Connect to external Amazon S3 buckets

Every language in Cloudera Machine Learning has libraries available for uploading to and downloading from Amazon S3.

To work with external S3 buckets in Python, do the following:

  • Add your Amazon Web Services access keys to your project's environment variables as AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY.
  • Add your Ozone S3 gateway to the environment variables as OZONE_S3_GATEWAY.


# Install Boto to the project
!pip3 install boto3

# Make sure below environment variables are set
# ozone s3 gateway : os.environ['OZONE_S3_GATEWAY']
# s3 keys from os.environ['AWS_ACCESS_KEY_ID'] and os.environ['AWS_SECRET_ACCESS_KEY']

import os
import boto3

# Use Boto to connect to S3 and get a list of objects from a bucket
conn = boto3.session.Session()

s3g = os.environ['OZONE_S3_GATEWAY']
access_key = os.environ['AWS_ACCESS_KEY_ID']
secret_key = os.environ['AWS_SECRET_ACCESS_KEY']

s3_client = conn.client(

test_bucket = 'testozones3'
all_buckets = s3_client.list_buckets()
print(f"All S3 Buckets are {[i['Name'] for i in all_buckets['Buckets']]}")

s3_client.put_object(Bucket=test_bucket, Key='')

all_objs = s3_client.list_objects(Bucket=test_bucket)
print(f"All keys in {bucket_name} are {[i['Key']for i in all_objs['Contents']]}")

s3_client.get_object(Bucket=test_bucket, Key='')

ssl = "true" if s3g.startswith("https") else "false"
s3a_path = f"s3a://{test_bucket}/"

hadoop_opts = f"-Dfs.s3a.access.key='{access_key}' -Dfs.s3a.secret.key='{secret_key}' -Dfs.s3a.endpoint='{s3g}' -Dfs.s3a.connection.ssl.enabled={ssl}"

!hdfs dfs {hadoop_opts} -ls "s3a://{test_bucket}/"