AWS signature version 4 signing process for the python requests module
Python
Switch branches/tags
Nothing to show
Clone or download

README.md

Build Status

AWS Signature Version 4 Signing Process with python requests

This package allows you to authenticate to AWS with Amazon's signature version 4 signing process with the python requests library.

Tested with both python 2.7 and 3.4.

Installation

pip install aws-requests-auth

Motivation

This code came about because Amazon's Elasticsearch Service does not currently support VPC. This authentication class allows us to talk to our Elasticsearch cluster via IAM.

Conceivably, the authentication class is flexible enough to be used with any AWS service that supports the signature version 4 signing process. However, I've only tested it with the Elasticsearch service.

Usage

import requests
from aws_requests_auth.aws_auth import AWSRequestsAuth

# let's talk to our AWS Elasticsearch cluster
auth = AWSRequestsAuth(aws_access_key='YOURKEY',
                       aws_secret_access_key='YOURSECRET',
                       aws_host='search-service-foobar.us-east-1.es.amazonaws.com',
                       aws_region='us-east-1',
                       aws_service='es')

response = requests.get('http://search-service-foobar.us-east-1.es.amazonaws.com',
                        auth=auth)
print response.content

{
  "status" : 200,
  "name" : "Stevie Hunter",
  "cluster_name" : "elasticsearch",
  "version" : {
    "number" : "1.5.2",
    etc....
  },
  "tagline" : "You Know, for Search"
}

elasticsearch-py Client Usage Example

It's possible to inject the AWSRequestsAuth class directly into the elasticsearch-py library so you can talk to your Amazon AWS cluster directly through the elasticsearch-py client.

from aws_requests_auth.aws_auth import AWSRequestsAuth
from elasticsearch import Elasticsearch, RequestsHttpConnection

es_host = 'search-service-foobar.us-east-1.es.amazonaws.com'
auth = AWSRequestsAuth(aws_access_key='YOURKEY',
                       aws_secret_access_key='YOURSECRET',
                       aws_host=es_host,
                       aws_region='us-east-1',
                       aws_service='es')

# use the requests connection_class and pass in our custom auth class
es_client = Elasticsearch(host=es_host,
                          port=80,
                          connection_class=RequestsHttpConnection,
                          http_auth=auth)
print es_client.info()

Temporary Security Credentials

If you are using AWS STS to grant temporary access to your Elasticsearch resource, you can use the aws_token keyword argument to include your credentials in AWSRequestsAuth. See issue #9 and PR #11 for additional details.

AWS Lambda Quickstart Example

If you are using an AWS lamba to talk to your Elasticsearch cluster and you've assigned an IAM role to your lambda function that allows the lambda to communicate with your Elasticserach cluster, you can instantiate an instance of AWSRequestsAuth by reading your credentials from environment variables:

import os
from aws_requests_auth.aws_auth import AWSRequestsAuth

def lambda_handler(event, context):
    auth = AWSRequestsAuth(aws_access_key=os.environ['AWS_ACCESS_KEY_ID'],
                           aws_secret_access_key=os.environ['AWS_SECRET_ACCESS_KEY'],
                           aws_token=os.environ['AWS_SESSION_TOKEN'],
                           aws_host='search-service-foobar.us-east-1.es.amazonaws.com',
                           aws_region='us-east-1',
                           aws_service='es')
    print 'My lambda finished executing'                           

'AWS_ACCESS_KEY_ID', 'AWS_SECRET_ACCESS_KEY', 'AWS_SESSION_TOKEN' are reserved environment variables in AWS lambdas.

Using Boto To Automatically Gather AWS Credentials

botocore (the core functionality of boto3) is not a strict requirement of aws-requests-auth, but we do provide some convenience methods if you'd like to use botocore to automatically retrieve your AWS credentials for you.

botocore can dynamically pull AWS credentials from environment variables, AWS config files, IAM Role, and other locations. Dynamic credential fetching can come in handy if you need to run a program leveraging aws-requests-auth in several places where you may authenticate in different manners. For example, you may rely on a .aws/credentials file when running on your local machine, but use an IAM role when running your program in a docker container in the cloud.

To take advantage of these conveniences, and help you authenticate wherever botocore finds AWS credentials, you can import the boto_utils file and initialize BotoAWSRequestsAuth as follows:

# note that this line will fail if you do not have botocore installed
# botocore installation instructions available here:
# https://boto3.readthedocs.io/en/latest/guide/quickstart.html#installation
from aws_requests_auth.boto_utils import BotoAWSRequestsAuth

auth = BotoAWSRequestsAuth(aws_host='search-service-foobar.us-east-1.es.amazonaws.com',
                           aws_region='us-east-1',
                           aws_service='es')

Credentials are only accessed when needed at runtime, and they will be refreshed using the underlying methods in botocore if needed.