This module provides low-level protocol support for Apache Kafka as well as high-level consumer and producer classes. Request batching is supported by the protocol as well as broker-aware request routing. Gzip and Snappy compression is also supported for message sets.
Copyright 2013, 2014, 2015 David Arthur under Apache License, v2.0. See LICENSE
Copyright 2014, 2015 Cyan, Inc. under Apache License, v2.0. See LICENSE
Copyright 2015 Ciena Corporation under Apache License, v2.0. See LICENSE
This project began as a port of the kafka-python library to Twisted.
See CONTRIBUTING.md for how to contribute.
- 0.8.0 *
- 0.8.1 *
- 0.8.1.1 *
- 0.8.2.1
- 0.9.0.1
* See Errata, below.
- CPython 2.7.3
- PyPy 2.6.1
[ Note: This code is not meant to be runable. See producer_example
and consumer_example
for runable example code. ]
from afkak.client import KafkaClient
from afkak.consumer import Consumer
from afkak.producer import Producer
from afkak.common import (OFFSET_EARLIEST, PRODUCER_ACK_ALL_REPLICAS,
PRODUCER_ACK_LOCAL_WRITE)
kClient = KafkaClient("localhost:9092")
# To send messages
producer = Producer(kClient)
d1 = producer.send_messages("my-topic", msgs=["some message"])
d2 = producer.send_messages("my-topic", msgs=["takes a list", "of messages"])
# To get confirmations/errors on the sends, add callbacks to the returned deferreds
d1.addCallbacks(handleResponses, handleErrors)
# To wait for acknowledgements
# PRODUCER_ACK_LOCAL_WRITE : server will wait till the data is written to
# a local log before sending response
# [ the default ]
# PRODUCER_ACK_ALL_REPLICAS : server will block until the message is committed
# by all in sync replicas before sending a response
producer = Producer(kClient,
req_acks=Producer.PRODUCER_ACK_LOCAL_WRITE,
ack_timeout=2000)
responseD = producer.send_messages("my-topic", msgs=["message"])
# Using twisted's @inlineCallbacks:
responses = yield responseD
if response:
print(response[0].error)
print(response[0].offset)
# To send messages in batch: You can use a producer with any of the
# partitioners for doing this. The following producer will collect
# messages in batch and send them to Kafka after 20 messages are
# collected or every 60 seconds (whichever comes first). You can
# also batch by number of bytes.
# Notes:
# * If the producer dies before the messages are sent, the caller would
# * not have had the callbacks called on the send_messages() returned
# * deferreds, and so can retry.
# * Calling producer.stop() before the messages are sent will
# errback() the deferred(s) returned from the send_messages call(s)
producer = Producer(kClient, batch_send=True,
batch_send_every_n=20,
batch_send_every_t=60)
responseD1 = producer.send_messages("my-topic", msgs=["message"])
responseD2 = producer.send_messages("my-topic", msgs=["message 2"])
# To consume messages
# define a function which takes a list of messages to process and
# possibly returns a deferred which fires when the processing is
# complete.
def processor_func(consumer, messages):
# Store_Messages_In_Database may return a deferred
result = store_messages_in_database(messages)
# record last processed message
consumer.commit()
return result
the_partition = 3 # Consume only from partition 3.
consumer = Consumer(kClient, "my-topic", the_partition, processor_func)
d = consumer.start(OFFSET_EARLIEST) # Start reading at earliest message
# The deferred returned by consumer.start() will fire when an error
# occurs that can't handled by the consumer, or when consumer.stop()
# is called
yield d
consumer.stop()
kClient.close()
from afkak.client import KafkaClient
from afkak.producer import Producer
from afkak.partitioner import HashedPartitioner, RoundRobinPartitioner
kafka = KafkaClient("localhost:9092")
# Use the HashedPartitioner so that the producer will use the optional key
# argument on send_messages()
producer = Producer(kafka, partitioner_class=HashedPartitioner)
producer.send_messages("my-topic", "key1", ["some message"])
producer.send_messages("my-topic", "key2", ["this method"])
from afkak.client import KafkaClient
kafka = KafkaClient("localhost:9092")
req = ProduceRequest(topic="my-topic", partition=1,
messages=[KafkaProdocol.encode_message("some message")])
resps = afkak.send_produce_request(payloads=[req], fail_on_error=True)
kafka.close()
resps[0].topic # "my-topic"
resps[0].partition # 1
resps[0].error # 0 (hopefully)
resps[0].offset # offset of the first message sent in this request
Afkak releases are available on PyPI.
Because the Afkak dependencies Twisted and python-snappy have binary extension modules you will need to install the Python development headers for the interpreter you wish to use. You'll need all of these to run Afkak's tests:
Debian/Ubuntu: | sudo apt-get install build-essential python-dev pypy-dev libsnappy-dev
|
OS X | brew install python pypy snappy pip install virtualenv |
Then Afkak can be installed with pip as usual:
make toxu
The integration tests will actually start up real local ZooKeeper instance and Kafka brokers, and send messages in using the client.
The makefile knows how to download several versions of Kafka. This will run just the integration tests against Kafka 0.8.1.1
KAFKA_VER=0.8.1.1 make toxi
make toxa
make toxik
Under Kafka 0.8.1 sometimes the test_consumer_integration:TestConsumerIntegration.test_consumer test will fail. This is due to an issue with Kafka where it will report the topic metadata including a leader, but will fail with UnknownTopicOrPartition when an attempt to write messages to the topic at the leader.
Due to the way the Kafka API is versioned, there is no way for the client to know the API version of which the server is capable. Afkak uses the version=1 API for the Offset Commit Request API call. Due to this, Afkak is not compatible with versions older than 0.8.2.1 for offset storage.