[PYTHON] How to store CSV data in Amazon Kinesis Streams with standard input

This article describes a python script that populates Kinesis Streams with standard input from localhost or EC2.

environment

・ Amazon linux ・ Language: python 2.7, shell -Input data format: CSV

The python script to be stored in Kinesis Streams is described below. In this script, it is aggregated and stored at 500 records / second *. .. (Please note that the meaning of aggregation is not aggrigation to one record using KPL)

• If set to 500 or more, an error will occur saying "failed to satisfy constraint: Member must have length less than or equal to 500".

script

buffer_insert.py

import sys
import json
import random
import boto3
import time

def create_json(buffered_data, streamname):
    jdat = {}
    dat = []
    jdat["StreamName"] = streamname      
    for rec in buffered_data :
        dat.append({"Data" : rec, "PartitionKey" : str(random.randint(1,1000))})
    jdat["Records"] = dat
    return jdat

if __name__ == '__main__':

  args = sys.argv
  streamname=args[1]

  cnt = 0
  buf = []

  client = boto3.client('kinesis')
  while 1:
      if len(buf) == 500:
          ret = client.put_records(**create_json(buf,streamname ))
          time.sleep(1)
          print ret
          buf = []
      line = sys.stdin.readline()
      if not line:
          break
      buf.append(line[:-1])

Since the credential information of access-key and secret-key is not set in the above script, please set it in client () if necessary. See the boto3 documentation below for details. http://boto3.readthedocs.io/en/latest/guide/configuration.html

The execution is as follows. -Created by Kinesis_streams_test of Streams ・ Enter the data test.csv

>Run
cat test.csv | python buffer_insert.py kinesis_streams_test

>result
{u'ShardId': u'shardId-000000000000', u'SequenceNumber': u'49571046543942031937439475265378143459129484557577879554'}, {u'ShardId': u'shardId-000000000001', u'SequenceNumber': u'49571046543964332682638005888401204447079899259962654738'}, {u'ShardId': u'shardId-000000000000', u'SequenceNumber': u'49571046543942031937439475265379352384949099186752585730'}, {u'ShardId': u'shardId-000000000000', u'SequenceNumber': u'49571046543942031937439475265380561310768713815927291906'}, {u'ShardId': u'shardId-000000000000', u'SequenceNumber': u'49571046543942031937439475265381770236588328445101998082'}, {u'ShardId': u'shardId-000000000000', u'SequenceNumber': u'49571046543942031937439475265382979162407943074276704258'}, {u'ShardId': u'shardId-000000000001', u'SequenceNumber': u'49571046543964332682638005888402413372899513889137360914'}, {u'ShardId': u'shardId-000000000001', u'SequenceNumber': u'49571046543964332682638005888403622298719128518312067090'}], 'ResponseMetadata': {'HTTPStatusCode': 200, 'RequestId': 'e8d1fc47-17eb-1f03-b9e4-17c6595f9a22'}}

The script to check if it can be stored in Kinesis Streams is described below. (Looking sideways, from the AWS CLI instead of python ... you can of course get it from python.)

get_record.sh

#!/bin/bash

stream_name=$1
shard_array_index=${2:-0}
shard_iterator_type=${3:-TRIM_HORIZON}


shard_id=$(aws kinesis describe-stream --stream-name ${stream_name} --query "StreamDescription.Shards[${shard_array_index}].ShardId" --output text)
echo $shard_id
shard_iterator=$(aws kinesis get-shard-iterator --stream-name ${stream_name} --shard-id ${shard_id} --shard-iterator-type ${shard_iterator_type} --query "ShardIterator" --output text)
echo $shard_iterator
aws kinesis get-records --shard-iterator ${shard_iterator}

Since the put_recorded data is Base64 encoded, it is necessary to perform decoding processing on the consumer side. The AWS CLI does not support base64, so you need to use a Base64 decoder (such as https://www.base64decode.org/).

Next time, I will introduce a script that aggregates data using KPL and stores it in Kinesis Streams.