Confluent's Python Client for Apache KafkaTM
confluent-kafka-python provides a high-level Producer, Consumer and AdminClient compatible with all
Apache KafkaTM brokers >= v0.8, Confluent Cloud
and the Confluent Platform. The client is:
-
Reliable - It's a wrapper around librdkafka (provided automatically via binary wheels) which is widely deployed in a diverse set of production scenarios. It's tested using the same set of system tests as the Java client and more. It's supported by Confluent.
-
Performant - Performance is a key design consideration. Maximum throughput is on par with the Java client for larger message sizes (where the overhead of the Python interpreter has less impact). Latency is on par with the Java client.
-
Future proof - Confluent, founded by the
creators of Kafka, is building a streaming platform
with Apache Kafka at its core. It's high priority for us that client features keep
pace with core Apache Kafka and components of the Confluent Platform.
See the API documentation for more info.
License: Apache License v2.0
Usage
Below are some examples of typical usage. For more examples, see the examples directory or the confluentinc/examples github repo for a Confluent Cloud example.
Producer
from confluent_kafka import Producer
p = Producer({'bootstrap.servers': 'mybroker1,mybroker2'})
def delivery_report(err, msg):
""" Called once for each message produced to indicate delivery result.
Triggered by poll() or flush(). """
if err is not None:
print('Message delivery failed: {}'.format(err))
else:
print('Message delivered to {} [{}]'.format(msg.topic(), msg.partition()))
for data in some_data_source:
# Trigger any available delivery report callbacks from previous produce() calls
p.poll(0)
# Asynchronously produce a message, the delivery report callback
# will be triggered from poll() above, or flush() below, when the message has
# been successfully delivered or failed permanently.
p.produce('mytopic', data.encode('utf-8'), callback=delivery_report)
# Wait for any outstanding messages to be delivered and delivery report
# callbacks to be triggered.
p.flush()
High-level Consumer
from confluent_kafka import Consumer, KafkaError
c = Consumer({
'bootstrap.servers': 'mybroker',
'group.id': 'mygroup',
'auto.offset.reset': 'earliest'
})
c.subscribe(['mytopic'])
while True:
msg = c.poll(1.0)
if msg is None:
continue
if msg.error():
print("Consumer error: {}".format(msg.error()))
continue
print('Received message: {}'.format(msg.value().decode('utf-8')))
c.close()
AvroProducer
from confluent_kafka import avro
from confluent_kafka.avro import AvroProducer
value_schema_str = """
{
"namespace": "my.test",
"name": "value",
"type": "record",
"fields" : [
{
"name" : "name",
"type" : "string"
}
]
}
"""
key_schema_str = """
{
"namespace": "my.test",
"name": "key",
"type": "record",
"fields" : [
{
"name" : "name",
"type" : "string"
}
]
}
"""
value_schema = avro.loads(value_schema_str)
key_schema = avro.loads(key_schema_str)
value = {"name": "Value"}
key = {"name": "Key"}
def delivery_report(err, msg):
""" Called once for each message produced to indicate delivery result.
Triggered by poll() or flush(). """
if err is not None:
print('Message delivery failed: {}'.format(err))
else:
print('Message delivered to {} [{}]'.format(msg.topic(), msg.partition()))
avroProducer = AvroProducer({
'bootstrap.servers': 'mybroker,mybroker2',
'on_delivery': delivery_report,
'schema.registry.url': 'http://schema_registry_host:port'
}, default_key_schema=key_schema, default_value_schema=value_schema)
avroProducer.produce(topic='my_topic', value=value, key=key)
avroProducer.flush()
AvroConsumer
from confluent_kafka import KafkaError
from confluent_kafka.avro import AvroConsumer
from confluent_kafka.avro.serializer import SerializerError
c = AvroConsumer({
'bootstrap.servers': 'mybroker,mybroker2',
'group.id': 'groupid',
'schema.registry.url': 'http://127.0.0.1:8081'})
c.subscribe(['my_topic'])
while True:
try:
msg = c.poll(10)
except SerializerError as e:
print("Message deserialization failed for {}: {}".format(msg, e))
break
if msg is None:
continue
if msg.error():
print("AvroConsumer error: {}".format(msg.error()))
continue
print(msg.value())
c.close()
AdminClient
Create topics:
from confluent_kafka.admin import AdminClient, NewTopic
a = AdminClient({'bootstrap.servers': 'mybroker'})
new_topics = [NewTopic(topic, num_partitions=3, replication_factor=1) for topic in ["topic1", "topic2"]]
# Note: In a multi-cluster production scenario, it is more typical to use a replication_factor of 3 for durability.
# Call create_topics to asynchronously create topics. A dict
# of <topic,future> is returned.
fs = a.create_topics(new_topics)
# Wait for each operation to finish.
for topic, f in fs.items():
try:
f.result() # The result itself is None
print("Topic {} created".format(topic))
except Exception as e:
print("Failed to create topic {}: {}".format(topic, e))
Thread Safety
The Producer, Consumer and AdminClient are all thread safe.
Install
Install self-contained binary wheels
$ pip install confluent-kafka
NOTE: The pre-built Linux wheels do NOT contain SASL Kerberos/GSSAPI support.
If you need SASL Kerberos/GSSAPI support you must install librdkafka and
its dependencies using the repositories below and then build
confluent-kafka using the command in the "Install from
source from PyPi" section below.
Install AvroProducer and AvroConsumer
$ pip install "confluent-kafka[avro]"
Install from source from PyPi
(requires librdkafka + dependencies to be installed separately):
$ pip install --no-binary :all: confluent-kafka
For source install, see Prerequisites below.
Broker Compatibility
The Python client (as well as the underlying C library librdkafka) supports
all broker versions >= 0.8.
But due to the nature of the Kafka protocol in broker versions 0.8 and 0.9 it
is not safe for a client to assume what protocol version is actually supported
by the broker, thus you will need to hint the Python client what protocol
version it may use. This is done through two configuration settings:
broker.version.fallback=YOUR_BROKER_VERSION(default 0.9.0.1)api.version.request=true, false(default true)
When using a Kafka 0.10 broker or later you don't need to do anything
(api.version.request=true is the default).
If you use Kafka broker 0.9 or 0.8 you must set
api.version.request=false and set
broker.version.fallback to your broker version,
e.g broker.version.fallback=0.9.0.1.
More info here:
https://github.com/edenhill/librdkafka/wiki/Broker-version-compatibility
SSL certificates
If you're connecting to a Kafka cluster through SSL you will need to configure
the client with 'security.protocol': 'SSL' (or 'SASL_SSL' if SASL
authentication is used).
The client will use CA certificates to verify the broker's certificate.
The embedded OpenSSL library will look for CA certificates in /usr/lib/ssl/certs/
or /usr/lib/ssl/cacert.pem. CA certificates are typically provided by the
Linux distribution's ca-certificates package which needs to be installed
through apt, yum, et.al.
If your system stores CA certificates in another location you will need to
configure the client with 'ssl.ca.location': '/path/to/cacert.pem'.
Alternatively, the CA certificates can be provided by the certifi
Python package. To use certifi, add an import certifi line and configure the
client's CA location with 'ssl.ca.location': certifi.where().
Prerequisites
- Python >= 2.7 or Python 3.x
- librdkafka >= 0.11.5 (latest release is embedded in wheels)
librdkafka is embedded in the macosx manylinux wheels, for other platforms, SASL Kerberos/GSSAPI support or
when a specific version of librdkafka is desired, following these guidelines:
-
For Debian/Ubuntu based systems, add this APT repo and then do
sudo apt-get install librdkafka-dev python-dev:
http://docs.confluent.io/current/installation.html#installation-apt -
For RedHat and RPM-based distros, add this YUM repo and then do
sudo yum install librdkafka-devel python-devel:
http://docs.confluent.io/current/installation.html#rpm-packages-via-yum -
On OSX, use homebrew and do
brew install librdkafka
Developer Notes
Instructions on building and testing confluent-kafka-python can be found here.