Connect to S3 data from PySpark

I've solved adding --packages org.apache.hadoop:hadoop-aws:2.7.1 into spark-submit command.

It will download all hadoop missing packages that will allow you to execute spark jobs with S3.

Then in your job you need to set your AWS credentials like:

sc._jsc.hadoopConfiguration().set("fs.s3n.awsAccessKeyId", aws_id)sc._jsc.hadoopConfiguration().set("fs.s3n.awsSecretAccessKey", aws_key)

Other option about setting your credentials is define them into spark/conf/spark-env:

#!/usr/bin/env bashAWS_ACCESS_KEY_ID='xxxx'AWS_SECRET_ACCESS_KEY='xxxx'SPARK_WORKER_CORES=1 # to set the number of cores to use on this machineSPARK_WORKER_MEMORY=1g # to set how much total memory workers have to give executors (e.g. 1000m, 2g)SPARK_EXECUTOR_INSTANCES=10 #, to set the number of worker processes per node

More info:

• How to Run PySpark on AWS
• AWS Credentials

I would suggest going through this link.

In my case, I used Instance profile credentials to access s3 data.

Instance profile credentials– used on EC2 instances, and delivered through the Amazon EC2 metadata service. The AWS SDK for Java uses the InstanceProfileCredentialsProvider to load these credentials.

Note

Instance profile credentials are used only if AWS_CONTAINER_CREDENTIALS_RELATIVE_URI is not set. See EC2ContainerCredentialsProviderWrapper for more information.

For pyspark, I use setting to access s3 content.

def get_spark_context(app_name):    # configure    conf = pyspark.SparkConf()    # init & return    sc = pyspark.SparkContext.getOrCreate(conf=conf)    # s3a config    sc._jsc.hadoopConfiguration().set('fs.s3a.endpoint',                                      's3.eu-central-1.amazonaws.com')    sc._jsc.hadoopConfiguration().set(        'fs.s3a.aws.credentials.provider',        'com.amazonaws.auth.InstanceProfileCredentialsProvider,'        'com.amazonaws.auth.profile.ProfileCredentialsProvider'    )    return pyspark.SQLContext(sparkContext=sc)

More on spark context here.

Please refer this for type S3 access.