pytest-kafka 0.8.1

Pytest fixture factories for Zookeeper, Kafka server and Kafka consumer. Read the API docs.

from pathlib import Path
from pytest_kafka import (
    make_zookeeper_process, make_kafka_server, make_kafka_consumer,
    terminate,
)

ROOT = Path(__file__).parent
KAFKA_SCRIPTS = ROOT / 'kafka/bin/'
KAFKA_BIN = str(KAFKA_SCRIPTS / 'kafka-server-start.sh')
ZOOKEEPER_BIN = str(KAFKA_SCRIPTS / 'zookeeper-server-start.sh')

# You can pass a custom teardown function (or parametrise ours). Just don't call it `teardown`
# or Pytest will interpret it as a module-scoped teardown function.
teardown_fn = partial(terminate, signal_fn=Popen.kill)
zookeeper_proc = make_zookeeper_process(ZOOKEEPER_BIN, teardown_fn=teardown_fn)
kafka_server = make_kafka_server(KAFKA_BIN, 'zookeeper_proc', teardown_fn=teardown_fn)
kafka_consumer = make_kafka_consumer(
    'kafka_server', seek_to_beginning=True, kafka_topics=['topic'])

This creates 3 fixtures:

1. zookeeper_proc - Zookeeper process

2. kafka_server - Kafka process

3. kafka_consumer - usable kafka.KafkaConsumer instance

ZOOKEEPER_BIN and KAFKA_BIN are paths to launch scripts in your Kafka distribution. Check this project’s setup.py to see a way of installing Kafka for development.

It is advised to pass seek_to_beginning=True because otherwise some messages may not be captured by the consumer. This requires knowing the topics upfront because without topics there’s no partitions to seek.

Kafka server is known to take a couple of seconds to terminate gracefully. You probably don’t need that, so you can pass partial(terminate, signal_fn=Popen.kill) to make it killed with SIGKILL and waited for afterwards.

It’s possible to create multiple Kafka fixtures forming a cluster by passing the same Zookeeper fixture to them. For an example, check the tests.

Session-scoped fixtures are also available. Consult the test suite.

pip install -e .[dev]
./pytest_kafka/install.py  # will install kafka to ./kafka