A short Lambda Function the can be sent CloudWatch Logs (in the case Flow Logs) and send them to Kinesis Firehose for storage in S3. A full writeup can be found on my site http://mlapida.com/thoughts/exporting-cloudwatch-logs-to-s3-lambda

CloudWatchLogsKinesisFirehose-Lambda.py

import boto3
import logging
import json
import gzip
from StringIO import StringIO

logger = logging.getLogger()
logger.setLevel(logging.INFO)

client = boto3.client('firehose')

def lambda_handler(event, context):
    
    #capture the CloudWatch log data
    outEvent = str(event['awslogs']['data'])
    
    #decode and unzip the log data
    outEvent = gzip.GzipFile(fileobj=StringIO(outEvent.decode('base64','strict'))).read()
    
    #convert the log data from JSON into a dictionary
    cleanEvent = json.loads(outEvent)
    
    #initiate a list
    s = []
    
    #set the name of the Kinesis Firehose Stream
    firehoseName = 'FlowLogTest'
    
    #loop through the events line by line
    for t in cleanEvent['logEvents']:
        
        #Transform the data and store it in the "Data" field. 
        p={
            #Fields in FlowLogs - [version, accountid, interfaceid, srcaddr, dstaddr, srcport, dstport, protocol, packets, bytes, start, stop, action, logstatus]
            'Data': str(t['extractedFields']['start']) + "," + str(t['extractedFields']['dstaddr']) + "," + str(t['extractedFields']['srcaddr']) + "," + str(t['extractedFields']['packets'])+"\n"
        }
        
        #write the data to our list
        s.insert(len(s),p)
        
        #limit of 500 records per batch. Break it up if you have to.
        if len(s) > 499:
            #send the response to Firehose in bulk
            SendToFireHose(firehoseName, s)
            
            #Empty the list
            s = []
    
    #when done, send the response to Firehose in bulk
    if len(s) > 0:
        SendToFireHose(firehoseName, s)

#function to send record to Kinesis Firehose
def SendToFireHose(streamName, records):
    response = client.put_record_batch(
        DeliveryStreamName = streamName,
        Records=records
    )
    
    #log the number of data points written to Kinesis
    print "Wrote the following records to Firehose: " + str(len(records))

Thanks @mlapida for the example. I left everything as a json for the logs since logstash and splunk know how to parse json log lines
s.append({'Data': "{d}\n".format(d=json.dumps(t['extractedFields']))})

this was extremely helpful for figuring out how to properly parse the CloudWatch flow log, thanks!

@mlapida - Thank you for this example. I am trying to demo this but running into some issues. Namely, I am getting the following error and given I'm a noob with Python, I can't seem to figure out how to go about resolve this error condition.

I'm trying to use this example not for VPC Flow logs, but I'm streaming Java application logs from Beanstalk to CloudWatch and from there, I'm streaming the CloudWatch log group to the Lambda function via Stream Subscription.

Was hoping someone could help me understand the issue and point me in the right direction.

---

initial_value must be unicode or None, not str: TypeError
Traceback (most recent call last):
File "/var/task/lambda_function.py", line 18, in lambda_handler
outEvent = gzip.GzipFile(fileobj=StringIO(outEvent.decode('base64','strict'))).read()
TypeError: initial_value must be unicode or None, not str

For anyone who wants to use this snippet with Python 3, the following decoding and unzipping code worked for me:

    # capture the CloudWatch log data
    out_event = str(event['awslogs']['data'])

    # decode and unzip the log data
    decoded = base64.b64decode(out_event)
    file = io.BytesIO(decoded)
    out_event = gzip.GzipFile(fileobj=file).read()

    # convert the log data from JSON into a dictionary
    clean_event = json.loads(out_event)

@cshanes please share the final code