    }),
])

if __name__ == '__main__':
    parse_command_line()
    application.listen(8888)
    tornado.ioloop.IOLoop.instance().start()

Contact:

Please use the GitHub issues: https://github.com/datadog/brod/issues

• Forked from https://github.com/dsully/pykafka which was inspired by Alejandro Crosa's kafka-rb: https://github.com/acrosa/kafka-rb