Accessing S3 data in another Amazon account from EMR Serverless - Amazon EMR
PrerequisitesUse an S3 bucket policyUse an assumed roleAssumed role examples
Services or capabilities described in Amazon Web Services documentation might vary by Region. To see the differences applicable to the China Regions, see Getting Started with Amazon Web Services in China (PDF).

Accessing S3 data in another Amazon account from EMR Serverless

You can run Amazon EMR Serverless jobs from one Amazon account and configure them to access data in Amazon S3 buckets that belong to another Amazon account. This page describes how to configure cross-account access to S3 from EMR Serverless.

Jobs that run on EMR Serverless can use an S3 bucket policy or an assumed role to access data in Amazon S3 from a different Amazon account.

Prerequisites

To set up cross-account access for Amazon EMR Serverless, you must complete tasks while signed in to two Amazon accounts:

  • AccountA – This is the Amazon account where you have created an Amazon EMR Serverless application. Before you set up cross-account access, you must have the following ready in this account:

    • An Amazon EMR Serverless application where you want to run jobs.

    • A job execution role that has the required permissions to run jobs in the application. For more information, see Job runtime roles for Amazon EMR Serverless.

  • AccountB – This is the Amazon account that contains the S3 bucket that you want your Amazon EMR Serverless jobs to access.

Use an S3 bucket policy to access cross-account S3 data

To access the S3 bucket in account B from account A, attach the following policy to the S3 bucket in account B.

{
   "Version": "2012-10-17",
   "Statement": [
      {
         "Sid": "Example permissions 1",
         "Effect": "Allow",
         "Principal": {
            "AWS": "arn:aws:iam::AccountA:root"
         },
         "Action": [
            "s3:ListBucket"
         ],
         "Resource": [
            "arn:aws:s3:::bucket_name_in_AccountB"
         ]
      },
      {
         "Sid": "Example permissions 2",
         "Effect": "Allow",
         "Principal": {
            "AWS": "arn:aws:iam::AccountA:root"
         },
         "Action": [
            "s3:PutObject",
            "s3:GetObject",
            "s3:DeleteObject"
         ],
         "Resource": [
            "arn:aws:s3:::bucket_name_in_AccountB/*"
         ]
      }
   ]
}

For more information about S3 cross-account access with S3 bucket policies, see Example 2: Bucket owner granting cross-account bucket permissions in the Amazon Simple Storage Service User Guide.

Use an assumed role to access cross-account S3 data

Another way to set up cross-account access for Amazon EMR Serverless is with the AssumeRole action from the Amazon Security Token Service (Amazon STS). Amazon STS is a global web service that lets you request temporary, limited-privilege credentials for users. You can make API calls to EMR Serverless and Amazon S3 with the temporary security credentials that you create with AssumeRole.

The following steps illustrate how to use an assumed role to access cross-account S3 data from EMR Serverless:

  1. Create an Amazon S3 bucket, cross-account-bucket, in AccountB. For more information, see Creating a bucket in the Amazon Simple Storage Service User Guide. If you want to have cross-account access to DynamoDB, you can also create a DynamoDB table in AccountB. For more information, see Creating a DynamoDB table in the Amazon DynamoDB Developer Guide.

  2. Create a Cross-Account-Role-B IAM role in AccountB that can access the cross-account-bucket.

    1. Sign in to the Amazon Web Services Management Console and open the IAM console at https://console.amazonaws.cn/iam/.

    2. Choose Roles and create a new role: Cross-Account-Role-B. For more information about how to create IAM roles, see Creating IAM roles in the IAM User Guide.

    3. Create an IAM policy that specifies the permissions for Cross-Account-Role-B to access the cross-account-bucket S3 bucket, as the following policy statement demonstrates. Then attach the IAM policy to Cross-Account-Role-B. For more information, see Creating IAM policies in the IAM User Guide.

    {
"Version": "2012-10-17",
    "Statement": [
        {
            "Effect": "Allow",
            "Action": "s3:*",
            "Resource": [
                "arn:aws:s3:::cross-account-bucket",
                "arn:aws:s3:::cross-account-bucket/*"
            ]
        }
    ]
}

    If you require DynamoDB access, create an IAM policy that specifies permissions to access the cross-account DynamoDB table. Then attach the IAM policy to Cross-Account-Role-B. For more information, see Amazon DynamoDB: Allows access to a specific table in the IAM User Guide.

    The following is a policy to allow access to the DynamoDB table CrossAccountTable.

    {
"Version": "2012-10-17",
    "Statement": [
        {
            "Effect": "Allow",
            "Action": "dynamodb:*",
            "Resource": "arn:aws:dynamodb:MyRegion:AccountB:table/CrossAccountTable"
        }
    ]
}

  3. Edit the trust relationship for the Cross-Account-Role-B role.

    1. To configure the trust relationship for the role, choose the Trust Relationships tab in the IAM console for the role Cross-Account-Role-B that you created in Step 2.

    2. Select Edit Trust Relationship.

    3. Add the following policy document. This allows Job-Execution-Role-A in AccountA to assume the Cross-Account-Role-B role.

      {
"Version": "2012-10-17",
  "Statement": [
    {
        "Effect": "Allow",
        "Principal": {
            "AWS": "arn:aws:iam::AccountA:role/Job-Execution-Role-A"
        },
        "Action": "sts:AssumeRole"
    }
  ]
}

  4. Grant Job-Execution-Role-A in AccountA the Amazon STS AssumeRole permission to assume Cross-Account-Role-B.

    1. In the IAM console for Amazon account AccountA, select Job-Execution-Role-A.

    2. Add the following policy statement to the Job-Execution-Role-A to allow the AssumeRole action on the Cross-Account-Role-B role.

      {
"Version": "2012-10-17",
    "Statement": [
        {
            "Effect": "Allow",
            "Action": "sts:AssumeRole",
            "Resource": "arn:aws:iam::AccountB:role/Cross-Account-Role-B"
        }
    ]
}

Assumed role examples

You can use a single assumed role to access all S3 resources in an account, or with Amazon EMR 6.11 and higher, you can configure multiple IAM roles to assume when you access different cross-account S3 buckets.

Access S3 resources with one assumed role

Note

When you configure a job to use a single assumed role, all S3 resources throughout the job use that role, including the entryPoint script.

If you want to use a single assumed role to access all S3 resources in account B, specify the following configurations:

  1. Specify EMRFS configuration fs.s3.customAWSCredentialsProvider to spark.hadoop.fs.s3.customAWSCredentialsProvider=com.amazonaws.emr.AssumeRoleAWSCredentialsProvider.

  2. For Spark, use spark.emr-serverless.driverEnv.ASSUME_ROLE_CREDENTIALS_ROLE_ARN and spark.executorEnv.ASSUME_ROLE_CREDENTIALS_ROLE_ARN to specify the environment variables on driver and executors.

  3. For Hive, use hive.emr-serverless.launch.env.ASSUME_ROLE_CREDENTIALS_ROLE_ARN, tez.am.emr-serverless.launch.env.ASSUME_ROLE_CREDENTIALS_ROLE_ARN, and tez.task.emr-serverless.launch.env.ASSUME_ROLE_CREDENTIALS_ROLE_ARN to specify the environment variables on Hive driver, Tez application master, and Tez task containers.

The following examples show how to use an assumed role to start an EMR Serverless job run with cross-account access.

Spark

The following example shows how to use an assumed role to start an EMR Serverless Spark job run with cross-account access to S3.

aws emr-serverless start-job-run \
    --application-id application-id \
    --execution-role-arn job-role-arn \
    --job-driver '{
        "sparkSubmit": {
            "entryPoint": "entrypoint_location",
            "entryPointArguments": [":argument_1:", ":argument_2:"],
            "sparkSubmitParameters": "--conf spark.executor.cores=4 --conf spark.executor.memory=20g --conf spark.driver.cores=4 --conf spark.driver.memory=8g --conf spark.executor.instances=1"
        }
    }' \
     --configuration-overrides '{
        "applicationConfiguration": [{
            "classification": "spark-defaults",
            "properties": {
                "spark.hadoop.fs.s3.customAWSCredentialsProvider": "spark.hadoop.fs.s3.customAWSCredentialsProvider=com.amazonaws.emr.AssumeRoleAWSCredentialsProvider",
                "spark.emr-serverless.driverEnv.ASSUME_ROLE_CREDENTIALS_ROLE_ARN": "arn:aws:iam::AccountB:role/Cross-Account-Role-B",
                "spark.executorEnv.ASSUME_ROLE_CREDENTIALS_ROLE_ARN": "arn:aws:iam::AccountB:role/Cross-Account-Role-B"
            }
        }]
    }'
Hive

The following example shows how to use an assumed role to start an EMR Serverless Hive job run with cross-account access to S3.

aws emr-serverless start-job-run \
    --application-id application-id \
    --execution-role-arn job-role-arn \
    --job-driver '{
        "hive": {
            "query": "query_location",
            "parameters": "hive_parameters"
        }
    }' \
    --configuration-overrides '{
        "applicationConfiguration": [{
            "classification": "hive-site",
            "properties": {
                "fs.s3.customAWSCredentialsProvider": "com.amazonaws.emr.serverless.credentialsprovider.AssumeRoleAWSCredentialsProvider",
                "hive.emr-serverless.launch.env.ASSUME_ROLE_CREDENTIALS_ROLE_ARN": "arn:aws:iam::AccountB:role/Cross-Account-Role-B",
                "tez.am.emr-serverless.launch.env.ASSUME_ROLE_CREDENTIALS_ROLE_ARN": "arn:aws:iam::AccountB:role/Cross-Account-Role-B",
                "tez.task.emr-serverless.launch.env.ASSUME_ROLE_CREDENTIALS_ROLE_ARN": "arn:aws:iam::AccountB:role/Cross-Account-Role-B"
            }
        }]
    }'

Access S3 resources with multiple assumed roles

With EMR Serverless releases 6.11.0 and higher, you can configure multiple IAM roles to assume when you access different cross-account buckets. If you want to access different S3 resources with different assumed roles in account B, use following configurations when you start the job run:

  1. Specify EMRFS configuration fs.s3.customAWSCredentialsProvider to com.amazonaws.emr.serverless.credentialsprovider.BucketLevelAssumeRoleCredentialsProvider.

  2. Specify EMRFS configuration fs.s3.bucketLevelAssumeRoleMapping to define the mapping from S3 bucket name to the IAM role in account B to assume. The value should be in format of bucket1->role1;bucket2->role2.

For example, you can use arn:aws:iam::AccountB:role/Cross-Account-Role-B-1 to access bucket bucket1, and use arn:aws:iam::AccountB:role/Cross-Account-Role-B-2 to access bucket bucket2. The following examples show how to start an EMR Serverless job run with cross-account access through multiple assumed roles.

Spark

The following example shows how to use multiple assumed roles to create an EMR Serverless Spark job run.

aws emr-serverless start-job-run \
    --application-id application-id \
    --execution-role-arn job-role-arn \
    --job-driver '{
        "sparkSubmit": {
            "entryPoint": "entrypoint_location",
            "entryPointArguments": [":argument_1:", ":argument_2:"],
            "sparkSubmitParameters": "--conf spark.executor.cores=4 --conf spark.executor.memory=20g --conf spark.driver.cores=4 --conf spark.driver.memory=8g --conf spark.executor.instances=1"
        }
    }' \
     --configuration-overrides '{
        "applicationConfiguration": [{
            "classification": "spark-defaults",
            "properties": {
                "spark.hadoop.fs.s3.customAWSCredentialsProvider": "com.amazonaws.emr.serverless.credentialsprovider.BucketLevelAssumeRoleCredentialsProvider",
                "spark.hadoop.fs.s3.bucketLevelAssumeRoleMapping": "bucket1->arn:aws:iam::AccountB:role/Cross-Account-Role-B-1;bucket2->arn:aws:iam::AccountB:role/Cross-Account-Role-B-2"
            }
        }]
    }'
Hive

The following examples show how to use multiple assumed roles to create an EMR Serverless Hive job run.

aws emr-serverless start-job-run \
    --application-id application-id \
    --execution-role-arn job-role-arn \
    --job-driver '{
        "hive": {
            "query": "query_location",
            "parameters": "hive_parameters"
        }
    }' \
    --configuration-overrides '{
        "applicationConfiguration": [{
            "classification": "hive-site",
            "properties": {
                "fs.s3.customAWSCredentialsProvider": "com.amazonaws.emr.serverless.credentialsprovider.AssumeRoleAWSCredentialsProvider",
                "fs.s3.bucketLevelAssumeRoleMapping": "bucket1->arn:aws:iam::AccountB:role/Cross-Account-Role-B-1;bucket2->arn:aws:iam::AccountB:role/Cross-Account-Role-B-2"
            }
        }]
    }'