The Kinesis Aggregation/Deaggregation Modules for Python provide the ability to do in-memory aggregation and deaggregation of standard Kinesis user records using the Kinesis Aggregated Record Format to allow for more efficient transmission of records.
The Python record aggregation/deaggregation modules are available on the Python Package Index (PyPI) as aws_kinesis_agg. You can install it via the pip command line tool:
pip install aws_kinesis_agg
Alternately, you can simply copy the aws_kinesis_agg module from this repository and use it directly with the caveat that the Google protobuf module must also be available (if you install via pip, this dependency will be handled for you).
The aggregator.py module contains Python classes that allow you to aggregate records using the Kinesis Aggregated Record Format. Using record aggregation improves throughput and reduces costs when writing producer applications that publish data to Amazon Kinesis.
Caution - this module is only suitable for low-value messages which are processed in aggregate. Do not use Kinesis Aggregation for data which is sensitive or where every message must be delivered, and where the KCL (including with AWS Lambda) is used for processing. DATA LOSS CAN OCCUR.
The record aggregation module provides a simple interface for creating protocol buffers encoded data in a producer application. The aws_kinesis_agg module provides methods for efficiently packing individual records into larger aggregated records, and deaggregating large records into a set of 'real' user records.
When using aggregation, you create a managing class which helps you to target the correct Kinesis Shard, and then provide a partition key, raw data and (optionally) an explicit hash key for each record. You can choose to either provide a callback function that will be invoked when a fully-packed aggregated record is available or you can add records and check byte sizes or number of records until the aggregated record is suitably full. You're guaranteed that any aggregated record returned from the RecordAggregator object will fit within a single PutRecord request to Kinesis. As you produce records in your producer application, you will aggregate them using a base RecordAggregator object, which provides methods to do both iterative aggregation and callback-based aggregation.
There are two ways to create aggregated user records. The first is to use a raw RecordAggregator, which can aggregate messages which are targeted for a single Shard, or use the AggregationManager to aggregate messages which may span Shards. From version 1.2.0, we highly recommend the use of AggregationManager to limit any exposure to data loss.
You can construct a raw RecordAggregator class with:
import aws_kinesis_agg as agg
kinesis_aggregator = agg.RecordAggregator()
The iterative aggregation method involves adding records one at a time to the RecordAggregator and checking the response to determine when a full aggregated record is available. The add_user_record method returns None when there is room for more records in the existing aggregated record and returns an AggRecord object when a full aggregated record is available for transmission.
for rec in records:
result = kinesis_aggregator.add_user_record(rec.PartitionKey, rec.Data, rec.ExplicitHashKey)
if result:
#Send the result to Kinesis
To use callback-based aggregation, you must register a callback via the on_record_complete method. As you add individual records to the RecordAggregator object, you will receive a callback (on a separate thread) whenever a new fully-packed aggregated record is available.
def my_callback(agg_record):
#Send the record to Kinesis
...
kinesis_aggregator.on_record_complete(my_callback)
for rec in records:
kinesis_aggregator.add_user_record(rec.PartitionKey, rec.Data, rec.ExplicitHashKey)
This repository includes an example script that uses the record aggregation module aggregator.py to aggregate records and transmit them to Amazon Kinesis using callback-based aggregation. You can find this example functionality in the file kinesis_publisher.py, which you can use as a template for your own applications to to easily build and transmit encoded data.
The example below assumes you are running Python version 2.7.x and also requires you to install and configure the boto3 module. You can install boto3 via pip install boto3 or any other normal Python install mechanism. To configure the example to be able to publish to your Kinesis stream, make sure you follow the instructions in the Boto3 Configuration Guide. The example below has been stripped down for brevity, but you can still find the full working version at kinesis_publisher.py. The abridged example is:
import boto3
import aws_kinesis_agg.aggregator
kinesis_client = None
def send_record(agg_record):
global kinesis_client
pk, ehk, data = agg_record.get_contents()
kinesis_client.put_record(StreamName='MyKinesisStreamName',
Data=data,
PartitionKey=pk,
ExplicitHashKey=ehk)
if __name__ == '__main__':
kinesis_client = boto3.client('kinesis', region_name='us-west-2')
kinesis_agg = aws_kinesis_agg.aggregator.RecordAggregator()
kinesis_agg.on_record_complete(send_record)
for i in range(0,1024):
pk, ehk, data = get_record(...)
kinesis_agg.add_user_record(pk, data, ehk)
#Clear out any remaining records that didn't trigger a callback yet
send_record(kinesis_agg.clear_and_get())
The deaggregator.py module contains Python classes that allow you to deaggregate records that were transmitted using the Kinesis Aggregated Record Format, including those transmitted by the Kinesis Producer Library. This library will allow you to deaggregate aggregated records in any Python environment, including AWS Lambda.
The record deaggregation module provides a simple interface for working with Kinesis aggregated message data in a consumer application. The aws_kinesis_agg module provides methods for both bulk and generator-based processing.
When using deaggregation, you provide an aggregated Kinesis Record and get back multiple Kinesis User Records. If a Kinesis Record that is provided is not an aggregated Kinesis record, that's perfectly fine - you'll just get a single record output from the single record input. A Kinesis user record which is returned from deaggregation looks like:
{
'eventVersion' : String - The version number of the Kinesis event used
'eventID' : String - The unique ID of this Kinesis event
'kinesis' :
{
'partitionKey' : String - The Partition Key provided when the record was submitted
'explicitHashKey' : String - The hash value used to explicitly determine the shard the data record is assigned to by overriding the partition key hash (or None if absent)
'data' : String - The original data transmitted by the producer (base64 encoded)
'kinesisSchemaVersion' : String - The version number of the Kinesis message schema used,
'sequenceNumber' : BigInt - The sequence number assigned to the record on submission to Kinesis
'subSequenceNumber' : Int - The sub-sequence number for the User Record in the aggregated record, if aggregation was in use by the producer
'aggregated' : Boolean - Always True for a user record extracted from a Kinesis aggregated record
},
'invokeIdentityArn' : String - The ARN of the IAM user used to invoke this Lambda function
'eventName' : String - Always "aws:kinesis:record" for a Kinesis record
'eventSourceARN' : String - The ARN of the source Kinesis stream
'eventSource' : String - Always "aws:kinesis" for a Kinesis record
'awsRegion' : String - The name of the source region for the event (e.g. "us-east-1")
}
To get started, import the aws_kinesis_agg module:
import aws_kinesis_agg
Next, when you receive a Kinesis Record in your consumer application, you will extract the user records using the deaggregation methods available in the aws_kinesis_agg module.
IMPORTANT: The deaggregation methods available in the aws_kinesis_agg module expect input records in the same dictionary-based format that they are normally received in from AWS Lambda. See the Programming Model for Authoring Lambda Functions in Python section of the AWS documentation for more details.
The bulk conversion method of deaggregation takes in a list of Kinesis Records, extracts all the aggregated user records and accumulates them into a list. Any records that are passed in to this method that are not Kinesis aggregated records will be returned unchanged. The method returns a list of Kinesis user records in the same format as they are normally delivered by Lambda's Kinesis event handler.
user_records = deaggregate_records(raw_kinesis_records)
The generator-based conversion method of deaggregation uses a Python generator function to extract user records from a raw Kinesis Record one at a time in an iterative fashion. Any records that are passed in to this method that are not Kinesis aggregated records will be returned unchanged. For example, you could use this code to iterate through each deaggregated record:
for record in iter_deaggregate_records(raw_kinesis_records):
#Process each record
pass
This module includes two example AWS Lambda function in the file lambda_function.py that give you the ability to easily build new functions to process Kinesis aggregated data via AWS Lambda.
from __future__ import print_function
from aws_kinesis_agg.deaggregator import deaggregate_records
import base64
def lambda_bulk_handler(event, context):
raw_kinesis_records = event['Records']
#Deaggregate all records in one call
user_records = deaggregate_records(raw_kinesis_records)
#Iterate through deaggregated records
for record in user_records:
# Kinesis data in Python Lambdas is base64 encoded
payload = base64.b64decode(record['kinesis']['data'])
#TODO: Process each record
return 'Successfully processed {} records.'.format(len(user_records))
from __future__ import print_function
from aws_kinesis_agg.deaggregator import iter_deaggregate_records
import base64
def lambda_generator_handler(event, context):
raw_kinesis_records = event['Records']
record_count = 0
#Deaggregate all records using a generator function
for record in iter_deaggregate_records(raw_kinesis_records):
# Kinesis data in Python Lambdas is base64 encoded
payload = base64.b64decode(record['kinesis']['data'])
#TODO: Process each record
record_count += 1
return 'Successfully processed {} records.'.format(record_count)
One easy way to get started processing Kinesis data is to use AWS Lambda. By building on top of the existing lambda_function.py module in this repository, you can take advantage of Kinesis message deaggregation features without having to write boilerplate code.
When you're ready to make a build and upload to AWS Lambda, you have two choices:
- Follow the existing instructions at Creating a Deployment Package (Python)
OR
- At the root of this Python project, you can find a sample build file called make_lambda_build.py. This file is a platform-agnostic build script that will take the existing Python project in this demo and package it in a single build file called
python_lambda_build.zipthat you can upload directly to AWS Lambda.
In order to use the build script, make sure that the python pip tool is available on your command line. If you have other pip dependencies, make sure to add them to the PIP_DEPENDENCIES list at the top of the make_lambda_build.py. Then run this command:
python make_lambda_build.py --output_dir <mydir>
The build script will create a new folder called build, copy all the Python source files, download any necessary dependencies via pip and create the file <mydir>/python_lambda_build.zip that you can deploy to AWS Lambda.
If you choose to make your own Python zip file to deploy to AWS Lambda, be aware that the Google protobuf module normally relies on using a Python pth setting to make the root google module importable. If you see an error in your AWS Lambda logs such as:
"Unable to import module 'lambda_function': No module named google.protobuf"
You can go into the google module folder (the same folder containing the protobuf folder) and make an empty file called __init__.py. Once you rezip everything and redeploy, this should fix the error above.
NOTE: If you used the provided make_lambda_build.py script, this issue is already handled for you.
Copyright Amazon.com, Inc. or its affiliates. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.