Access Spark Data as a Remote Oracle Database

The Oracle Database Gateway for ODBC and Heterogeneous Services technology enable you to connect to ODBC data sources as remote Oracle databases. This article shows how to use the CData ODBC Driver for Spark to create a database link from Spark to Oracle and to query Spark data through the SQL*Plus tool. You can also create the database link and execute queries from SQL Developer.

Connect to Spark as an ODBC Data Source

Information for connecting to Spark follows, along with different instructions for configuring a DSN in Windows and Linux environments.

Set the Server, Database, User, and Password connection properties to connect to SparkSQL.

Windows

If you have not already, first specify connection properties in an ODBC DSN (data source name). This is the last step of the driver installation. You can use the Microsoft ODBC Data Source Administrator to create and configure ODBC DSNs.

Note: If you need to modify the DSN or create other Spark DSNs, you must use a system DSN and the bitness of the DSN must match your Oracle system. You can access and create 32-bit DSNs on a 64-bit system by opening the 32-bit ODBC Data Source Administrator from C:\Windows\SysWOW64\odbcad32.exe.

Linux

If you are installing the CData ODBC Driver for Spark in a Linux environment, the driver installation predefines a system DSN. You can modify the DSN by editing the system data sources file (/etc/odbc.ini) and defining the required connection properties.

/etc/odbc.ini

For specific information on using these configuration files, please refer to the help documentation (installed and found online).

Set Connection Properties for Compatibility with Oracle

The driver provides several connection properties that streamline accessing Spark data just as you would an Oracle database. Set the following properties when working with Spark data in SQL*Plus and SQL Developer. For compatibility with Oracle, you will need to set the following connection properties, in addition to authentication and other required connection properties.

• MapToWVarchar=False

  Set this property to map string data types to SQL_VARCHAR instead of SQL_WVARCHAR. By default, the driver uses SQL_WVARCHAR to accommodate various international character sets. You can use this property to avoid the ORA-28528 Heterogeneous Services data type conversion error when the Unicode type is returned.

• MaximumColumnSize=4000

  Set this property to restrict the maximum column size to 4000 characters.

• IncludeDualTable=True

  Set this property to mock the Oracle DUAL table. SQL Developer uses this table to test the connection.

Linux Configuration

In Linux environments, Oracle uses UTF-8 to communicate with the unixODBC Driver manager, whereas the default driver encoding is UTF-16. To resolve this, open the file /opt/cdata/cdata-driver-for-sparksql/lib/cdata.odbc.sparksql.ini in a text editor and set the encoding.

cdata.odbc.sparksql.ini

Configure the ODBC Gateway, Oracle Net, and Oracle Database

Follow the procedure below to set up an ODBC gateway to Spark data that enables you to query live Spark data as an Oracle database.

1. Create the file initmysparkdb.ora in the folder oracle-home-directory/hs/admin and add the following setting:

   initmysparkdb.ora

   HS_FDS_CONNECT_INFO = "CData SparkSQL Sys"
2. Add an entry to the listener.ora file. This file is located in oracle-home-directory/NETWORK/admin.

   If you are using the Database Gateway for ODBC, your listener.ora needs to have a SID_LIST_LISTENER entry that resembles the following:

   listener.ora

   SID_LIST_LISTENER = (SID_LIST = (SID_DESC = (SID_NAME = mysparkdb) (ORACLE_HOME = your-oracle-home) (PROGRAM = dg4odbc) ) )

   If you are using Heterogeneous Services, your listener.ora needs to have a SID_LIST_LISTENER entry that resembles the following:

   listener.ora

   SID_LIST_LISTENER = (SID_LIST = (SID_DESC = (SID_NAME = mysparkdb) (ORACLE_HOME = your-oracle-home) (PROGRAM = hsodbc) ) )

3. Add the connect descriptor below in tnsnames.ora, located in oracle-home-directory/NETWORK/admin:

   tnsnames.ora

   mysparkdb = (DESCRIPTION= (ADDRESS=(PROTOCOL=tcp)(HOST=localhost)(PORT=1521)) (CONNECT_DATA=(SID=mysparkdb)) (HS=OK) )
4. Restart the listener.

5. Test the configuration with the following command:

   tnsping mysparkdb

6. Open SQL*Plus and create the database link with the command below:

   CREATE DATABASE LINK mysparkdb CONNECT TO "user" IDENTIFIED BY "password" USING 'mysparkdb';

You can now execute queries in SQL*Plus like the one below (note the double quotation marks around the table name):