Επεξεργασία

Κοινή χρήση μέσω


CREATE EXTERNAL DATA SOURCE (Transact-SQL)

Creates an external data source for querying using SQL Server, Azure SQL Database, Azure SQL Managed Instance, Azure Synapse Analytics, Analytics Platform System (PDW), or Azure SQL Edge.

This article provides the syntax, arguments, remarks, permissions, and examples for whichever SQL product you choose.

Select a product

In the following row, select the product name you're interested in, and only that product's information is displayed.

Overview: SQL Server 2016

Applies to: SQL Server 2016 (13.x) and later versions

Creates an external data source for PolyBase queries. External data sources are used to establish connectivity and support these primary use cases:

  • Data virtualization and data load using PolyBase
  • Bulk load operations using BULK INSERT or OPENROWSET

Note

This syntax varies in different versions of SQL Server. Use the version selector dropdown to choose the appropriate version.
To view the features of SQL Server 2019 (15.x), visit CREATE EXTERNAL DATA SOURCE. To view the features of SQL Server 2022 (16.x), visit CREATE EXTERNAL DATA SOURCE.

Syntax for SQL Server 2016

For more information about the syntax conventions, see Transact-SQL syntax conventions.

CREATE EXTERNAL DATA SOURCE <data_source_name>
WITH
  ( [ LOCATION = '<prefix>://<path>[:<port>]' ]
    [ [ , ] CREDENTIAL = <credential_name> ]
    [ [ , ] TYPE = { HADOOP } ]
    [ [ , ] RESOURCE_MANAGER_LOCATION = '<resource_manager>[:<port>]' )
[ ; ]

Arguments

data_source_name

Specifies the user-defined name for the data source. The name must be unique within the database in SQL Server.

LOCATION = '<prefix>://<path[:port]>'

Provides the connectivity protocol and path to the external data source.

External Data Source Connector location prefix Location path Supported locations by product / service Authentication
Cloudera CDH or Hortonworks HDP hdfs <Namenode>[:port] SQL Server 2016 (13.x) to SQL Server 2019 (15.x) Anonymous or basic authentication
Azure Storage account(V2) wasb[s] <container>@<storage_account>.blob.core.windows.net Starting with SQL Server 2016 (13.x)
Hierarchical Namespace not supported
Azure Storage account key

Location path:

  • <Namenode> = the machine name, name service URI, or IP address of the Namenode in the Hadoop cluster. PolyBase must resolve any DNS names used by the Hadoop cluster.
  • port = The port that the external data source is listening on. In Hadoop, the port can be found using the fs.defaultFS configuration parameter. The default is 8020.
  • <container> = the container of the storage account holding the data. Root containers are read-only, data can't be written back to the container.
  • <storage_account> = the storage account name of the Azure resource.
  • <server_name> = the host name.
  • <instance_name> = the name of the SQL Server named instance. Used if you have SQL Server Browser Service running on the target instance.

Additional notes and guidance when setting the location:

  • The SQL Server Database Engine doesn't verify the existence of the external data source when the object is created. To validate, create an external table using the external data source.
  • Use the same external data source for all tables when querying Hadoop to ensure consistent querying semantics.
  • wasbs is optional but recommended in SQL Server 2016 (13.x) for accessing Azure Storage Accounts as data will be sent using a secure TLS/SSL connection.
  • To ensure successful PolyBase queries during a Hadoop Namenode fail-over, consider using a virtual IP address for the Namenode of the Hadoop cluster. If you don't, execute an ALTER EXTERNAL DATA SOURCE command to point to the new location.

CREDENTIAL = credential_name

Specifies a database-scoped credential for authenticating to the external data source.

CREDENTIAL is only required if the data has been secured. CREDENTIAL isn't required for data sets that allow anonymous access.

To create a database scoped credential, see CREATE DATABASE SCOPED CREDENTIAL (Transact-SQL).

TYPE = [ HADOOP ]

Specifies the type of the external data source being configured. In SQL Server 2016, this parameter is always required, and should only be specified as HADOOP. Supports connections to Cloudera CDH, Hortonworks HDP, or an Azure Storage account. The behavior of this parameter is different in later versions of SQL Server.

For an example of using TYPE = HADOOP to load data from an Azure Storage account, see Create external data source to access data in Azure Storage using the wasb:// interface

RESOURCE_MANAGER_LOCATION = 'ResourceManager_URI[:port]'

Configure this optional value when connecting to Cloudera CDH, Hortonworks HDP, or an Azure Storage account only. For a complete list of supported Hadoop versions, see PolyBase Connectivity Configuration (Transact-SQL).

When the RESOURCE_MANAGER_LOCATION is defined, the query optimizer makes a cost-based decision to improve performance. A MapReduce job can be used to push down the computation to Hadoop. Specifying the RESOURCE_MANAGER_LOCATION can significantly reduce the volume of data transferred between Hadoop and SQL Server, which can lead to improved query performance.

If the Resource Manager isn't specified, pushing compute to Hadoop is disabled for PolyBase queries. Create external data source to reference Hadoop with push-down enabled provides a concrete example and further guidance.

The RESOURCE_MANAGER_LOCATION value is not validated when you create the external data source. Entering an incorrect value may cause query failure at execution time whenever push-down is attempted as the provided value would not be able to resolve.

In order for PolyBase to function correctly with a Hadoop external data source, the ports for the following Hadoop cluster components must be open:

  • HDFS ports
    • Namenode
    • Datanode
  • Resource Manager
    • Job submission
    • Job history

If the port isn't specified, the default value is chosen using the current setting for 'hadoop connectivity' configuration.

Hadoop Connectivity Default Resource Manager Port
1 50300
2 50300
3 8021
4 8032
5 8050
6 8032
7 8050
8 8032

The following table shows the default ports for these components. There is Hadoop version dependency as well as the possibility of custom configuration that doesn't use the default port assignment.

Hadoop cluster component Default Port
NameNode 8020
DataNode (Data transfer, non-privilege IPC port) 50010
DataNode (Data transfer, privilege IPC port) 1019
Resource Manager Job Submission (Hortonworks 1.3) 50300
Resource Manager Job Submission (Cloudera 4.3) 8021
Resource Manager Job Submission (Hortonworks 2.0 on Windows, Cloudera 5.x on Linux) 8032
Resource Manager Job Submission (Hortonworks 2.x, 3.0 on Linux, Hortonworks 2.1-3 on Windows) 8050
Resource Manager Job History 10020

Permissions

Requires CONTROL permission on database in SQL Server.

Locking

Takes a shared lock on the EXTERNAL DATA SOURCE object.

Security

PolyBase supports proxy based authentication for most external data sources. Create a database scoped credential to create the proxy account.

Examples

Important

For information on how to install and enable PolyBase, see Install PolyBase on Windows

A. Create external data source to reference Hadoop

To create an external data source to reference your Hortonworks HDP or Cloudera CDH Hadoop cluster, specify the machine name, or IP address of the Hadoop Namenode and port.

CREATE EXTERNAL DATA SOURCE MyHadoopCluster
WITH (
    LOCATION = 'hdfs://10.10.10.10:8050',
    TYPE = HADOOP
);

B. Create external data source to reference Hadoop with push-down enabled

Specify the RESOURCE_MANAGER_LOCATION option to enable push-down computation to Hadoop for PolyBase queries. Once enabled, PolyBase makes a cost-based decision to determine whether the query computation should be pushed to Hadoop.

CREATE EXTERNAL DATA SOURCE MyHadoopCluster
WITH (
    LOCATION = 'hdfs://10.10.10.10:8020',
    TYPE = HADOOP,
    RESOURCE_MANAGER_LOCATION = '10.10.10.10:8050'
);

C. Create external data source to reference Kerberos-secured Hadoop

To verify if the Hadoop cluster is Kerberos-secured, check the value of hadoop.security.authentication property in Hadoop core-site.xml. To reference a Kerberos-secured Hadoop cluster, you must specify a database scoped credential that contains your Kerberos username and password. The database master key is used to encrypt the database scoped credential secret.

-- Create a database master key if one does not already exist, using your own password.
-- This key is used to encrypt the credential secret in next step.
CREATE MASTER KEY ENCRYPTION BY PASSWORD = '<password>';

-- Create a database scoped credential with Kerberos user name and password.
CREATE DATABASE SCOPED CREDENTIAL HadoopUser1
    WITH IDENTITY = '<hadoop_user_name>',
    SECRET = '<hadoop_password>';

-- Create an external data source with CREDENTIAL option.
CREATE EXTERNAL DATA SOURCE MyHadoopCluster
WITH (
    LOCATION = 'hdfs://10.10.10.10:8050',
    CREDENTIAL = HadoopUser1,
    TYPE = HADOOP,
    RESOURCE_MANAGER_LOCATION = '10.10.10.10:8050'
);

D. Create external data source to access data in Azure Storage using the wasb:// interface

In this example, the external data source is an Azure V2 Storage account named logs. The storage container is called daily. The Azure Storage external data source is for data transfer only. It doesn't support predicate push-down. Hierarchical namespaces are not supported when accessing data via the wasb:// interface.

This example shows how to create the database scoped credential for authentication to an Azure V2 Storage account. Specify the Azure Storage account key in the database credential secret. You can specify any string in database scoped credential identity as it isn't used during authentication to Azure Storage. Note that when connecting to the Azure Storage via the WASB[s] connector, authentication must be done with a storage account key, not with a shared access signature (SAS).

In SQL Server 2016 (13.x), TYPE should be set to HADOOP even when accessing Azure Storage.

-- Create a database master key if one does not already exist, using your own password.
-- This key is used to encrypt the credential secret in next step.
CREATE MASTER KEY ENCRYPTION BY PASSWORD = '<password>';

-- Create a database scoped credential with Azure storage account key as the secret.
CREATE DATABASE SCOPED CREDENTIAL AzureStorageCredential
    WITH IDENTITY = '<my_account>',
    SECRET = '<azure_storage_account_key>';

-- Create an external data source with CREDENTIAL option.
CREATE EXTERNAL DATA SOURCE MyAzureStorage
WITH (
    LOCATION = 'wasbs://daily@logs.blob.core.windows.net/',
    CREDENTIAL = AzureStorageCredential,
    TYPE = HADOOP
);

Next steps

Overview: SQL Server 2017

Applies to: SQL Server 2017 (14.x) only

Creates an external data source for PolyBase queries. External data sources are used to establish connectivity and support these primary use cases:

  • Data virtualization and data load using PolyBase
  • Bulk load operations using BULK INSERT or OPENROWSET

Note

This syntax varies in different versions of SQL Server on Linux. Use the version selector dropdown to choose the appropriate version.
To view the features of SQL Server 2019 (15.x), visit CREATE EXTERNAL DATA SOURCE. To view the features of SQL Server 2022 (16.x), visit CREATE EXTERNAL DATA SOURCE.

Note

This syntax varies in different versions of SQL Server. Use the version selector dropdown to choose the appropriate version.
To view the features of SQL Server 2019 (15.x), visit CREATE EXTERNAL DATA SOURCE. To view the features of SQL Server 2022 (16.x), visit CREATE EXTERNAL DATA SOURCE.

Syntax for SQL Server 2017

For more information about the syntax conventions, see Transact-SQL syntax conventions.

CREATE EXTERNAL DATA SOURCE <data_source_name>
WITH
  ( [ LOCATION = '<prefix>://<path>[:<port>]' ]
    [ [ , ] CREDENTIAL = <credential_name> ]
    [ [ , ] TYPE = { HADOOP | BLOB_STORAGE } ]
    [ [ , ] RESOURCE_MANAGER_LOCATION = '<resource_manager>[:<port>]' )
[ ; ]

Arguments

data_source_name

Specifies the user-defined name for the data source. The name must be unique within the database in SQL Server.

LOCATION = '<prefix>://<path[:port]>'

Provides the connectivity protocol and path to the external data source.

External Data Source Connector location prefix Location path Supported locations by product / service Authentication
Cloudera CDH or Hortonworks HDP hdfs <Namenode>[:port] SQL Server 2016 (13.x) to SQL Server 2019 (15.x) only Anonymous or basic authentication
Azure Storage account(V2) wasb[s] <container>@<storage_account>.blob.core.windows.net Starting with SQL Server 2016 (13.x)
Hierarchical Namespace not supported
Azure Storage account key
Bulk Operations https <storage_account>.blob.core.windows.net/<container> Starting with SQL Server 2017 (14.x) Shared access signature (SAS)

Location path:

  • <Namenode> = the machine name, name service URI, or IP address of the Namenode in the Hadoop cluster. PolyBase must resolve any DNS names used by the Hadoop cluster.
  • port = The port that the external data source is listening on. In Hadoop, the port can be found using the fs.defaultFS configuration parameter. The default is 8020.
  • <container> = the container of the storage account holding the data. Root containers are read-only, data can't be written back to the container.
  • <storage_account> = the storage account name of the Azure resource.
  • <server_name> = the host name.
  • <instance_name> = the name of the SQL Server named instance. Used if you have SQL Server Browser Service running on the target instance.

Additional notes and guidance when setting the location:

  • The SQL Server Database Engine doesn't verify the existence of the external data source when the object is created. To validate, create an external table using the external data source.
  • Use the same external data source for all tables when querying Hadoop to ensure consistent querying semantics.
  • Specify the Driver={<Name of Driver>} when connecting via ODBC.
  • wasbs is optional but recommended in SQL Server 2017 (14.x) for accessing Azure Storage Accounts as data will be sent using a secure TLS/SSL connection.
  • To ensure successful PolyBase queries during a Hadoop Namenode fail-over, consider using a virtual IP address for the Namenode of the Hadoop cluster. If you don't, execute an ALTER EXTERNAL DATA SOURCE command to point to the new location.

CREDENTIAL = credential_name

Specifies a database-scoped credential for authenticating to the external data source.

Additional notes and guidance when creating a credential:

  • CREDENTIAL is only required if the data has been secured. CREDENTIAL isn't required for data sets that allow anonymous access.
  • When the TYPE = BLOB_STORAGE, the credential must be created using SHARED ACCESS SIGNATURE as the identity.
  • TYPE = BLOB_STORAGE is only permitted for bulk operations; you cannot create external tables for an external data source with TYPE = BLOB_STORAGE.
  • Note that when connecting to the Azure Storage via the WASB[s] connector, authentication must be done with a storage account key, not with a shared access signature (SAS).
  • When TYPE = HADOOP the credential must be created using the storage account key as the SECRET.

There are multiple ways to create a shared access signature:

  • You can create an SAS token by navigating to the Azure portal -> <Your_Storage_Account> -> Shared access signature -> Configure permissions -> Generate SAS and connection string. For more information, see Generate a shared access signature.

  • You can create and configure an SAS with Azure Storage Explorer.

  • You can create an SAS programmatically via PowerShell, Azure CLI, .NET, and REST API. For more information, see Grant limited access to Azure Storage resources using shared access signatures (SAS).

  • The SAS token should be configured as follows:

    • When an SAS token is generated, it includes a question mark ('?') at the beginning of the token. Exclude the leading ? when configured as the SECRET.
    • Use a valid expiration period (all dates are in UTC time).
    • Grant at least read permission on the file that should be loaded (for example srt=o&sp=r). Multiple shared access signatures can be created for different use cases. Permissions should be granted as follows:
    Action Permission
    Read data from a file Read
    Read data from multiple files and subfolders Read and List

For an example of using a CREDENTIAL with SHARED ACCESS SIGNATURE and TYPE = BLOB_STORAGE, see Create an external data source to execute bulk operations and retrieve data from Azure Storage into SQL Database

To create a database scoped credential, see CREATE DATABASE SCOPED CREDENTIAL (Transact-SQL).

TYPE = [ HADOOP | BLOB_STORAGE ]

Specifies the type of the external data source being configured. This parameter isn't always required, and should only be specified when connecting to Cloudera CDH, Hortonworks HDP, an Azure Storage account, or an Azure Data Lake Storage Gen2.

  • Use HADOOP when the external data source is Cloudera CDH, Hortonworks HDP, an Azure Storage account, or an Azure Data Lake Storage Gen2.
  • Use BLOB_STORAGE when executing bulk operations from Azure Storage account using BULK INSERT or OPENROWSET. Introduced with SQL Server 2017 (14.x). Use HADOOP when intending to CREATE EXTERNAL TABLE against Azure Storage.

Note

TYPE should be set to HADOOP even when accessing Azure Storage.

For an example of using TYPE = HADOOP to load data from an Azure Storage account, see Create external data source to access data in Azure Storage using the wasb:// interface

RESOURCE_MANAGER_LOCATION = 'ResourceManager_URI[:port]'

Configure this optional value when connecting to Cloudera CDH, Hortonworks HDP, or an Azure Storage account only. For a complete list of supported Hadoop versions, see PolyBase Connectivity Configuration (Transact-SQL).

When the RESOURCE_MANAGER_LOCATION is defined, the query optimizer will make a cost-based decision to improve performance. A MapReduce job can be used to push down the computation to Hadoop. Specifying the RESOURCE_MANAGER_LOCATION can significantly reduce the volume of data transferred between Hadoop and SQL Server, which can lead to improved query performance.

If the Resource Manager isn't specified, pushing compute to Hadoop is disabled for PolyBase queries. Create external data source to reference Hadoop with push-down enabled provides a concrete example and further guidance.

The RESOURCE_MANAGER_LOCATION value is not validated when you create the external data source. Entering an incorrect value may cause query failure at execution time whenever push-down is attempted as the provided value would not be able to resolve.

In order for PolyBase to function correctly with a Hadoop external data source, the ports for the following Hadoop cluster components must be open:

  • HDFS ports
    • Namenode
    • Datanode
  • Resource Manager
    • Job submission
    • Job history

If the port isn't specified, the default value is chosen using the current setting for 'hadoop connectivity' configuration.

Hadoop Connectivity Default Resource Manager Port
1 50300
2 50300
3 8021
4 8032
5 8050
6 8032
7 8050
8 8032

The following table shows the default ports for these components. Note that there is Hadoop version dependency as well as the possibility of custom configuration that doesn't use the default port assignment.

Hadoop cluster component Default Port
NameNode 8020
DataNode (Data transfer, non-privilege IPC port) 50010
DataNode (Data transfer, privilege IPC port) 1019
Resource Manager Job Submission (Hortonworks 1.3) 50300
Resource Manager Job Submission (Cloudera 4.3) 8021
Resource Manager Job Submission (Hortonworks 2.0 on Windows, Cloudera 5.x on Linux) 8032
Resource Manager Job Submission (Hortonworks 2.x, 3.0 on Linux, Hortonworks 2.1-3 on Windows) 8050
Resource Manager Job History 10020

Permissions

Requires CONTROL permission on database in SQL Server.

Locking

Takes a shared lock on the EXTERNAL DATA SOURCE object.

Security

PolyBase supports proxy based authentication for most external data sources. Create a database scoped credential to create the proxy account.

An SAS token with type HADOOP is unsupported. It's only supported with type = BLOB_STORAGE when a storage account access key is used instead. Attempting to create an external data source with type HADOOP and a SAS credential fails with the following error:

Msg 105019, Level 16, State 1 - EXTERNAL TABLE access failed due to internal error: 'Java exception raised on call to HdfsBridge_Connect. Java exception message: Parameters provided to connect to the Azure storage account are not valid.: Error [Parameters provided to connect to the Azure storage account are not valid.] occurred while accessing external file.'

Examples

Important

For information on how to install and enable PolyBase, see Install PolyBase on Windows

A. Create external data source to reference Hadoop

To create an external data source to reference your Hortonworks HDP or Cloudera CDH Hadoop cluster, specify the machine name, or IP address of the Hadoop Namenode and port.

CREATE EXTERNAL DATA SOURCE MyHadoopCluster
WITH (
    LOCATION = 'hdfs://10.10.10.10:8050',
    TYPE = HADOOP
);

B. Create external data source to reference Hadoop with push-down enabled

Specify the RESOURCE_MANAGER_LOCATION option to enable push-down computation to Hadoop for PolyBase queries. Once enabled, PolyBase makes a cost-based decision to determine whether the query computation should be pushed to Hadoop.

CREATE EXTERNAL DATA SOURCE MyHadoopCluster
WITH (
    LOCATION = 'hdfs://10.10.10.10:8020',
    TYPE = HADOOP,
    RESOURCE_MANAGER_LOCATION = '10.10.10.10:8050'
);

C. Create external data source to reference Kerberos-secured Hadoop

To verify if the Hadoop cluster is Kerberos-secured, check the value of hadoop.security.authentication property in Hadoop core-site.xml. To reference a Kerberos-secured Hadoop cluster, you must specify a database scoped credential that contains your Kerberos username and password. The database master key is used to encrypt the database scoped credential secret.

-- Create a database master key if one does not already exist, using your own password.
-- This key is used to encrypt the credential secret in next step.
CREATE MASTER KEY ENCRYPTION BY PASSWORD = '<password>';

-- Create a database scoped credential with Kerberos user name and password.
CREATE DATABASE SCOPED CREDENTIAL HadoopUser1
    WITH IDENTITY = '<hadoop_user_name>',
    SECRET = '<hadoop_password>';

-- Create an external data source with CREDENTIAL option.
CREATE EXTERNAL DATA SOURCE MyHadoopCluster
WITH (
    LOCATION = 'hdfs://10.10.10.10:8050',
    CREDENTIAL = HadoopUser1,
    TYPE = HADOOP,
    RESOURCE_MANAGER_LOCATION = '10.10.10.10:8050'
);

D. Create external data source to access data in Azure Storage using the wasb:// interface

In this example, the external data source is an Azure V2 Storage account named logs. The storage container is called daily. The Azure Storage external data source is for data transfer only. It doesn't support predicate push-down. Hierarchical namespaces are not supported when accessing data via the wasb:// interface. Note that when connecting to the Azure Storage via the WASB[s] connector, authentication must be done with a storage account key, not with a shared access signature (SAS).

This example shows how to create the database scoped credential for authentication to an Azure V2 Storage account. Specify the Azure Storage account key in the database credential secret. You can specify any string in database scoped credential identity as it isn't used during authentication to Azure Storage.

-- Create a database master key if one does not already exist, using your own password.
-- This key is used to encrypt the credential secret in next step.
CREATE MASTER KEY ENCRYPTION BY PASSWORD = '<password>';

-- Create a database scoped credential with Azure storage account key as the secret.
CREATE DATABASE SCOPED CREDENTIAL AzureStorageCredential
    WITH IDENTITY = '<my_account>',
    SECRET = '<azure_storage_account_key>';

-- Create an external data source with CREDENTIAL option.
CREATE EXTERNAL DATA SOURCE MyAzureStorage
WITH (
    LOCATION = 'wasbs://daily@logs.blob.core.windows.net/',
    CREDENTIAL = AzureStorageCredential,
    TYPE = HADOOP
);

Examples: Bulk operations

Important

Do not add a trailing /, file name, or shared access signature parameters at the end of the LOCATION URL when configuring an external data source for bulk operations.

E. Create an external data source for bulk operations retrieving data from Azure Storage

Applies to: SQL Server 2017 (14.x) and later.

Use the following data source for bulk operations using BULK INSERT or OPENROWSET. The credential must set SHARED ACCESS SIGNATURE as the identity, mustn't have the leading ? in the SAS token, must have at least read permission on the file that should be loaded (for example srt=o&sp=r), and the expiration period should be valid (all dates are in UTC time). For more information on shared access signatures, see Using Shared Access Signatures (SAS).

CREATE DATABASE SCOPED CREDENTIAL AccessAzureInvoices
    WITH IDENTITY = 'SHARED ACCESS SIGNATURE',
    -- Remove ? from the beginning of the SAS token
    SECRET = '<azure_storage_account_key>';

CREATE EXTERNAL DATA SOURCE MyAzureInvoices
WITH (
    LOCATION = 'https://newinvoices.blob.core.windows.net/week3',
    CREDENTIAL = AccessAzureInvoices,
    TYPE = BLOB_STORAGE
);

To see this example in use, see the BULK INSERT example.

Next steps

Overview: SQL Server 2019

Applies to: SQL Server 2019 (15.x) and later

Creates an external data source for PolyBase queries. External data sources are used to establish connectivity and support these primary use cases:

  • Data virtualization and data load using PolyBase
  • Bulk load operations using BULK INSERT or OPENROWSET

Note

This syntax varies in different versions of SQL Server. Use the version selector dropdown to choose the appropriate version.
To view the features of SQL Server 2022 (16.x), visit CREATE EXTERNAL DATA SOURCE.

Note

This syntax varies in different versions of SQL Server. Use the version selector dropdown to choose the appropriate version.
To view the features of SQL Server 2022 (16.x), visit CREATE EXTERNAL DATA SOURCE.

Syntax for SQL Server 2019

For more information about the syntax conventions, see Transact-SQL syntax conventions.

CREATE EXTERNAL DATA SOURCE <data_source_name>
WITH
  ( [ LOCATION = '<prefix>://<path>[:<port>]' ]
    [ [ , ] CONNECTION_OPTIONS = '<key_value_pairs>'[,...]]
    [ [ , ] CREDENTIAL = <credential_name> ]
    [ [ , ] PUSHDOWN = { ON | OFF } ]
    [ [ , ] TYPE = { HADOOP | BLOB_STORAGE } ]
    [ [ , ] RESOURCE_MANAGER_LOCATION = '<resource_manager>[:<port>]' )
[ ; ]

Arguments

data_source_name

Specifies the user-defined name for the data source. The name must be unique within the database in SQL Server.

LOCATION = '<prefix>://<path[:port]>'

Provides the connectivity protocol and path to the external data source.

External Data Source Connector location prefix Location path Supported locations by product / service Authentication
Cloudera CDH or Hortonworks HDP hdfs <Namenode>[:port] SQL Server 2016 (13.x) to SQL Server 2019 (15.x) Anonymous or basic authentication
Azure Storage account(V2) wasb[s] <container>@<storage_account>.blob.core.windows.net Starting with SQL Server 2016 (13.x)
Hierarchical Namespace not supported
Azure Storage account key
SQL Server sqlserver <server_name>[\<instance_name>][:port] Starting with SQL Server 2019 (15.x) SQL authentication only
Oracle oracle <server_name>[:port] Starting with SQL Server 2019 (15.x) Basic authentication only
Teradata teradata <server_name>[:port] Starting with SQL Server 2019 (15.x) Basic authentication only
MongoDB or Cosmos DB API for MongoDB mongodb <server_name>[:port] Starting with SQL Server 2019 (15.x) Basic authentication only
Generic ODBC odbc <server_name>[:port] Starting with SQL Server 2019 (15.x) - Windows only Basic authentication only
Bulk Operations https <storage_account>.blob.core.windows.net/<container> Starting with SQL Server 2017 (14.x) Shared access signature (SAS)
Azure Data Lake Storage Gen2 abfs[s] abfss://<container>@<storage _account>.dfs.core.windows.net Starting with SQL Server 2019 (15.x) CU11+. Storage Access Key
SQL Server Big Data Clusters data pool sqldatapool sqldatapool://controller-svc/default Only supported in SQL Server 2019 Big Data Clusters Basic authentication only
SQL Server Big Data Clusters storage pool sqlhdfs sqlhdfs://controller-svc/default Only supported in SQL Server 2019 Big Data Clusters Basic authentication only

Location path:

  • <Namenode> = the machine name, name service URI, or IP address of the Namenode in the Hadoop cluster. PolyBase must resolve any DNS names used by the Hadoop cluster.
  • port = The port that the external data source is listening on. In Hadoop, the port can be found using the fs.defaultFS configuration parameter. The default is 8020.
  • <container> = the container of the storage account holding the data. Root containers are read-only, data can't be written back to the container.
  • <storage_account> = the storage account name of the Azure resource.
  • <server_name> = the host name.
  • <instance_name> = the name of the SQL Server named instance. Used if you have SQL Server Browser Service running on the target instance.

Additional notes and guidance when setting the location:

  • The SQL Server Database Engine doesn't verify the existence of the external data source when the object is created. To validate, create an external table using the external data source.
  • Use the same external data source for all tables when querying Hadoop to ensure consistent querying semantics.
  • You can use the sqlserver connector to connect SQL Server 2019 (15.x) to another SQL Server, or to Azure SQL Database.
  • Specify the Driver={<Name of Driver>} when connecting via ODBC.
  • Using wasbs or abfss is optional but recommended in SQL Server 2019 (15.x) for accessing Azure Storage Accounts as data will be sent using a secure TLS/SSL connection.
  • The abfs or abfss APIs are supported when accessing Azure Storage Accounts starting with SQL Server 2019 (15.x) CU11. For more information, see the Azure Blob Filesystem driver (ABFS).
  • The Hierarchical Namespace option for Azure Storage Accounts(V2) using abfs[s] is supported via Azure Data Lake Storage Gen2 starting with SQL Server 2019 (15.x) CU11+. The Hierarchical Namespace option is otherwise not supported, and this option should remain disabled.
  • To ensure successful PolyBase queries during a Hadoop Namenode fail-over, consider using a virtual IP address for the Namenode of the Hadoop cluster. If you don't, execute an ALTER EXTERNAL DATA SOURCE command to point to the new location.
  • The sqlhdfs and sqldatapool types are supported for connecting between the master instance and storage pool of a big data cluster. For Cloudera CDH or Hortonworks HDP, use hdfs. For more information on using sqlhdfs for querying SQL Server Big Data Clusters storage pools, see Query HDFS in SQL Server 2019 Big Data Cluster.
  • SQL Server support for HDFS Cloudera (CDP) and Hortonworks (HDP) external data sources will be retired and will not be included in SQL Server 2022 (16.x). For more information, see Big data options on the Microsoft SQL Server platform.

CONNECTION_OPTIONS = key_value_pair

Specified for SQL Server 2019 (15.x) and later. Specifies additional options when connecting over ODBC to an external data source. To use multiple connection options, separate them by a semi-colon.

Applies to generic ODBC connections, as well as built-in ODBC connectors for SQL Server, Oracle, Teradata, MongoDB, and Azure Cosmos DB API for MongoDB.

The key_value_pair is the keyword and the value for a specific connection option. The available keywords and values depend on the external data source type. The name of the driver is required as a minimum, but there are other options such as APP='<your_application_name>' or ApplicationIntent= ReadOnly|ReadWrite that are also useful to set and can assist with troubleshooting.

Possible key value pairs are specific to the provider for the external data source vendor. For more information for each provider, see CREATE EXTERNAL DATA SOURCE (Transact-SQL) CONNECTION_OPTIONS.

Starting in SQL Server 2019 (15.x) cumulative update 19, additional keywords were introduced to support Oracle TNS files:

  • The keyword TNSNamesFile specifies the filepath to the tnsnames.ora file located on the Oracle server.
  • The keyword ServerName specifies the alias used inside the tnsnames.ora that will be used to replace the host name and the port.

Pushdown = ON | OFF

Specified for SQL Server 2019 (15.x) only. States whether computation can be pushed down to the external data source. It is ON by default.

PUSHDOWN is supported when connecting to SQL Server, Oracle, Teradata, MongoDB, the Azure Cosmos DB API for MongoDB, or ODBC at the external data source level.

Enabling or disabling push-down at the query level is achieved through a hint.

CREDENTIAL = credential_name

Specifies a database-scoped credential for authenticating to the external data source.

Additional notes and guidance when creating a credential:

  • CREDENTIAL is only required if the data has been secured. CREDENTIAL isn't required for data sets that allow anonymous access.
  • When the TYPE = BLOB_STORAGE, the credential must be created using SHARED ACCESS SIGNATURE as the identity.
    • TYPE = BLOB_STORAGE is only permitted for bulk operations; you cannot create external tables for an external data source with TYPE = BLOB_STORAGE.

There are multiple ways to create a shared access signature:

  • You can create an SAS token by navigating to the Azure portal -> <Your_Storage_Account> -> Shared access signature -> Configure permissions -> Generate SAS and connection string. For more information, see Generate a shared access signature.

  • You can create and configure an SAS with Azure Storage Explorer.

  • You can create an SAS programmatically via PowerShell, Azure CLI, .NET, and REST API. For more information, see Grant limited access to Azure Storage resources using shared access signatures (SAS).

  • The SAS token should be configured as follows:

    • When an SAS token is generated, it includes a question mark ('?') at the beginning of the token. Exclude the leading ? when configured as the SECRET.
    • Use a valid expiration period (all dates are in UTC time).
    • Grant at least read permission on the file that should be loaded (for example srt=o&sp=r). Multiple shared access signatures can be created for different use cases. Permissions should be granted as follows:
    Action Permission
    Read data from a file Read
    Read data from multiple files and subfolders Read and List

For an example of using a CREDENTIAL with SHARED ACCESS SIGNATURE and TYPE = BLOB_STORAGE, see Create an external data source to execute bulk operations and retrieve data from Azure Storage into SQL Database

To create a database scoped credential, see CREATE DATABASE SCOPED CREDENTIAL (Transact-SQL).

TYPE = [ HADOOP | BLOB_STORAGE ]

Specifies the type of the external data source being configured. This parameter isn't always required, and should only be specified when connecting to Cloudera CDH, Hortonworks HDP, an Azure Storage account, or an Azure Data Lake Storage Gen2.

  • In SQL Server 2019 (15.x), do not specify TYPE unless connecting to Cloudera CDH, Hortonworks HDP, an Azure Storage account.
  • Use HADOOP when the external data source is Cloudera CDH, Hortonworks HDP, an Azure Storage account, or an Azure Data Lake Storage Gen2.
  • Use BLOB_STORAGE when executing bulk operations from Azure Storage account using BULK INSERT, or OPENROWSET with SQL Server 2017 (14.x). Use HADOOP when intending to CREATE EXTERNAL TABLE against Azure Storage.
  • SQL Server support for HDFS Cloudera (CDP) and Hortonworks (HDP) external data sources will be retired and will not be included in SQL Server 2022 (16.x). For more information, see Big data options on the Microsoft SQL Server platform.

For an example of using TYPE = HADOOP to load data from an Azure Storage account, see Create external data source to access data in Azure Storage using the wasb:// interface

RESOURCE_MANAGER_LOCATION = 'ResourceManager_URI[:port]'

In SQL Server 2019 (15.x), do not specify RESOURCE_MANAGER_LOCATION unless connecting to Cloudera CDH, Hortonworks HDP, an Azure Storage account.

Configure this optional value when connecting to Cloudera CDH, Hortonworks HDP, or an Azure Storage account only. For a complete list of supported Hadoop versions, see PolyBase Connectivity Configuration (Transact-SQL).

When the RESOURCE_MANAGER_LOCATION is defined, the query optimizer makes a cost-based decision to improve performance. A MapReduce job can be used to push down the computation to Hadoop. Specifying the RESOURCE_MANAGER_LOCATION can significantly reduce the volume of data transferred between Hadoop and SQL Server, which can lead to improved query performance.

If the Resource Manager isn't specified, pushing compute to Hadoop is disabled for PolyBase queries. Create external data source to reference Hadoop with push-down enabled provides a concrete example and further guidance.

The RESOURCE_MANAGER_LOCATION value is not validated when you create the external data source. Entering an incorrect value may cause query failure at execution time whenever push-down is attempted as the provided value would not be able to resolve.

In order for PolyBase to function correctly with a Hadoop external data source, the ports for the following Hadoop cluster components must be open:

  • HDFS ports
    • Namenode
    • Datanode
  • Resource Manager
    • Job submission
    • Job history

If the port isn't specified, the default value is chosen using the current setting for 'hadoop connectivity' configuration.

Hadoop Connectivity Default Resource Manager Port
1 50300
2 50300
3 8021
4 8032
5 8050
6 8032
7 8050
8 8032

The following table shows the default ports for these components. Note that there is Hadoop version dependency as well as the possibility of custom configuration that doesn't use the default port assignment.

Hadoop cluster component Default Port
NameNode 8020
DataNode (Data transfer, non-privilege IPC port) 50010
DataNode (Data transfer, privilege IPC port) 1019
Resource Manager Job Submission (Hortonworks 1.3) 50300
Resource Manager Job Submission (Cloudera 4.3) 8021
Resource Manager Job Submission (Hortonworks 2.0 on Windows, Cloudera 5.x on Linux) 8032
Resource Manager Job Submission (Hortonworks 2.x, 3.0 on Linux, Hortonworks 2.1-3 on Windows) 8050
Resource Manager Job History 10020

Permissions

Requires CONTROL permission on database in SQL Server.

Locking

Takes a shared lock on the EXTERNAL DATA SOURCE object.

Security

PolyBase supports proxy based authentication for most external data sources. Create a database scoped credential to create the proxy account.

When you connect to the storage or data pool in SQL Server 2019 Big Data Cluster, the user's credentials are passed through to the back-end system. Create logins in the data pool itself to enable pass through authentication.

An SAS token with type HADOOP is unsupported. It's only supported with type = BLOB_STORAGE when a storage account access key is used instead. Attempting to create an external data source with type HADOOP and a SAS credential fails with the following error:

Msg 105019, Level 16, State 1 - EXTERNAL TABLE access failed due to internal error: 'Java exception raised on call to HdfsBridge_Connect. Java exception message: Parameters provided to connect to the Azure storage account are not valid.: Error [Parameters provided to connect to the Azure storage account are not valid.] occurred while accessing external file.'

Examples

Important

For information on how to install and enable PolyBase, see Install PolyBase on Windows

A. Create external data source in SQL Server 2019 to reference Oracle

To create an external data source that references Oracle, ensure you have a database scoped credential. You may optionally also enable or disable push-down of computation against this data source.

-- Create a database master key if one does not already exist, using your own password.
-- This key is used to encrypt the credential secret in next step.
CREATE MASTER KEY ENCRYPTION BY PASSWORD = '<password>';

-- Create a database scoped credential with Azure storage account key as the secret.
CREATE DATABASE SCOPED CREDENTIAL OracleProxyAccount
    WITH IDENTITY = 'oracle_username',
    SECRET = 'oracle_password';

CREATE EXTERNAL DATA SOURCE MyOracleServer
WITH (
    LOCATION = 'oracle://145.145.145.145:1521',
    CREDENTIAL = OracleProxyAccount,
    PUSHDOWN = ON
);

Optionally, the external data source to Oracle can use proxy authentication to provide fine grain access control. A proxy user can be configured to have limited access compared to the user being impersonated.

CREATE DATABASE SCOPED CREDENTIAL [OracleProxyCredential]
    WITH IDENTITY = 'oracle_username',
    SECRET = 'oracle_password';

CREATE EXTERNAL DATA SOURCE [OracleSalesSrvr]
WITH (
    LOCATION = 'oracle://145.145.145.145:1521',
    CONNECTION_OPTIONS = 'ImpersonateUser=%CURRENT_USER',
    CREDENTIAL = [OracleProxyCredential]
);

Alternatively, you can use TNS authentication.

Starting in SQL Server 2019 (15.x) Cumulative Update 19, CREATE EXTERNAL DATA SOURCE now supports the use of TNS files when connecting to Oracle. The CONNECTION_OPTIONS parameter was expanded and now uses TNSNamesFile and ServerName as variables to browse the tnsnames.ora file and establish connection with the server.

In the example below, during runtime SQL Server will search for the tnsnames.ora file location specified by TNSNamesFile and search for the host and network port specified by ServerName.

CREATE EXTERNAL DATA SOURCE [external_data_source_name]
WITH (
    LOCATION = N'oracle://XE',
    CREDENTIAL = [OracleCredentialTest],
    CONNECTION_OPTIONS = N'TNSNamesFile=C:\Temp\tnsnames.ora;ServerName=XE'
);

For additional examples to other data sources such as MongoDB, see Configure PolyBase to access external data in MongoDB.

B. Create external data source to reference Hadoop

To create an external data source to reference your Hortonworks HDP or Cloudera CDH Hadoop cluster, specify the machine name, or IP address of the Hadoop Namenode and port.

CREATE EXTERNAL DATA SOURCE MyHadoopCluster
WITH (
    LOCATION = 'hdfs://10.10.10.10:8050',
    TYPE = HADOOP
);

C. Create external data source to reference Hadoop with push-down enabled

Specify the RESOURCE_MANAGER_LOCATION option to enable push-down computation to Hadoop for PolyBase queries. Once enabled, PolyBase makes a cost-based decision to determine whether the query computation should be pushed to Hadoop.

CREATE EXTERNAL DATA SOURCE MyHadoopCluster
WITH (
    LOCATION = 'hdfs://10.10.10.10:8020',
    TYPE = HADOOP,
    RESOURCE_MANAGER_LOCATION = '10.10.10.10:8050'
);

D. Create external data source to reference Kerberos-secured Hadoop

To verify if the Hadoop cluster is Kerberos-secured, check the value of hadoop.security.authentication property in Hadoop core-site.xml. To reference a Kerberos-secured Hadoop cluster, you must specify a database scoped credential that contains your Kerberos username and password. The database master key is used to encrypt the database scoped credential secret.

-- Create a database master key if one does not already exist, using your own password.
-- This key is used to encrypt the credential secret in next step.
CREATE MASTER KEY ENCRYPTION BY PASSWORD = '<password>';

-- Create a database scoped credential with Kerberos user name and password.
CREATE DATABASE SCOPED CREDENTIAL HadoopUser1
    WITH IDENTITY = '<hadoop_user_name>',
    SECRET = '<hadoop_password>';

-- Create an external data source with CREDENTIAL option.
CREATE EXTERNAL DATA SOURCE MyHadoopCluster
WITH (
    LOCATION = 'hdfs://10.10.10.10:8050',
    CREDENTIAL = HadoopUser1,
    TYPE = HADOOP,
    RESOURCE_MANAGER_LOCATION = '10.10.10.10:8050'
);

E. Create external data source to access data in Azure Storage using the wasb:// interface

In this example, the external data source is an Azure V2 Storage account named logs. The storage container is called daily. The Azure Storage external data source is for data transfer only. It doesn't support predicate push-down. Hierarchical namespaces are not supported when accessing data via the wasb:// interface. Note that when connecting to the Azure Storage via the WASB[s] connector, authentication must be done with a storage account key, not with a shared access signature (SAS).

This example shows how to create the database scoped credential for authentication to an Azure V2 Storage account. Specify the Azure Storage account key in the database credential secret. You can specify any string in database scoped credential identity as it isn't used during authentication to Azure Storage.

-- Create a database master key if one does not already exist, using your own password.
-- This key is used to encrypt the credential secret in next step.
CREATE MASTER KEY ENCRYPTION BY PASSWORD = '<password>';

-- Create a database scoped credential with Azure storage account key as the secret.
CREATE DATABASE SCOPED CREDENTIAL AzureStorageCredential
    WITH IDENTITY = '<my_account>',
    SECRET = '<azure_storage_account_key>';

-- Create an external data source with CREDENTIAL option.
CREATE EXTERNAL DATA SOURCE MyAzureStorage
WITH (
    LOCATION = 'wasbs://daily@logs.blob.core.windows.net/',
    CREDENTIAL = AzureStorageCredential,
    TYPE = HADOOP
);

F. Create external data source to reference a SQL Server named instance via PolyBase connectivity

Applies to: SQL Server 2019 (15.x) and later

To create an external data source that references a named instance of SQL Server, use CONNECTION_OPTIONS to specify the instance name.

In the following example, WINSQL2019 is the host name and SQL2019 is the instance name. 'Server=%s\SQL2019' is the key value pair.

CREATE EXTERNAL DATA SOURCE SQLServerInstance2
WITH (
    LOCATION = 'sqlserver://WINSQL2019',
    CONNECTION_OPTIONS = 'Server=%s\SQL2019',
    CREDENTIAL = SQLServerCredentials
);

Alternatively, you can use a port to connect to a SQL Server default instance.

CREATE EXTERNAL DATA SOURCE SQLServerInstance2
WITH (
    LOCATION = 'sqlserver://WINSQL2019:58137',
    CREDENTIAL = SQLServerCredentials
);

G. Create external data source to reference a readable secondary replica of Always On availability group

Applies to: SQL Server 2019 (15.x) and later

To create an external data source that references a readable secondary replica of SQL Server, use CONNECTION_OPTIONS to specify the ApplicationIntent=ReadOnly. In addition, you will need to either set the availability database as Database={dbname} in CONNECTION_OPTIONS, or set the availability database as the default database of the login used for the database scoped credential. You will need to do this on all availability replicas of the availability group.

First, create the database scoped credential, storing credentials for a SQL authenticated login. The SQL ODBC Connector for PolyBase only supports basic authentication. Before you create a database scoped credential, the database must have a master key to protect the credential. For more information, see CREATE MASTER KEY. The following sample creates a database scoped credential, provide your own login and password.

CREATE DATABASE SCOPED CREDENTIAL SQLServerCredentials
    WITH IDENTITY = 'username',
    SECRET = 'password';

Next, create the new external data source.

Whether you included Database=dbname in the CONNECTION_OPTIONS or set the availability database as the default database for the login in the database scoped credential, you must still provide the database name via a three-part name in the CREATE EXTERNAL TABLE statement, within the LOCATION parameter. For an example, see CREATE EXTERNAL TABLE.

In the following example, WINSQL2019AGL is the availability group listener name and dbname is the name of the database to be the target of the CREATE EXTERNAL TABLE statement.

CREATE EXTERNAL DATA SOURCE SQLServerInstance2
WITH (
    LOCATION = 'sqlserver://WINSQL2019AGL',
    CONNECTION_OPTIONS = 'ApplicationIntent=ReadOnly; Database=dbname',
    CREDENTIAL = SQLServerCredentials
);

You can demonstrate the redirection behavior of the availability group by specifying ApplicationIntent and creating an external table on the system view sys.servers. In the following sample script, two external data sources are created, and one external table is created for each. Use the views to test which server is responding to the connection. Similar outcomes can also be achieved via the read-only routing feature. For more information, see Configure read-only routing for an Always On availability group.

CREATE EXTERNAL DATA SOURCE [DataSource_SQLInstanceListener_ReadOnlyIntent]
WITH (
    LOCATION = 'sqlserver://WINSQL2019AGL',
    CONNECTION_OPTIONS = 'ApplicationIntent=ReadOnly; Database=dbname',
    CREDENTIAL = [SQLServerCredentials]
);
GO

CREATE EXTERNAL DATA SOURCE [DataSource_SQLInstanceListener_ReadWriteIntent]
WITH (
    LOCATION = 'sqlserver://WINSQL2019AGL',
    CONNECTION_OPTIONS = 'ApplicationIntent=ReadWrite',
    CREDENTIAL = [SQLServerCredentials]
);
GO

Inside the database in the availability group, create a view to return sys.servers and the name of the local instance, which helps you identify which replica is responding to the query. For more information, see sys.servers.

CREATE VIEW vw_sys_servers
AS
SELECT [name]
FROM sys.servers
WHERE server_id = 0;
GO

Then, create an external table on the source instance:

CREATE EXTERNAL TABLE vw_sys_servers_ro (name SYSNAME NOT NULL)
WITH (
    DATA_SOURCE = [DataSource_SQLInstanceListener_ReadOnlyIntent],
    LOCATION = N'dbname.dbo.vw_sys_servers'
);
GO

CREATE EXTERNAL TABLE vw_sys_servers_rw (name SYSNAME NOT NULL)
WITH (
    DATA_SOURCE = [DataSource_SQLInstanceListener_ReadWriteIntent],
    LOCATION = N'dbname.dbo.vw_sys_servers'
);
GO

SELECT [name]
FROM dbo.vw_sys_servers_ro;--should return secondary replica instance

SELECT [name]
FROM dbo.vw_sys_servers_rw;--should return primary replica instance
GO

Examples: Bulk operations

Important

Do not add a trailing /, file name, or shared access signature parameters at the end of the LOCATION URL when configuring an external data source for bulk operations.

H. Create an external data source for bulk operations retrieving data from Azure Storage

Applies to: SQL Server 2017 (14.x) and SQL Server 2019 (15.x)

Use the following data source for bulk operations using BULK INSERT or OPENROWSET. The credential must set SHARED ACCESS SIGNATURE as the identity, mustn't have the leading ? in the SAS token, must have at least read permission on the file that should be loaded (for example srt=o&sp=r), and the expiration period should be valid (all dates are in UTC time). For more information on shared access signatures, see Using Shared Access Signatures (SAS).

CREATE DATABASE SCOPED CREDENTIAL AccessAzureInvoices
    WITH IDENTITY = 'SHARED ACCESS SIGNATURE',
    -- Remove ? from the beginning of the SAS token
    SECRET = '<azure_shared_access_signature>';

CREATE EXTERNAL DATA SOURCE MyAzureInvoices
WITH (
    LOCATION = 'https://newinvoices.blob.core.windows.net/week3',
    CREDENTIAL = AccessAzureInvoices,
    TYPE = BLOB_STORAGE
);

To see this example in use, see the BULK INSERT example.

I. Create external data source to access data in Azure Storage using the abfs:// interface

Applies to: SQL Server 2019 (15.x) CU11 and later

In this example, the external data source is an Azure Data Lake Storage Gen2 account logs, using the Azure Blob Filesystem driver (ABFS). The storage container is called daily. The Azure Data Lake Storage Gen2 external data source is for data transfer only, as predicate push-down is not supported.

This example shows how to create the database scoped credential for authentication to an Azure Data Lake Storage Gen2 account. Specify the Azure Storage account key in the database credential secret. You can specify any string in database scoped credential identity as it isn't used during authentication to Azure Storage.

-- Create a database master key if one does not already exist, using your own password.
-- This key is used to encrypt the credential secret in next step.
CREATE MASTER KEY ENCRYPTION BY PASSWORD = '<password>';

-- Create a database scoped credential with Azure storage account key as the secret.
CREATE DATABASE SCOPED CREDENTIAL AzureStorageCredential
    WITH IDENTITY = '<my_account>',
    SECRET = '<azure_storage_account_key>';

-- Create an external data source with CREDENTIAL option.
CREATE EXTERNAL DATA SOURCE MyAzureStorage
WITH (
    LOCATION = 'abfss://daily@logs.dfs.core.windows.net/',
    CREDENTIAL = AzureStorageCredential,
    TYPE = HADOOP
);

J. Create external data source using generic ODBC to PostgreSQL

As in previous examples, first create a database master key and database scoped credential. The database scoped credential will be used for the external data source. This example also assumes that a generic ODBC data provider for PostgreSQL is installed on the server.

In this example, the generic ODBC data provider is used to connect to a PostgreSQL database server in the same network, where the fully qualified domain name of the PostgreSQL server is POSTGRES1, using the default port of TCP 5432.

CREATE EXTERNAL DATA SOURCE POSTGRES1
WITH (
    LOCATION = 'odbc://POSTGRES1.domain:5432',
    CONNECTION_OPTIONS = 'Driver={PostgreSQL Unicode(x64)};',
    CREDENTIAL = postgres_credential
);

Next steps

Overview: SQL Server 2022

Applies to: SQL Server 2022 (16.x) and later

Creates an external data source for PolyBase queries. External data sources are used to establish connectivity and support these primary use cases:

  • Data virtualization and data load using PolyBase
  • Bulk load operations using BULK INSERT or OPENROWSET

Note

This syntax varies in different versions of SQL Server. Use the version selector dropdown to choose the appropriate version. This content applies to SQL Server 2022 (16.x) and later.

Syntax for SQL Server 2022 and later

For more information about the syntax conventions, see Transact-SQL syntax conventions.

CREATE EXTERNAL DATA SOURCE <data_source_name>
WITH
  ( [ LOCATION = '<prefix>://<path>[:<port>]' ]
    [ [ , ] CONNECTION_OPTIONS = '<key_value_pairs>'[,...]]
    [ [ , ] CREDENTIAL = <credential_name> ]
    [ [ , ] PUSHDOWN = { ON | OFF } ]
  )
[ ; ]

Arguments

data_source_name

Specifies the user-defined name for the data source. The name must be unique within the database in SQL Server.

LOCATION = '<prefix>://<path[:port]>'

Provides the connectivity protocol and path to the external data source.

External Data Source Connector location prefix Location path Supported locations by product / service Authentication
Azure Storage Account(V2) abs abs://<container_name>@<storage_account_name>.blob.core.windows.net/
or
abs://<storage_account_name>.blob.core.windows.net/<container_name>
Starting with SQL Server 2022 (16.x)
Hierarchical Namespace is supported.
Shared access signature (SAS)
Azure Data Lake Storage Gen2 adls adls://<container_name>@<storage_account_name>.dfs.core.windows.net/
or
adls://<storage_account_name>.dfs.core.windows.net/<container_name>
Starting with SQL Server 2022 (16.x) Shared access signature (SAS)
SQL Server sqlserver <server_name>[\<instance_name>][:port] Starting with SQL Server 2019 (15.x) SQL authentication only
Oracle oracle <server_name>[:port] Starting with SQL Server 2019 (15.x) Basic authentication only
Teradata teradata <server_name>[:port] Starting with SQL Server 2019 (15.x) Basic authentication only
MongoDB or Cosmos DB API for MongoDB mongodb <server_name>[:port] Starting with SQL Server 2019 (15.x) Basic authentication only
Generic ODBC odbc <server_name>[:port] Starting with SQL Server 2019 (15.x) - Windows only Basic authentication only
Bulk Operations https <storage_account>.blob.core.windows.net/<container> Starting with SQL Server 2017 (14.x) Shared access signature (SAS)
S3-compatible object storage s3 - S3-compatible: s3://<server_name>:<port>/
- AWS S3: s3://<bucket_name>.S3.amazonaws.com[:port]/<folder>
or s3://s3.amazonaws.com[:port]/<bucket_name>/<folder>
Starting with SQL Server 2022 (16.x) Basic or pass-through (STS) *

* Must be a database scoped credential, where the IDENTITY is hard-coded to IDENTITY = 'S3 Access Key' and the SECRET argument is in the format = '<AccessKeyID>:<SecretKeyID>' or use pass-through (STS) authorization. For more information, see Configure PolyBase to access external data in S3-compatible object storage.

Location path:

  • port = The port that the external data source is listening on. Optional in many cases, depending on network configuration.
  • <container_name> = the container of the storage account holding the data. Root containers are read-only, data can't be written back to the container.
  • <storage_account> = the storage account name of the Azure resource.
  • <server_name> = the host name.
  • <instance_name> = the name of the SQL Server named instance. Used if you have SQL Server Browser Service running on the target instance.
  • <ip_address>:<port> = For S3-compatible object storage only (starting with SQL Server 2022 (16.x)), the endpoint and port used to connect to the S3-compatible storage.
  • <bucket_name> = For S3-compatible object storage only (starting with SQL Server 2022 (16.x)), specific to the storage platform.
  • <region> = For S3-compatible object storage only (starting with SQL Server 2022 (16.x)), specific to the storage platform.
  • <folder> = Part of the storage path within the storage URL.

Additional notes and guidance when setting the location:

  • The SQL Server Database Engine doesn't verify the existence of the external data source when the object is created. To validate, create an external table using the external data source.
  • You can use the sqlserver connector to connect SQL Server 2019 (15.x) to another SQL Server or to Azure SQL Database.
  • Specify the Driver={<Name of Driver>} when connecting via ODBC.
  • The Hierarchical Namespace option for Azure Storage Accounts(V2) using the prefix adls is supported via Azure Data Lake Storage Gen2 in SQL Server 2022 (16.x).
  • SQL Server support for HDFS Cloudera (CDP) and Hortonworks (HDP) external data sources are retired and not included in SQL Server 2022 (16.x). There is no need to use the TYPE argument in SQL Server 2022 (16.x).
  • For more information on S3-compatible object storage and PolyBase starting with SQL Server 2022 (16.x), see Configure PolyBase to access external data in S3-compatible object storage. For an example of querying a parquet file within S3-compatible object storage, see Virtualize parquet file in a S3-compatible object storage with PolyBase.
  • Differing from previous versions, in SQL Server 2022 (16.x), the prefix used for Azure Storage Account (v2) changed from wasb[s] to abs.
  • Differing from previous versions, in SQL Server 2022 (16.x), the prefix used for Azure Data Lake Storage Gen2 changed from abfs[s] to adls.
  • For an example using PolyBase to virtualize a CSV file in Azure Storage, see Virtualize CSV file with PolyBase.
  • For an example using PolyBase to virtualize a delta table in ADLS Gen2, see Virtualize delta table with PolyBase.
  • SQL Server 2022 (16.x) fully supports two URL formats for both Azure Storage Account v2 (abs) and Azure Data Lake Gen2 (adls).
    • The LOCATION path can use the formats: <container>@<storage_account_name>.. (recommended) or <storage_account_name>../<container>. For example:
      • Azure Storage Account v2: abs://<container>@<storage_account_name>.blob.core.windows.net (recommended) or abs://<storage_account_name>.blob.core.windows.net/<container>.
      • Azure Data Lake Gen2 supports: adls://<container>@<storage_account_name>.blob.core.windows.net (recommended) or adls://<storage_account_name>.dfs.core.windows.net/<container>.

CONNECTION_OPTIONS = key_value_pair

Specified for SQL Server 2019 (15.x) and later. Specifies additional options when connecting over ODBC to an external data source. To use multiple connection options, separate them by a semi-colon.

Applies to generic ODBC connections, as well as built-in ODBC connectors for SQL Server, Oracle, Teradata, MongoDB, and Azure Cosmos DB API for MongoDB.

The key_value_pair is the keyword and the value for a specific connection option. The available keywords and values depend on the external data source type. The name of the driver is required as a minimum, but there are other options such as APP='<your_application_name>' or ApplicationIntent= ReadOnly|ReadWrite that are also useful to set and can assist with troubleshooting.

Possible key value pairs are specific to the driver. For more information for each provider, see CREATE EXTERNAL DATA SOURCE (Transact-SQL) CONNECTION_OPTIONS.

Starting in Applies to: SQL Server 2022 (16.x) cumulative update 2, additional keywords were introduced to support Oracle TNS files:

  • The keyword TNSNamesFile specifies the filepath to the tnsnames.ora file located on the Oracle server.
  • The keyword ServerName specifies the alias used inside the tnsnames.ora that will be used to replace the host name and the port.

PUSHDOWN = ON | OFF

Applies to: SQL Server 2019 (15.x) and later. States whether computation can be pushed down to the external data source. It is on by default.

PUSHDOWN is supported when connecting to SQL Server, Oracle, Teradata, MongoDB, the Azure Cosmos DB API for MongoDB, or ODBC at the external data source level.

Enabling or disabling push-down at the query level is achieved through a hint.

CREDENTIAL = credential_name

Specifies a database-scoped credential for authenticating to the external data source.

Additional notes and guidance when creating a credential:

There are multiple ways to create a shared access signature:

  • You can create an SAS token by navigating to the Azure portal -> <Your_Storage_Account> -> Shared access signature -> Configure permissions -> Generate SAS and connection string. For more information, see Generate a shared access signature.

  • You can create and configure an SAS with Azure Storage Explorer.

  • You can create an SAS programmatically via PowerShell, Azure CLI, .NET, and REST API. For more information, see Grant limited access to Azure Storage resources using shared access signatures (SAS).

  • The SAS token should be configured as follows:

    • When an SAS token is generated, it includes a question mark ('?') at the beginning of the token. Exclude the leading ? when configured as the SECRET.
    • Use a valid expiration period (all dates are in UTC time).
    • Grant at least read permission on the file that should be loaded (for example srt=o&sp=r). Multiple shared access signatures can be created for different use cases. Permissions should be granted as follows:
    Action Permission
    Read data from a file Read
    Read data from multiple files and subfolders Read and List
    Use Create External Table as Select (CETAS) Read, Create, List and Write
  • For Azure Blob Storage and Azure Data Lake Gen 2:

    • Allowed services: Blob must be selected to generate the SAS token
    • Allowed resource types: Container and Object must be selected to generate the SAS token

For an example of using a CREDENTIAL with S3-compatible object storage and PolyBase, see Configure PolyBase to access external data in S3-compatible object storage.

To create a database scoped credential, see CREATE DATABASE SCOPED CREDENTIAL (Transact-SQL).

Permissions

Requires CONTROL permission on database in SQL Server.

Locking

Takes a shared lock on the EXTERNAL DATA SOURCE object.

Security

PolyBase supports proxy based authentication for most external data sources. Create a database scoped credential to create the proxy account.

Upgrade to SQL Server 2022

Starting in SQL Server 2022 (16.x), Hadoop external data sources are no longer supported. It is required to manually recreate external data sources previously created with TYPE = HADOOP, and any external table that uses this external data source.

Users will also need to configure their external data sources to use new connectors when connecting to Azure Storage.

External Data Source From To
Azure Blob Storage wasb[s] abs
ADLS Gen2 abfs[s] adls

Examples

Important

For information on how to install and enable PolyBase, see Install PolyBase on Windows

A. Create external data source in SQL Server to reference Oracle

To create an external data source that references Oracle, ensure you have a database scoped credential. You may optionally also enable or disable push-down of computation against this data source.

-- Create a database master key if one does not already exist, using your own password.
-- This key is used to encrypt the credential secret in next step.
CREATE MASTER KEY ENCRYPTION BY PASSWORD = '<password>';

-- Create a database scoped credential with Azure storage account key as the secret.
CREATE DATABASE SCOPED CREDENTIAL OracleProxyAccount
    WITH IDENTITY = 'oracle_username',
    SECRET = 'oracle_password';

CREATE EXTERNAL DATA SOURCE MyOracleServer
WITH (
    LOCATION = 'oracle://145.145.145.145:1521',
    CREDENTIAL = OracleProxyAccount,
    PUSHDOWN = ON
);

Optionally, the external data source to Oracle can use proxy authentication to provide fine grain access control. A proxy user can be configured to have limited access compared to the user being impersonated.

CREATE DATABASE SCOPED CREDENTIAL [OracleProxyCredential]
    WITH IDENTITY = 'oracle_username',
    SECRET = 'oracle_password';

CREATE EXTERNAL DATA SOURCE [OracleSalesSrvr]
WITH (
    LOCATION = 'oracle://145.145.145.145:1521',
    CONNECTION_OPTIONS = 'ImpersonateUser=%CURRENT_USER',
    CREDENTIAL = [OracleProxyCredential]
);

Alternatively, you can authenticate using TNS.

Starting in Applies to: SQL Server 2022 (16.x) Cumulative Update 2, CREATE EXTERNAL DATA SOURCE now supports the use of TNS files when connecting to Oracle. The CONNECTION_OPTIONS parameter was expanded and now uses TNSNamesFile and ServerName as variables to browse the tnsnames.ora file and establish connection with the server.

In the example below, during runtime SQL Server will search for the tnsnames.ora file location specified by TNSNamesFile and search for the host and network port specified by ServerName.

CREATE EXTERNAL DATA SOURCE [external_data_source_name]
WITH (
    LOCATION = N'oracle://XE',
    CREDENTIAL = [OracleCredentialTest],
    CONNECTION_OPTIONS = N'TNSNamesFile=C:\Temp\tnsnames.ora;ServerName=XE'
);

B. Create external data source to reference a SQL Server named instance via PolyBase connectivity

Applies to: SQL Server 2019 (15.x) and later

To create an external data source that references a named instance of SQL Server, use CONNECTION_OPTIONS to specify the instance name.

First, create the database scoped credential, storing credentials for a SQL authenticated login. The SQL ODBC Connector for PolyBase only supports basic authentication. Before you create a database scoped credential, the database must have a master key to protect the credential. For more information, see CREATE MASTER KEY. The following sample creates a database scoped credential, provide your own login and password.

CREATE DATABASE SCOPED CREDENTIAL SQLServerCredentials
    WITH IDENTITY = 'username',
    SECRET = 'password';

In the following example, WINSQL2019 is the host name and SQL2019 is the instance name. 'Server=%s\SQL2019' is the key value pair.

CREATE EXTERNAL DATA SOURCE SQLServerInstance2
WITH (
    LOCATION = 'sqlserver://WINSQL2019',
    CONNECTION_OPTIONS = 'Server=%s\SQL2019',
    CREDENTIAL = SQLServerCredentials
);

Alternatively, you can use a port to connect to a SQL Server default instance.

CREATE EXTERNAL DATA SOURCE SQLServerInstance2
WITH (
    LOCATION = 'sqlserver://WINSQL2019:58137',
    CREDENTIAL = SQLServerCredentials
);

C. Create external data source to reference a readable secondary replica of Always On availability group

Applies to: SQL Server 2019 (15.x) and later

To create an external data source that references a readable secondary replica of SQL Server, use CONNECTION_OPTIONS to specify the ApplicationIntent=ReadOnly. In addition, you will need to either set the availability database as Database={dbname} in CONNECTION_OPTIONS, or set the availability database as the default database of the login used for the database scoped credential. You will need to do this on all availability replicas of the availability group.

First, create the database scoped credential, storing credentials for a SQL authenticated login. The SQL ODBC Connector for PolyBase only supports basic authentication. Before you create a database scoped credential, the database must have a master key to protect the credential. For more information, see CREATE MASTER KEY. The following sample creates a database scoped credential, provide your own login and password.

CREATE DATABASE SCOPED CREDENTIAL SQLServerCredentials
    WITH IDENTITY = 'username',
    SECRET = 'password';

Next, create the new external data source.

Whether you included Database=dbname in the CONNECTION_OPTIONS or set the availability database as the default database for the login in the database scoped credential, you must still provide the database name via a three-part name in the CREATE EXTERNAL TABLE statement, within the LOCATION parameter. For an example, see CREATE EXTERNAL TABLE.

In the following example, WINSQL2019AGL is the availability group listener name and dbname is the name of the database to be the target of the CREATE EXTERNAL TABLE statement.

CREATE EXTERNAL DATA SOURCE SQLServerInstance2
WITH (
    LOCATION = 'sqlserver://WINSQL2019AGL',
    CONNECTION_OPTIONS = 'ApplicationIntent=ReadOnly; Database=dbname',
    CREDENTIAL = SQLServerCredentials
);

You can demonstrate the redirection behavior of the availability group by specifying ApplicationIntent and creating an external table on the system view sys.servers. In the following sample script, two external data sources are created, and one external table is created for each. Use the views to test which server is responding to the connection. Similar outcomes can also be achieved via the read-only routing feature. For more information, see Configure read-only routing for an Always On availability group.

CREATE EXTERNAL DATA SOURCE [DataSource_SQLInstanceListener_ReadOnlyIntent]
WITH (
    LOCATION = 'sqlserver://WINSQL2019AGL',
    CONNECTION_OPTIONS = 'ApplicationIntent=ReadOnly; Database=dbname',
    CREDENTIAL = [SQLServerCredentials]
);
GO

CREATE EXTERNAL DATA SOURCE [DataSource_SQLInstanceListener_ReadWriteIntent]
WITH (
    LOCATION = 'sqlserver://WINSQL2019AGL',
    CONNECTION_OPTIONS = 'ApplicationIntent=ReadWrite',
    CREDENTIAL = [SQLServerCredentials]
);
GO

Inside the database in the availability group, create a view to return sys.servers and the name of the local instance, which helps you identify which replica is responding to the query. For more information, see sys.servers.

CREATE VIEW vw_sys_servers
AS
SELECT [name]
FROM sys.servers
WHERE server_id = 0;
GO

Then, create an external table on the source instance:

CREATE EXTERNAL TABLE vw_sys_servers_ro (name SYSNAME NOT NULL)
WITH (
    DATA_SOURCE = [DataSource_SQLInstanceListener_ReadOnlyIntent],
    LOCATION = N'dbname.dbo.vw_sys_servers'
);
GO

CREATE EXTERNAL TABLE vw_sys_servers_rw (name SYSNAME NOT NULL)
WITH (
    DATA_SOURCE = [DataSource_SQLInstanceListener_ReadWriteIntent],
    LOCATION = N'dbname.dbo.vw_sys_servers'
);
GO

SELECT [name]
FROM dbo.vw_sys_servers_ro;--should return secondary replica instance

SELECT [name]
FROM dbo.vw_sys_servers_rw;--should return primary replica instance
GO

D. Create external data source to query a parquet file in S3-compatible object storage via PolyBase

Applies to: SQL Server 2022 (16.x) and later

The following sample script creates an external data source s3_ds in the source user database in SQL Server. The external data source references the s3_dc database scoped credential.

CREATE DATABASE SCOPED CREDENTIAL s3_dc
    WITH IDENTITY = 'S3 Access Key', -- for S3-compatible object storage the identity must always be S3 Access Key
    SECRET = '<access_key_id>:<secret_key_id>' -- provided by the S3-compatible object storage
GO

CREATE EXTERNAL DATA SOURCE s3_ds
WITH (
    LOCATION = 's3://<ip_address>:<port>/',
    CREDENTIAL = s3_dc
);
GO

Verify the new external data source with sys.external_data_sources.

SELECT * FROM sys.external_data_sources;

Then, the following example demonstrates using T-SQL to query a parquet file stored in S3-compatible object storage via OPENROWSET query. For more information, see Virtualize parquet file in a S3-compatible object storage with PolyBase.

SELECT *
FROM OPENROWSET (
    BULK '/<bucket>/<parquet_folder>',
    FORMAT = 'PARQUET',
    DATA_SOURCE = 's3_ds'
) AS [cc];

E. Create external data source using generic ODBC to PostgreSQL

As in previous examples, first create a database master key and database scoped credential. The database scoped credential will be used for the external data source. This example also assumes that a generic ODBC data provider for PostgreSQL is installed on the server.

In this example, the generic ODBC data provider is used to connect to a PostgreSQL database server in the same network, where the fully qualified domain name of the PostgreSQL server is POSTGRES1, using the default port of TCP 5432.

CREATE EXTERNAL DATA SOURCE POSTGRES1
WITH (
    LOCATION = 'odbc://POSTGRES1.domain:5432',
    CONNECTION_OPTIONS = 'Driver={PostgreSQL Unicode(x64)};',
    CREDENTIAL = postgres_credential
);

Azure Storage

Create a shared access signature

For both Azure Blob Storage and Azure Data Lake Gen2, the supported authentication method is shared access signature (SAS). One simple way to generate a shared access signature token follow the steps that follow. For more information, see CREDENTIAL.

  1. Navigate to the Azure portal, and the desired Storage Account.
  2. Navigate to your desired Container under Data Storage menu.
  3. Select Shared access tokens.
  4. Choose the appropriate permission based on the desired action, for reference use the table bellow:
Action Permission
Read data from a file Read
Read data from multiple files and subfolders Read and List
Use Create External Table as Select (CETAS) Read, Create and Write
  1. Choose the token expiration date.
  2. Generate SAS token and URL.
  3. Copy the SAS token.

F. Create external data source to access data in Azure Blob Storage using the abs:// interface

Applies to: SQL Server 2022 (16.x) and later

Starting in SQL Server 2022 (16.x), use a new prefix abs for Azure Storage Account v2. The abs prefix supports authentication using SHARED ACCESS SIGNATURE. The abs prefix replaces wasb, used in previous versions. HADOOP is not longer supported, there is no more need to use TYPE = BLOB_STORAGE.

The Azure storage account key is no longer needed, instead using SAS Token as we can see in the following example:

-- Create a database master key if one does not already exist, using your own password.
-- This key is used to encrypt the credential secret in next step.
CREATE MASTER KEY ENCRYPTION BY PASSWORD = '<password>';
GO

CREATE DATABASE SCOPED CREDENTIAL AzureStorageCredentialv2
    WITH IDENTITY = 'SHARED ACCESS SIGNATURE', -- to use SAS the identity must be fixed as-is
    SECRET = '<Blob_SAS_Token>';
GO

-- Create an external data source with CREDENTIAL option.
CREATE EXTERNAL DATA SOURCE MyAzureStorage
WITH (
    LOCATION = 'abs://<container>@<storage_account_name>.blob.core.windows.net/',
    CREDENTIAL = AzureStorageCredentialv2,
);

For a more detailed example on how to access CSV files stored in Azure Blob Storage, see Virtualize CSV file with PolyBase.

G. Create external data source to access data in Azure Data Lake Gen2

Applies to: SQL Server 2022 (16.x) and later versions

Starting in SQL Server 2022 (16.x), use a new prefix adls for Azure Data Lake Gen2, replacing abfs used in previous versions. The adls prefix also supports SAS token as authentication method as shown in this example:

--Create a database scoped credential using SAS Token
CREATE DATABASE SCOPED CREDENTIAL datalakegen2
    WITH IDENTITY = 'SHARED ACCESS SIGNATURE',
    SECRET = '<DataLakeGen2_SAS_Token>';
GO

CREATE EXTERNAL DATA SOURCE data_lake_gen2_dfs
WITH (
    LOCATION = 'adls://<container>@<storage_account>.dfs.core.windows.net',
    CREDENTIAL = datalakegen2
);

For a more detailed example on how to access delta files stored on Azure Data Lake Gen2, see Virtualize delta table with PolyBase.

Examples: Bulk Operations

Important

Do not add a trailing /, file name, or shared access signature parameters at the end of the LOCATION URL when configuring an external data source for bulk operations.

H. Create an external data source for bulk operations retrieving data from Azure Storage

Applies to: SQL Server 2022 (16.x) and later.

Use the following data source for bulk operations using BULK INSERT or OPENROWSET. The credential must set SHARED ACCESS SIGNATURE as the identity, mustn't have the leading ? in the SAS token, must have at least read permission on the file that should be loaded (for example srt=o&sp=r), and the expiration period should be valid (all dates are in UTC time). For more information on shared access signatures, see Using Shared Access Signatures (SAS).

CREATE DATABASE SCOPED CREDENTIAL AccessAzureInvoices
WITH IDENTITY = 'SHARED ACCESS SIGNATURE',
    -- Remove ? from the beginning of the SAS token
    SECRET = '<azure_shared_access_signature>';

CREATE EXTERNAL DATA SOURCE MyAzureInvoices
WITH (
    LOCATION = 'abs://<container>@<storage_account_name>.blob.core.windows.net/',
    CREDENTIAL = AccessAzureInvoices,
);

Next steps

* SQL Database *  

 

Overview: Azure SQL Database

Applies to: Azure SQL Database

Creates an external data source for elastic queries. External data sources are used to establish connectivity and support these primary use cases:

  • Bulk load operations using BULK INSERT or OPENROWSET
  • Query remote SQL Database or Azure Synapse instances using SQL Database with elastic query
  • Query a sharded SQL Database using elastic query

Syntax

For more information about the syntax conventions, see Transact-SQL syntax conventions.

CREATE EXTERNAL DATA SOURCE <data_source_name>
WITH
  ( [ LOCATION = '<prefix>://<path>[:<port>]' ]
    [ [ , ] CREDENTIAL = <credential_name> ]
    [ [ , ] TYPE = { BLOB_STORAGE | RDBMS | SHARD_MAP_MANAGER } ]
    [ [ , ] DATABASE_NAME = '<database_name>' ]
    [ [ , ] SHARD_MAP_NAME = '<shard_map_manager>' ] )
[ ; ]

Arguments

data_source_name

Specifies the user-defined name for the data source. The name must be unique within the database in SQL Database.

LOCATION = '<prefix>://<path[:port]>'

Provides the connectivity protocol and path to the external data source.

External Data Source Connector location prefix Location path Availability
Bulk Operations https <storage_account>.blob.core.windows.net/<container>
Elastic Query (shard) Not required <shard_map_server_name>.database.windows.net
Elastic Query (remote) Not required <remote_server_name>.database.windows.net
EdgeHub edgehub edgehub:// Available in Azure SQL Edge only. EdgeHub is always local to the instance of Azure SQL Edge. As such there is no need to specify a path or port value.
Kafka kafka kafka://<kafka_bootstrap_server_name_ip>:<port_number> Available in Azure SQL Edge only.

Location path:

  • <shard_map_server_name> = The logical server name in Azure that is hosting the shard map manager. The DATABASE_NAME argument provides the database used to host the shard map and SHARD_MAP_NAME is used for the shard map itself.
  • <remote_server_name> = The target logical server name for the elastic query. The database name is specified using the DATABASE_NAME argument.

Additional notes and guidance when setting the location:

  • The Database Engine doesn't verify the existence of the external data source when the object is created. To validate, create an external table using the external data source.

CREDENTIAL = credential_name

Specifies a database-scoped credential for authenticating to the external data source.

Additional notes and guidance when creating a credential:

  • To load data from Azure Storage into Azure SQL Database, use a Shared Access Signature (SAS token).
  • CREDENTIAL is only required if the data has been secured. CREDENTIAL isn't required for data sets that allow anonymous access.
  • When the TYPE = BLOB_STORAGE, the credential must be created using SHARED ACCESS SIGNATURE as the identity.
  • When connecting to the Azure Storage via the WASB[s] connector, authentication must be done with a storage account key, not with a shared access signature (SAS).
  • When TYPE = HADOOP the credential must be created using the storage account key as the SECRET.
  • TYPE = BLOB_STORAGE is only permitted for bulk operations; you cannot create external tables for an external data source with TYPE = BLOB_STORAGE.

There are multiple ways to create a shared access signature:

  • You can create an SAS token by navigating to the Azure portal -> <Your_Storage_Account> -> Shared access signature -> Configure permissions -> Generate SAS and connection string. For more information, see Generate a shared access signature.

  • You can create and configure an SAS with Azure Storage Explorer.

  • You can create an SAS programmatically via PowerShell, Azure CLI, .NET, and REST API. For more information, see Grant limited access to Azure Storage resources using shared access signatures (SAS).

  • The SAS token should be configured as follows:

    • When an SAS token is generated, it includes a question mark ('?') at the beginning of the token. Exclude the leading ? when configured as the SECRET.
    • Use a valid expiration period (all dates are in UTC time).
    • Grant at least read permission on the file that should be loaded (for example srt=o&sp=r). Multiple shared access signatures can be created for different use cases. Permissions should be granted as follows:
    Action Permission
    Read data from a file Read
    Read data from multiple files and subfolders Read and List
    Use Create External Table as Select (CETAS) Read, Create and Write

For an example of using a CREDENTIAL with SHARED ACCESS SIGNATURE and TYPE = BLOB_STORAGE, see Create an external data source to execute bulk operations and retrieve data from Azure Storage into SQL Database

To create a database scoped credential, see CREATE DATABASE SCOPED CREDENTIAL (Transact-SQL).

TYPE = [ BLOB_STORAGE | RDBMS | SHARD_MAP_MANAGER]

Specifies the type of the external data source being configured. This parameter isn't always required.

  • Use RDBMS for cross-database queries using elastic query from SQL Database.
  • Use SHARD_MAP_MANAGER when creating an external data source when connecting to a sharded SQL Database.
  • Use BLOB_STORAGE when executing bulk operations with BULK INSERT, or OPENROWSET.

Important

Do not set TYPE if using any other external data source.

DATABASE_NAME = database_name

Configure this argument when the TYPE is set to RDBMS or SHARD_MAP_MANAGER.

TYPE Value of DATABASE_NAME
RDBMS The name of the remote database on the server provided using LOCATION
SHARD_MAP_MANAGER Name of the database operating as the shard map manager

For an example showing how to create an external data source where TYPE = RDBMS refer to Create an RDBMS external data source

SHARD_MAP_NAME = shard_map_name

Used when the TYPE argument is set to SHARD_MAP_MANAGER only to set the name of the shard map.

For an example showing how to create an external data source where TYPE = SHARD_MAP_MANAGER refer to Create a shard map manager external data source

Permissions

Requires CONTROL permission on database in Azure SQL Database.

Locking

Takes a shared lock on the EXTERNAL DATA SOURCE object.

Examples

A. Create a shard map manager external data source

To create an external data source to reference a SHARD_MAP_MANAGER, specify the SQL Database server name that hosts the shard map manager in SQL Database or a SQL Server database on a virtual machine.

CREATE MASTER KEY ENCRYPTION BY PASSWORD = '<password>';

CREATE DATABASE SCOPED CREDENTIAL ElasticDBQueryCred
    WITH IDENTITY = '<username>',
    SECRET = '<password>';

CREATE EXTERNAL DATA SOURCE MyElasticDBQueryDataSrc
WITH (
    TYPE = SHARD_MAP_MANAGER,
    LOCATION = '<server_name>.database.windows.net',
    DATABASE_NAME = 'ElasticScaleStarterKit_ShardMapManagerDb',
    CREDENTIAL = ElasticDBQueryCred,
    SHARD_MAP_NAME = 'CustomerIDShardMap'
);

For a step-by-step tutorial, see Getting started with elastic queries for sharding (horizontal partitioning).

B. Create an RDBMS external data source

To create an external data source to reference an RDBMS, specifies the SQL Database server name of the remote database in SQL Database.

CREATE MASTER KEY ENCRYPTION BY PASSWORD = '<password>';

CREATE DATABASE SCOPED CREDENTIAL SQL_Credential
    WITH IDENTITY = '<username>',
    SECRET = '<password>';

CREATE EXTERNAL DATA SOURCE MyElasticDBQueryDataSrc
WITH (
    TYPE = RDBMS,
    LOCATION = '<server_name>.database.windows.net',
    DATABASE_NAME = 'Customers',
    CREDENTIAL = SQL_Credential
);

For a step-by-step tutorial on RDBMS, see Getting started with cross-database queries (vertical partitioning).

Examples: Bulk operations

Important

Do not add a trailing /, file name, or shared access signature parameters at the end of the LOCATION URL when configuring an external data source for bulk operations.

C. Create an external data source for bulk operations retrieving data from Azure Storage

Use the following data source for bulk operations using BULK INSERT or OPENROWSET. The credential must set SHARED ACCESS SIGNATURE as the identity, mustn't have the leading ? in the SAS token, must have at least read permission on the file that should be loaded (for example srt=o&sp=r), and the expiration period should be valid (all dates are in UTC time). For more information on shared access signatures, see Using Shared Access Signatures (SAS).

CREATE DATABASE SCOPED CREDENTIAL AccessAzureInvoices
    WITH IDENTITY = 'SHARED ACCESS SIGNATURE',
    -- Remove ? from the beginning of the SAS token
    SECRET = '******srt=sco&sp=rwac&se=2017-02-01T00:55:34Z&st=2016-12-29T16:55:34Z***************';

CREATE EXTERNAL DATA SOURCE MyAzureInvoices
WITH (
    LOCATION = 'https://newinvoices.blob.core.windows.net/week3',
    CREDENTIAL = AccessAzureInvoices,
    TYPE = BLOB_STORAGE
);

To see this example in use, see BULK INSERT.

Examples: Azure SQL Edge

Important

For information on configuring external data for Azure SQL Edge, see Data streaming in Azure SQL Edge.

A. Create external data source to reference Kafka

Applies to: Azure SQL Edge only

In this example, the external data source is a Kafka server with IP address xxx.xxx.xxx.xxx and listening on port 1900. The Kafka external data source is only for data streaming and does not support predicate push down.

-- Create an External Data Source for Kafka
CREATE EXTERNAL DATA SOURCE MyKafkaServer
    WITH (LOCATION = 'kafka://xxx.xxx.xxx.xxx:1900');

B. Create external data source to reference EdgeHub

Applies to: Azure SQL Edge only

In this example, the external data source is a EdgeHub running on the same edge device as Azure SQL Edge. The edgeHub external data source is only for data streaming and does not support predicate push down.

-- Create an External Data Source for Kafka
CREATE EXTERNAL DATA SOURCE MyEdgeHub
    WITH (LOCATION = 'edgehub://');

Next steps

* Azure Synapse
Analytics *
 

 

Overview: Azure Synapse Analytics

Applies to: Azure Synapse Analytics

Creates an external data source for data virtualization. External data sources are used to establish connectivity and support the primary use case of data virtualization and data loading from external data sources. For more information, see Use external tables with Synapse SQL.

Important

To create an external data source to query a Azure Synapse Analytics resource using Azure SQL Database with elastic query, see SQL Database.

Syntax

For more information about the syntax conventions, see Transact-SQL syntax conventions.

CREATE EXTERNAL DATA SOURCE <data_source_name>
WITH
( [ LOCATION = '<prefix>://<path>[:<port>]' ]
  [ [ , ] CREDENTIAL = <credential_name> ]
  [ [ , ] TYPE = HADOOP ]
)
[ ; ]

Arguments

data_source_name

Specifies the user-defined name for the data source. The name must be unique within the Azure SQL Database in Azure Synapse Analytics.

LOCATION = '<prefix>://<path>'

Provides the connectivity protocol and path to the external data source.

External Data Source Connector location prefix Location path
Data Lake Storage* Gen1 adl <storage_account>.azuredatalake.net
Data Lake Storage Gen2 abfs[s] <container>@<storage_account>.dfs.core.windows.net
Azure Blob Storage wasbs <container>@<storage_account>.blob.core.windows.net
Azure Blob Storage https <storage_account>.blob.core.windows.net/<container>/subfolders
Data Lake Storage Gen1 http[s] <storage_account>.azuredatalakestore.net/webhdfs/v1
Data Lake Storage Gen2 http[s] <storage_account>.dfs.core.windows.net/<container>/subfolders
Data Lake Storage Gen2 wasb[s] <container>@<storage_account>.blob.core.windows.net

* Microsoft Azure Data Lake Storage Gen1 has limited support, Gen2 is recommended for all new development.

External Data Source Connector location prefix Dedicated SQL pools: PolyBase Dedicated SQL pools: native* Serverless SQL pools
Data Lake Storage** Gen1 adl No No Yes
Data Lake Storage Gen2 abfs[s] Yes Yes Yes
Azure Blob Storage wasbs Yes Yes*** Yes
Azure Blob Storage https No Yes Yes
Data Lake Storage Gen1 http[s] No No Yes
Data Lake Storage Gen2 http[s] Yes Yes Yes
Data Lake Storage Gen2 wasb[s] Yes Yes Yes

* Serverless and dedicated SQL pools in Azure Synapse Analytics use different code bases for data virtualization. Serverless SQL pools support a native data virtualization technology. Dedicated SQL pools support both native and PolyBase data virtualization. PolyBase data virtualization is used when the EXTERNAL DATA SOURCE is created with TYPE=HADOOP.

** Microsoft Azure Data Lake Storage Gen1 has limited support, Gen2 is recommended for all new development.

*** The more secure wasbs connector is recommended over wasb. Only native data virtualization in dedicated SQL pools (where TYPE does not equal HADOOP) support wasb.

Location path:

  • <container> = the container of the storage account holding the data. Root containers are read-only, data can't be written back to the container.
  • <storage_account> = the storage account name of the Azure resource.

Additional notes and guidance when setting the location:

  • The default option is to use enable secure SSL connections when provisioning Azure Data Lake Storage Gen2. When this is enabled, you must use abfss when a secure TLS/SSL connection is selected. Note that abfss works for unsecure TLS connections as well. For more information, see the Azure Blob Filesystem driver (ABFS).
  • Azure Synapse doesn't verify the existence of the external data source when the object is created. To validate, create an external table using the external data source.
  • Use the same external data source for all tables when querying Hadoop to ensure consistent querying semantics.
  • https: prefix enables you to use subfolder in the path. https is not available for all data access methods.
  • wasbs is recommended as data will be sent using a secure TLS connection.
  • Hierarchical Namespaces aren't supported with Azure V2 Storage Accounts when accessing data using the legacy wasb:// interface, but using wasbs:// supports Hierarchical Namespaces.

CREDENTIAL = credential_name

Optional. Specifies a database scoped credential for authenticating to the external data source. External data source without credential can access public storage account or use the caller's Microsoft Entra identity to access files on Azure storage.

Additional notes and guidance when creating a credential:

  • To load data from Azure Storage or Azure Data Lake Store (ADLS) Gen2 into Azure Synapse Analytics, use an Azure Storage Key.
  • CREDENTIAL is only required if the data has been secured. CREDENTIAL isn't required for data sets that allow anonymous access.

To create a database scoped credential, see CREATE DATABASE SCOPED CREDENTIAL (Transact-SQL).

  • In serverless SQL pool, database-scoped credentials can specify workspace managed identity, service principal name, or shared access signature (SAS) token. Access via a user identity, also known as Microsoft Entra passthrough, is also possible in the databased-scoped credential, as is anonymous access to publicly available storage. For more information, see Supported storage authorization types.

  • In dedicated SQL pool, database scoped credentials can specify shared access signature (SAS) token, storage access key, service principal, workspace managed identity, or Microsoft Entra passthrough.

TYPE = HADOOP

Optional, not recommended.

You can only specify TYPE with dedicated SQL pools. HADOOP is the only allowed value when specified. External data sources with TYPE=HADOOP are available only in dedicated SQL pools.

Use HADOOP for legacy implementations, otherwise it is recommended to use the newer native data access. Do not specify the TYPE argument to use the newer native data access.

For an example of using TYPE = HADOOP to load data from Azure Storage, see Create external data source to reference Azure Data Lake Store Gen 1 or 2 using a service principal.

Serverless and dedicated SQL pools in Azure Synapse Analytics use different code bases for data virtualization. Serverless SQL pools support a native data virtualization technology. Dedicated SQL pools support both native and PolyBase data virtualization. PolyBase data virtualization is used when the EXTERNAL DATA SOURCE is created with TYPE=HADOOP.

Permissions

Requires CONTROL permission on the database.

Locking

Takes a shared lock on the EXTERNAL DATA SOURCE object.

Security

Most external data sources support proxy based authentication, using a database-scoped credential to create the proxy account.

Shared Access Signature (SAS) keys are supported for authenticating to Azure Data Lake Store Gen 2 Storage Accounts. Customers who want to authenticate by using a Shared Access Signature must create a database scoped credential where IDENTITY = "Shared Access Signature" and enter a SAS token as the secret.

If you create a database scoped credential where IDENTITY = "Shared Access Signature" and use a storage key value as the secret, you'll get the following error message:

'HdfsBridge::isDirExist - Unexpected error encountered checking whether directory exists or not: AbfsRestOperationException: Operation failed: "Server failed to authenticate the request. Please refer to the information in the www-authenticate header.", 401, HEAD, [Storage path URL]'

Examples

A. Create external data source to access data in Azure Storage using the wasb:// interface

In this example, the external data source is an Azure Storage account V2 named logs. The storage container is called daily. The Azure Storage external data source is for data transfer only. It doesn't support predicate push-down. Hierarchical namespaces are not supported when accessing data via the wasb:// interface. Note that when connecting to the Azure Storage via the WASB[s] connector, authentication must be done with a storage account key, not with a shared access signature (SAS).

This example uses the legacy HADOOP Java-based access method. The following sample shows how to create the database scoped credential for authentication to Azure Storage. Specify the Azure Storage account key in the database credential secret. You can specify any string in database scoped credential identity as it isn't used during authentication to Azure storage.

-- Create a database master key if one does not already exist, using your own password.
-- This key is used to encrypt the credential secret in next step.
CREATE MASTER KEY ENCRYPTION BY PASSWORD = '<password>';

-- Create a database scoped credential with Azure storage account key as the secret.
CREATE DATABASE SCOPED CREDENTIAL AzureStorageCredential
    WITH IDENTITY = '<my_account>',
    SECRET = '<azure_storage_account_key>';

-- Create an external data source with CREDENTIAL option.
CREATE EXTERNAL DATA SOURCE MyAzureStorage
WITH (
    LOCATION = 'wasbs://daily@logs.blob.core.windows.net/',
    CREDENTIAL = AzureStorageCredential,
    TYPE = HADOOP
);

B. Create external data source to reference Azure Data Lake Store Gen 1 or 2 using a service principal

Azure Data Lake Store connectivity can be based on your ADLS URI and your Microsoft Entra application's service principal. Documentation for creating this application can be found at Data lake store authentication using Microsoft Entra ID.

-- If you do not have a Master Key on your DW you will need to create one.
CREATE MASTER KEY ENCRYPTION BY PASSWORD = '<password>';

-- These values come from your Microsoft Entra application used to authenticate to ADLS
CREATE DATABASE SCOPED CREDENTIAL ADLS_credential
WITH
    -- IDENTITY = '<clientID>@<OAuth2.0TokenEndPoint>' ,
    IDENTITY = '536540b4-4239-45fe-b9a3-629f97591c0c@https://login.microsoftonline.com/42f988bf-85f1-41af-91ab-2d2cd011da47/oauth2/token',
    -- SECRET = '<KEY>'
    SECRET = 'BjdIlmtKp4Fpyh9hIvr8HJlUida/seM5kQ3EpLAmeDI=';

-- For Gen 1 - Create an external data source
-- TYPE: HADOOP - PolyBase uses Hadoop APIs to access data in Azure Data Lake Storage.
-- LOCATION: Provide Data Lake Storage Gen 1 account name and URI
-- CREDENTIAL: Provide the credential created in the previous step
CREATE EXTERNAL DATA SOURCE AzureDataLakeStore
WITH (
    LOCATION = 'adl://newyorktaxidataset.azuredatalakestore.net',
    CREDENTIAL = ADLS_credential,
    TYPE = HADOOP
);

-- For Gen2 - Create an external data source
-- TYPE: HADOOP - PolyBase uses Hadoop APIs to access data in Azure Data Lake Storage.
-- LOCATION: Provide Data Lake Storage Gen2 account name and URI
-- CREDENTIAL: Provide the credential created in the previous step
CREATE EXTERNAL DATA SOURCE AzureDataLakeStore
WITH (
    -- Note the abfss endpoint when your account has secure transfer enabled
    LOCATION = 'abfss://data@newyorktaxidataset.dfs.core.windows.net',
    CREDENTIAL = ADLS_credential,
    TYPE = HADOOP
);

C. Create external data source to reference Azure Data Lake Store Gen2 using the storage account key

-- If you do not have a Master Key on your DW you will need to create one.
CREATE MASTER KEY ENCRYPTION BY PASSWORD = '<password>';

CREATE DATABASE SCOPED CREDENTIAL ADLS_credential
WITH
    -- IDENTITY = '<storage_account_name>' ,
    IDENTITY = 'newyorktaxidata',
    -- SECRET = '<storage_account_key>'
    SECRET = 'yz5N4+bxSb89McdiysJAzo+9hgEHcJRJuXbF/uC3mhbezES/oe00vXnZEl14U0lN3vxrFKsphKov16C0w6aiTQ==';

-- Note this example uses a Gen2 secured endpoint (abfss)
CREATE EXTERNAL DATA SOURCE < data_source_name >
WITH (
    LOCATION = 'abfss://2013@newyorktaxidataset.dfs.core.windows.net',
    CREDENTIAL = ADLS_credential,
    TYPE = HADOOP
);

D. Create external data source to Azure Data Lake Store Gen2 using abfs://

There is no need to specify SECRET when connecting to Azure Data Lake Store Gen2 account with Managed Identity mechanism.

-- If you do not have a Master Key on your DW you will need to create one
CREATE MASTER KEY ENCRYPTION BY PASSWORD = '<password>';

--Create database scoped credential with **IDENTITY = 'Managed Service Identity'**
CREATE DATABASE SCOPED CREDENTIAL msi_cred
WITH IDENTITY = 'Managed Service Identity';

--Create external data source with abfss:// scheme for connecting to your Azure Data Lake Store Gen2 account
CREATE EXTERNAL DATA SOURCE ext_datasource_with_abfss
WITH (
    TYPE = HADOOP,
    LOCATION = 'abfss://myfile@mystorageaccount.dfs.core.windows.net',
    CREDENTIAL = msi_cred
);

Next steps

* Analytics
Platform System (PDW) *
 

 

Overview: Analytics Platform System

Applies to: Analytics Platform System (PDW)

Creates an external data source for PolyBase queries. External data sources are used to establish connectivity and support the following use case: Data virtualization and data load using PolyBase.

Syntax

For more information about the syntax conventions, see Transact-SQL syntax conventions.

CREATE EXTERNAL DATA SOURCE <data_source_name>
WITH
  ( [ LOCATION = '<prefix>://<path>[:<port>]' ]
    [ [ , ] CREDENTIAL = <credential_name> ]
    [ [ , ] TYPE = HADOOP ]
    [ [ , ] RESOURCE_MANAGER_LOCATION = '<resource_manager>[:<port>]' )
[ ; ]

Arguments

data_source_name

Specifies the user-defined name for the data source. The name must be unique within the server in Analytics Platform System (PDW).

LOCATION = '<prefix>://<path[:port]>'

Provides the connectivity protocol and path to the external data source.

External Data Source Connector location prefix Location path
Cloudera CDH or Hortonworks HDP hdfs <Namenode>[:port]
Azure Storage Account wasb[s] <container>@<storage_account>.blob.core.windows.net

Location path:

  • <Namenode> = the machine name, name service URI, or IP address of the Namenode in the Hadoop cluster. PolyBase must resolve any DNS names used by the Hadoop cluster.
  • port = The port that the external data source is listening on. In Hadoop, the port can be found using the fs.defaultFS configuration parameter. The default is 8020.
  • <container> = the container of the storage account holding the data. Root containers are read-only, data can't be written back to the container.
  • <storage_account> = the storage account name of the Azure resource.

Additional notes and guidance when setting the location:

  • The PDW engine doesn't verify the existence of the external data source when the object is created. To validate, create an external table using the external data source.
  • Use the same external data source for all tables when querying Hadoop to ensure consistent querying semantics.
  • wasbs is recommended as data will be sent using a secure TLS connection.
  • Hierarchical Namespaces are not supported when used with Azure Storage accounts over wasb://.
  • To ensure successful PolyBase queries during a Hadoop Namenode fail-over, consider using a virtual IP address for the Namenode of the Hadoop cluster. If you don't, execute an ALTER EXTERNAL DATA SOURCE command to point to the new location.

CREDENTIAL = credential_name

Specifies a database-scoped credential for authenticating to the external data source.

Additional notes and guidance when creating a credential:

  • To load data from Azure Storage into Azure Synapse or PDW, use an Azure Storage Key.
  • CREDENTIAL is only required if the data has been secured. CREDENTIAL isn't required for data sets that allow anonymous access.

TYPE = [ HADOOP ]

Specifies the type of the external data source being configured. This parameter isn't always required.

  • Use HADOOP when the external data source is Cloudera CDH, Hortonworks HDP, or Azure Storage.

For an example of using TYPE = HADOOP to load data from Azure Storage, see Create external data source to reference Hadoop.

RESOURCE_MANAGER_LOCATION = 'ResourceManager_URI[:port]'

In SQL Server 2019 (15.x), do not specify RESOURCE_MANAGER_LOCATION unless connecting to Cloudera CDH, Hortonworks HDP, an Azure Storage account.

Configure this optional value when connecting to Cloudera CDH, Hortonworks HDP, or an Azure Storage account only. For a complete list of supported Hadoop versions, see PolyBase Connectivity Configuration (Transact-SQL).

When the RESOURCE_MANAGER_LOCATION is defined, the query optimizer makes a cost-based decision to improve performance. A MapReduce job can be used to push down the computation to Hadoop. Specifying the RESOURCE_MANAGER_LOCATION can significantly reduce the volume of data transferred between Hadoop and SQL, which can lead to improved query performance.

If the Resource Manager isn't specified, pushing compute to Hadoop is disabled for PolyBase queries. Create external data source to reference Hadoop with push-down enabled provides a concrete example and further guidance.

The RESOURCE_MANAGER_LOCATION value is not validated when you create the external data source. Entering an incorrect value may cause query failure at execution time whenever push-down is attempted as the provided value would not be able to resolve.

In order for PolyBase to function correctly with a Hadoop external data source, the ports for the following Hadoop cluster components must be open:

  • HDFS ports
    • Namenode
    • Datanode
  • Resource Manager
    • Job submission
    • Job history

If the port isn't specified, the default value is chosen using the current setting for 'hadoop connectivity' configuration.

Hadoop Connectivity Default Resource Manager Port
1 50300
2 50300
3 8021
4 8032
5 8050
6 8032
7 8050

The following table shows the default ports for these components. Note that there is Hadoop version dependency as well as the possibility of custom configuration that doesn't use the default port assignment.

Hadoop cluster component Default Port
NameNode 8020
DataNode (Data transfer, non-privilege IPC port) 50010
DataNode (Data transfer, privilege IPC port) 1019
Resource Manager Job Submission (Hortonworks 1.3) 50300
Resource Manager Job Submission (Cloudera 4.3) 8021
Resource Manager Job Submission (Hortonworks 2.0 on Windows, Cloudera 5.x on Linux) 8032
Resource Manager Job Submission (Hortonworks 2.x, 3.0 on Linux, Hortonworks 2.1-3 on Windows) 8050
Resource Manager Job History 10020

Permissions

Requires CONTROL permission on database in Analytics Platform System (PDW).

Note

In previous releases of PDW, create external data source required ALTER ANY EXTERNAL DATA SOURCE permissions.

Locking

Takes a shared lock on the EXTERNAL DATA SOURCE object.

Security

PolyBase supports proxy based authentication for most external data sources. Create a database scoped credential to create the proxy account.

An SAS token with type HADOOP is unsupported. It's only supported with type = BLOB_STORAGE when a storage account access key is used instead. Attempting to create an external data source with type HADOOP and a SAS credential fails with the following error:

Msg 105019, Level 16, State 1 - EXTERNAL TABLE access failed due to internal error: 'Java exception raised on call to HdfsBridge_Connect. Java exception message: Parameters provided to connect to the Azure storage account are not valid.: Error [Parameters provided to connect to the Azure storage account are not valid.] occurred while accessing external file.'

Examples

A. Create external data source to reference Hadoop

To create an external data source to reference your Hortonworks HDP or Cloudera CDH, specify the machine name, or IP address of the Hadoop Namenode and port.

CREATE EXTERNAL DATA SOURCE MyHadoopCluster
WITH (
    LOCATION = 'hdfs://10.10.10.10:8050',
    TYPE = HADOOP
);

B. Create external data source to reference Hadoop with push-down enabled

Specify the RESOURCE_MANAGER_LOCATION option to enable push-down computation to Hadoop for PolyBase queries. Once enabled, PolyBase makes a cost-based decision to determine whether the query computation should be pushed to Hadoop.

CREATE EXTERNAL DATA SOURCE MyHadoopCluster
WITH (
    LOCATION = 'hdfs://10.10.10.10:8020',
    TYPE = HADOOP,
    RESOURCE_MANAGER_LOCATION = '10.10.10.10:8050'
);

C. Create external data source to reference Kerberos-secured Hadoop

To verify if the Hadoop cluster is Kerberos-secured, check the value of hadoop.security.authentication property in Hadoop core-site.xml. To reference a Kerberos-secured Hadoop cluster, you must specify a database scoped credential that contains your Kerberos username and password. The database master key is used to encrypt the database scoped credential secret.

-- Create a database master key if one does not already exist, using your own password.
-- This key is used to encrypt the credential secret in next step.
CREATE MASTER KEY ENCRYPTION BY PASSWORD = '<password>';

-- Create a database scoped credential with Kerberos user name and password.
CREATE DATABASE SCOPED CREDENTIAL HadoopUser1
    WITH IDENTITY = '<hadoop_user_name>',
    SECRET = '<hadoop_password>';

-- Create an external data source with CREDENTIAL option.
CREATE EXTERNAL DATA SOURCE MyHadoopCluster
WITH (
    LOCATION = 'hdfs://10.10.10.10:8050',
    CREDENTIAL = HadoopUser1,
    TYPE = HADOOP,
    RESOURCE_MANAGER_LOCATION = '10.10.10.10:8050'
);

D. Create external data source to access data in Azure Storage using the wasb:// interface

In this example, the external data source is an Azure V2 Storage account named logs. The storage container is called daily. The Azure Storage external data source is for data transfer only. It doesn't support predicate push-down. Hierarchical namespaces are not supported when accessing data via the wasb:// interface. Note that when connecting to the Azure Storage via the WASB[s] connector, authentication must be done with a storage account key, not with a shared access signature (SAS).

This example shows how to create the database scoped credential for authentication to Azure storage. Specify the Azure storage account key in the database credential secret. You can specify any string in database scoped credential identity as it isn't used during authentication to Azure storage.

-- Create a database master key if one does not already exist, using your own password.
-- This key is used to encrypt the credential secret in next step.
CREATE MASTER KEY ENCRYPTION BY PASSWORD = '<password>';

-- Create a database scoped credential with Azure storage account key as the secret.
CREATE DATABASE SCOPED CREDENTIAL AzureStorageCredential
    WITH IDENTITY = '<my_account>',
        SECRET = '<azure_storage_account_key>';

-- Create an external data source with CREDENTIAL option.
CREATE EXTERNAL DATA SOURCE MyAzureStorage
WITH (
    LOCATION = 'wasbs://daily@logs.blob.core.windows.net/',
    CREDENTIAL = AzureStorageCredential,
    TYPE = HADOOP
);

Next steps

* SQL Managed Instance *  

Overview: Azure SQL Managed Instance

Applies to: Azure SQL Managed Instance

Creates an external data source in Azure SQL Managed Instance. For complete information, see Data virtualization with Azure SQL Managed Instance.

Data virtualization in Azure SQL Managed Instance provides access to external data in a variety of file formats via the OPENROWSET T-SQL syntax or the CREATE EXTERNAL TABLE T-SQL syntax.

Syntax

For more information about the syntax conventions, see Transact-SQL syntax conventions.

CREATE EXTERNAL DATA SOURCE <data_source_name>
WITH
  ( [ LOCATION = '<prefix>://<path>[:<port>]' ]
    [ [ , ] CREDENTIAL = <credential_name> ]
  )
[ ; ]

Arguments

data_source_name

Specifies the user-defined name for the data source. The name must be unique within the database.

LOCATION = '<prefix>://<path[:port]>'

Provides the connectivity protocol and path to the external data source.

External Data Source Location prefix Location path
Azure Blob Storage abs abs://<container>@<storage_account>.blob.core.windows.net/<path>/<file_name>
Azure Data Lake Service Gen2 adls adls://<container>@<storage_account>.dfs.core.windows.net/<path>/<file_name>

The Database Engine doesn't verify the existence of the external data source when the object is created. To validate, create an external table using the external data source.

Do not add a trailing /, file name, or shared access signature parameters at the end of the LOCATION URL when configuring an external data source for bulk operations.

CREDENTIAL = credential_name

Specifies a database-scoped credential for authenticating to the external data source.

Additional notes and guidance when creating a credential:

  • To load data from Azure Storage into Azure SQL Managed Instance, use a Shared Access Signature (SAS token).
  • CREDENTIAL is only required if the data has been secured. CREDENTIAL isn't required for data sets that allow anonymous access.
  • If a credential is required, the credential must be created using Managed Identity or SHARED ACCESS SIGNATURE as the IDENTITY. To create a database scoped credential, see CREATE DATABASE SCOPED CREDENTIAL (Transact-SQL).

To use the managed service identity for the database scoped credential:

  • Specify WITH IDENTITY = 'Managed Identity'

    • Use the system-assigned managed service identity of the Azure SQL Managed Instance, which must be enabled if it is to be used for this purpose.
  • Grant the Reader Azure RBAC role to the system assigned managed service identity of the Azure SQL Managed Instance to the necessary Azure Blob Storage containers. For example, via the Azure portal, see Assign Azure roles using the Azure portal.

To create a shared access signature (SAS) for the database scoped credential:

  • Specify WITH IDENTITY = 'SHARED ACCESS SIGNATURE', SECRET = ...

  • There are multiple ways to create a shared access signature:

  • The SAS token should be configured as follows:

    • When an SAS token is generated, it includes a question mark ('?') at the beginning of the token. Exclude the leading ? when configured as the SECRET.
    • Use a valid expiration period (all dates are in UTC time).
    • Grant at least read permission on the file that should be loaded (for example srt=o&sp=r). Multiple shared access signatures can be created for different use cases. Permissions should be granted as follows:
    Action Permission
    Read data from a file Read
    Read data from multiple files and subfolders Read and List
    Use Create External Table as Select (CETAS) Read, Create and Write

Permissions

Requires CONTROL permission on database in Azure SQL Managed Instance.

Locking

Takes a shared lock on the EXTERNAL DATA SOURCE object.

Examples

For more examples, see Data virtualization with Azure SQL Managed Instance.

A. Query external data from Azure SQL Managed Instance with OPENROWSET or an external table

For more examples, see Create external data source or see Data virtualization with Azure SQL Managed Instance.

  1. Create the database master key, if it doesn't exist.

    -- Optional: Create MASTER KEY if it doesn't exist in the database:
    CREATE MASTER KEY ENCRYPTION BY PASSWORD = '<Strong Password>'
    GO
    
  2. Create the database scoped credential using a SAS token. You can also use a managed identity.

    CREATE DATABASE SCOPED CREDENTIAL MyCredential
    WITH IDENTITY = 'SHARED ACCESS SIGNATURE',
    SECRET = '<KEY>' ; --Removing leading '?'
    GO
    
  3. Create the external data source using the credential.

    --Create external data source pointing to the file path, and referencing database-scoped credential:
    CREATE EXTERNAL DATA SOURCE MyPrivateExternalDataSource
    WITH (
        LOCATION = 'abs://public@pandemicdatalake.blob.core.windows.net/curated/covid-19/bing_covid-19_data/latest',
        CREDENTIAL = [MyCredential]
    );
    
  4. Query parquet data file in the external data source using the OPENROWSET T-SQL syntax, relying on schema inference to quickly explore data without knowing the schema.

    --Query data with OPENROWSET, relying on schema inference.
    SELECT TOP 10 *
    FROM OPENROWSET (
        BULK 'bing_covid-19_data.parquet',
        DATA_SOURCE = 'MyExternalDataSource',
        FORMAT = 'parquet'
    ) AS filerows;
    
  5. Or, query data using OPENROWSET the WITH clause, instead of relying on schema inference, which may query execution cost. On a CSV, schema inference is not supported.

    --Or, query data using the WITH clause on a CSV, where schema inference is not supported
    SELECT TOP 10 id,
        updated,
        confirmed,
        confirmed_change
    FROM OPENROWSET (
        BULK 'bing_covid-19_data.csv', DATA_SOURCE = 'MyExternalDataSource',
        FORMAT = 'CSV',
        FIRSTROW = 2
    ) WITH (
        id INT,
        updated DATE,
        confirmed INT,
        confirmed_change INT
    ) AS filerows;
    
  6. Or, create an EXTERNAL FILE FORMAT and an EXTERNAL TABLE, to query the data as a local table.

    -- Or, create an EXTERNAL FILE FORMAT and an EXTERNAL TABLE
    --Create external file format
    CREATE EXTERNAL FILE FORMAT DemoFileFormat
        WITH (FORMAT_TYPE = PARQUET)
    GO
    
    --Create external table:
    CREATE EXTERNAL TABLE tbl_TaxiRides (
        vendorID VARCHAR(100) COLLATE Latin1_General_BIN2,
        tpepPickupDateTime DATETIME2,
        tpepDropoffDateTime DATETIME2,
        passengerCount INT,
        tripDistance FLOAT,
        puLocationId VARCHAR(8000),
        doLocationId VARCHAR(8000),
        startLon FLOAT,
        startLat FLOAT,
        endLon FLOAT,
        endLat FLOAT,
        rateCodeId SMALLINT,
        storeAndFwdFlag VARCHAR(8000),
        paymentType VARCHAR(8000),
        fareAmount FLOAT,
        extra FLOAT,
        mtaTax FLOAT,
        improvementSurcharge VARCHAR(8000),
        tipAmount FLOAT,
        tollsAmount FLOAT,
        totalAmount FLOAT
    )
    WITH (
        LOCATION = 'yellow/puYear=*/puMonth=*/*.parquet',
        DATA_SOURCE = NYCTaxiExternalDataSource,
        FILE_FORMAT = MyFileFormat\.\./\.\./\.\./azure-sql/
    );
    GO
    
    --Then, query the data via an external table with T-SQL:
    SELECT TOP 10 *
    FROM tbl_TaxiRides;
    GO
    

Next steps