A transformational Lambda

Datenstreaming mit AWS Kinesis und Lambda

Maksim Pecherskiy

Data Engineer

In chapter 1...

In chapter 2...

In this chapter...

Previous approach

Processing once in S3 vs Lambda transform

Processing once in S3

Uses a lambda function fired on object write in S3
Longer delay until data can be transformed
Raw data is stored before cleaning
Requires Firehose destination to be S3

Processing via Lambda transform

Uses a lambda function fired mid-firehose-stream
Transformation is immediate
Only cleaned data is stored
Allows for other destination in Firehose

Incoming data

record_id	timestamp	vin	lon	lat	speed
939ed1d1-1740-420c-8906-445278573c7f	4:25:06.000	4FTEX4944AK844294	106.9447146	-6.3385652	25
f29a5b3d-d0fa-43c0-9e1a-e2a5cdb8be7a	8:10:47.000	3FTEX1G5XAK844393	108.580681	34.79925	37
ff8e7131-408d-463b-8d07-d016419b0656	20:26:44.000	2LAXX1C8XAK844292	114.392392	36.097577	90

Transformational Lambda

Sample event

{
  "invocationId": "invocationIdExample",
  "deliveryStreamArn": "arn:aws:firehose:us-east-1:458912630:deliverystream/gps-delivery-stream",
  "region": "us-east-1",
  "records": [
    {

      "recordId": "49546986683135544286507457936321625675700192471156785154",

      "approximateArrivalTimestamp": 1495072949453,

      "data": "NjQuMjQyLjg4LjEwIC0gLSBbMDcvTWFyLzIwMDQ6MTY6MTA6MDIgLTA4MDBdICJHRVQgL21haWxtYW4vbGlzdGluZm8vaHNkaXZpc2lvbiBIVFRQLzEuMSIgMjAwIDYyOTE=="

    }
  ]
}

Base64

Processing the data

import base64
from datetime import datetime as dt

def convert_timestamp(record_time_val):

    # Get today's date as string
    today = dt.today().strftime("%Y-%m-%d")

    # Combine today's date with the record's time and make datetime object
    new_ts = dt.strptime(f"{today} {record_time_val}", "%Y-%m-%d %H:%M:%S.%f")

    # Convert the datetime object to a nicely formatted string
    return new_ts.strftime("%Y-%m-%dT%H:%M:%S")

Processing the data

def lambda_handler(event, context):
    output = []

    for record in event['records']:
        payload = base64.b64decode(record['data'])

        payload = payload.decode()

        payload = payload.split(" ")

        payload[1] = convert_timestamp(payload[1])

Processing the data

    for record in event['records']:
        ...
        payload = " ".join(payload)
        payload_enc = base64.b64encode(payload.encode())

        output.append({

          'recordId': record['recordId'],

          'result': 'Ok',

          'data': payload_enc

        })

    return {'records': output}

A review

import base64
def lambda_handler(event, context):

    output = []
    for record in event['records']: # Iterate over the records

        payload = base64.b64decode(record['data']).decode() # Decode the payload

        # Modify it

        payload_enc = base64.b64encode(payload.encode()) # Re-encode it

        output.append({ # Put it in a dictionary
            'recordId': record['recordId'], 'result': 'Ok', 'data': payload_enc,
        })

    return {'records': output}

Creating Lambda in AWS

Script for create lambda

https://drive.google.com/file/d/1fTzIp2mgWt04nVdILc7ClyAMOF1yNxkv/view?usp=sharing

Let's practice!

Datenstreaming mit AWS Kinesis und Lambda