Skip to main content

Asynchronous Python client for AWS Kinesis

Project description


Python Kinesis Client library utilising asyncio

Based on Kinesis-Python project by Evan Borgstrom but with asyncio magic

The problem with Kinesis-Python is that all the data ends up in a single thread and being checkpointed from there - so despite having many processes, the client is clogged by checkpointing. Besides, it checkpoints every single record and this is not configurable.

This client is based on aioboto3 library and uses Python 3.6+ async methods.


import asyncio
from async_kinesis_client.kinesis_consumer import AsyncKinesisConsumer

async def read_stream():

    # This is a coroutine that reads all the records from a shard
    async def read_records(shard_reader):
        async for records in shard_reader.get_records():
            for r in records:
                print('Shard: {}; Record: {}'.format(shard_reader.shard_id, r))

    consumer = AsyncKinesisConsumer(

    # consumer will yield existing shards and will continue yielding
    # new shards if re-sharding happens             
    async for shard_reader in consumer.get_shard_readers():
        print('Got shard reader for shard id: {}'.format(shard_reader.shard_id))


AsyncShardReader and AsyncKinesisConsumer can be stopped from parallel coroutine by calling stop() method, consumer will stop all shard readers in that case. If you want to be notified of shard closing, catch ShardClosedException while reading records

AsyncShardReader exposes property millis_behind_latest which could be useful for determining application performance.

AsyncKinesisConsumer has following configuration methods:

set_checkpoint_interval(records) - how many records to skip before checkpointing

set_lock_duration(time) - how many seconds to hold the lock. Consumer would attempt to refresh the lock before that time

set_reader_sleep_time(time) - how long should shard reader wait (in seconds, fractions possible) if it did not receive any records from Kinesis stream

set_checkpoint_callback(coro) - set callback coroutine to be called before checkpointing next batch of records. Coroutine arguments: ShardId, SequenceNumber

Producer is rather trivial:

from async_kinesis_client.kinesis_producer import AsyncKinesisProducer

# ...

async def write_stream():
    producer = AsyncKinesisProducer(

    await producer.put_record(
        partition_key='string',     # optional, if none, default time-based key is used
        explicit_hash_key='string'  # optional

Sending multiple records at once:

from async_kinesis_client.kinesis_producer import AsyncKinesisProducer

# ...

async def write_stream():
    producer = AsyncKinesisProducer(

    records = [
            'Data': b'bytes',
            'PartitionKey': 'string',   # optional, if none, default time-based key is used
            'ExplicitHashKey': 'string' # optional

    response = await producer.put_records(

    # See boto3 docs for response structure:

AWS authentication. For testing outside AWS cloud, especially when Mutil-Factor Authentication is in use I find following snippet extremely useful:

import os
import aioboto3
from botocore import credentials
from aiobotocore import AioSession

    working_dir = os.path.join(os.path.expanduser('~'), '.aws/cli/cache')
    session = AioSession(profile=os.environ.get('AWS_PROFILE'))
    provider = session.get_component('credential_provider').get_provider('assume-role')
    provider.cache = credentials.JSONFileCache(working_dir)

This allows re-using cached session token after completing any aws command under awsudo, all you need is to set AWS_PROFILE environment variable.

Currently library still not tested enough for different network events. Use it on your own risk, you've been warned.

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

async-kinesis-client-0.2.14.tar.gz (14.0 kB view hashes)

Uploaded source

Built Distribution

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page