Page tree

Release 8.2.2


Contents:

   

Contents:


You can create a connection to Databricks Tables from the Designer Cloud Powered by Trifacta platform. This section describes how to create connections of this type.

  • Databricks Tables provides a JDBC-based interface for reading and writing datasets in ADLS or WASB. Using the underlying JDBC connection, you can access your ADLS or WASB data like a relational datastore, run jobs against it, and write results back to the datastore as JDBC tables. 
  • Your connection to Databricks Tables leverages the SSO authentication that is native to Databricks. 

Limitations

  • Ad-hoc publishing of generated results to Databricks Tables is not supported.
  • Integration with Kerberos or secure impersonation is not supported.
  • Some table types and publishing actions are not supported. For more information, see Using Databricks Tables.

Pre-requisites

  • Azure: The Designer Cloud Powered by Trifacta platform must be installed on Azure and integrated with an Azure Databricks cluster.
  • AWS: The Designer Cloud Powered by Trifacta platform must be installed on AWS and integrated with an AWS Databricks cluster.

  • This connection interacts with Databricks Tables through the Hive metastore that has been installed in the Databricks cluster.

    NOTE: External Hive metastores are not supported.

Insert Databricks Access Token

Each user must insert a Databricks Personal Access Token into the user profile. For more information, see Databricks Settings Page.

Enable

To enable Databricks Tables connections, please complete the following:

NOTE: Typically, you need only one connection to Databricks Tables, although you can create multiple connections.

NOTE: This connection is created with SSL automatically enabled.

Tip: Your SSL certificate should be created using 2047-bit or larger keys. In some environments, such as RHEL 8.x, 1024-bit encryption is no longer accepted per default cryptographic policies.

Steps:

  1. You can apply this change through the Admin Settings Page (recommended) or trifacta-conf.json. For more information, see Platform Configuration Methods.
  2. Locate the following parameter and set it to true:

    "feature.databricks.connection.enabled": true,
  3. To allow for direct publishing of job results to Databricks tables from the Run Job page, you must enable the following parameters. For more information on these settings, see Databricks Tables Table Settings.

    ParameterDescription
    feature.databricks.enableDeltaTableWritesSet this value to true to enable users to choose to write generated results to Databricks delta tables from the Run Job page.
    feature.databricks.enableExternalTableWritesSet this value to true to enable users to choose to write generated results to Databricks external tables from the Run Job page.
  4. Save your changes and restart the platform.



Create Connection

This connection can also be created via API. For details on values to use when creating via API, see Connection Types.

Please create a Databricks connection and then specify the following properties with the listed values:

NOTE: Host and port number connection information is taken from Databricks and does not need to be re-entered here.


PropertyDescription
Connect String options

Please insert any connection string options that you need. Connect String options are not required for this connection.

Test Connection
Click this button to test the specified connection.
Default Column Data Type Inference

Set to disabled to prevent the Designer Cloud Powered by Trifacta platform from applying its own type inference to each column on import. The default value is enabled.

Reference

Connection URL

The properties that you provide are inserted into the following URL, which connects  Designer Cloud Enterprise Edition to the connection:

jdbc:spark://<host>:<port>/<database><connect-string-options>

The Connection URL is mostly built up automatically using cluster configuration for the platform.

Connect string options

The connect string options are optional. If you are passing additional properties and values to complete the connection, the connect string options must be structured in the following manner:

;<prop1>=<val1>;<prop2>=<val2>...

where:

  • <prop> : the name of the property
  • <val> : the value for the property

delimiters:

  • ; : any set of connect string options must begin and end with a semi-colon.
    • A semi-colon can be omitted from the end of the connect string options.
  • = : property names and values must be separated with an equal sign (=).

Use HTTP

To enable the use of the HTTP protocol, specify the following in the connect string options:

;transportMode=http;

Use SSL

To enable the use of SSL for the connection, specify the following in the connect string options:

;ssl=1;

HTTP Path

When HTTP is enabled, you can specify the path as a connect string option:

;httpPath=sql/protocolv1/o/0/xxxx-xxxxxx-xxxxxxxx;

Authentication

You can specify a Databricks personal access token to use when authenticating to the database using the following connect string options. 

;AuthMech=3;UID=token;PWD=<Databricks-personal-access-token>

where:

  • <Databricks-personal-access-token> = the personal access token of the user who is connecting to the database.

Driver Information

This connection uses the following driver:

Use

For more information, see Using Databricks Tables.

Data Conversion

For more information on how values are converted during input and output with this database, see Databricks Tables Data Type Conversions.

Troubleshooting

For more information on error messages for this connection type, see https://kb.databricks.com/bi/jdbc-odbc-troubleshooting.html.

Failure when importing wide Databricks Tables table

If you are attempting to import a table containing a large number of columns (>200), you may encounter an error message similar to the following:

org.apache.spark.SparkException: Job aborted due to stage failure: Task 0 in stage 408.0 failed 4 times, most recent failure: Lost task 0.3 in stage 408.0 (TID 1342, 10.139.64.11, executor 11): org.apache.spark.SparkException: Kryo serialization failed: Buffer overflow. Available: 0, required: 1426050. To avoid this, increase spark.kryoserializer.buffer.max value.

The problem is that the serializer ran out of memory.

Solution:

To address this issue, you can increase the Kyroserializer buffer size. 

  1. You can apply this change through the Admin Settings Page (recommended) or trifacta-conf.json. For more information, see Platform Configuration Methods.
  2. Locate the spark.props section and add the following setting. Modify 2000 (2GB) depending on whether your import is successful:

    "spark.kryoserializer.buffer.max.mb": "2000"
  3. Save your changes and restart the platform. 
  4. Attempt to import the dataset again. If it fails, you can try incrementally raising the above value.

For more information on passing property values into Spark, see Configure for Spark.

This page has no comments.