This release includes numerous bug fixes, support for new distributions, and new capabilities, such as the option to disable initial type inference on schematized sources.
Install, Config & Admin:
Support for Ubuntu 16.04. See System Requirements.
Support for Cloudera 5.13. See Supported Deployment Scenarios for Cloudera.
NOTE: Support for CDH 5.10 has been deprecated. Please upgrade your Hadoop cluster. For more information, see End of Life and Deprecated Features.
|TD-27799||DATEDIF function does not work for inputs that are functions returning date values.|
|TD-27703||Spark job fails with scala.MatchError|
|TD-24121||When publishing multi-part files, different permissions are written to the parent directory when job was executed on Hadoop or Photon.|
|TD-27950||Transformer Page - Tools|
When you join with an imported dataset not in your flow and it takes longer than expected to collect its initial sample, you may encounter the following error:
Ubuntu 16 install for Azure: supervisord complains about "missing" Python packages.
This release introduces deployment management, which enables separation of development and production flows and their related jobs. Develop your flows in a Dev environment and, when ready, push to Prod, where they can be versioned and triggered for production execution. Additionally, you can create and manage all of your connections through the new Connections page. A revamped flow view streamlines object interactions and now supports starting and stopping of jobs without leaving flow view.
Details are below.
New objects in Flow View and better organization of them. See Flow View Page.
NOTE: Wrangled datasets are no longer objects in the . Their functionality has been moved to other and new objects. For more information, see Changes to the Object Model.
See Object Overview.
Upload your own avatar image. See User Profile Page.
NOTE: This feature may need to be enabled. See Miscellaneous Configuration.
For technical reasons, the name and default value of the following parameter has been changed in Release 4.2.
|Affected Releases||Parameter Name||Default Value||Max Value|
|Release 4.2 and later|
|Release 4.1.1 and earlier|
NOTE: Upgrading customers have the new configuration setting automatically set to the default:
For more information on changing this parameter value, see Configure Application Limits.
In Release 4.0, the
/docs endpoint was deprecated from use. This endpoint displayed a documentation page containing information on language, the command line interface, and .
In Release 4.2, this endpoint has been removed from the platform. Content has been superseded by the following content:
For more information on features that have been deprecated or removed, see End of Life and Deprecated Features.
If you are integrating with S3 sources, the platform now requires use of the s3a protocol. The s3n protocol is no longer supported.
No configuration changes in the are needed. See Enable S3 Access.
|TD-27748||Direct publish to Hive fails on wide datasets due to Avro limitations.|
SQL Server Database timing out with long load times.
|TD-27197||Column histogram does not update after adding |
|TD-27127||Send a Copy tab in Flow View sharing does not include all available users.|
|TD-27055||Job run on flow with complex recipes fails on Hadoop but succeeds on Photon.|
|TD-26837||Creating custom dictionaries fails on S3 backend datastore.|
|TD-26388||Orphaned bzip2 processes owned by the platform user accumulate on the node.|
|TD-26041||When editing a schedule that was set for 0 minutes after the hour, the schedule is displayed to execute at 15 minutes after the hour.|
|TD-25903||Overflow error when ROUND function is applied to large values.|
|TD-25733||Attempting a union of 12 datasets crashes UI.|
|TD-25709||Spark jobs fail if HDFS path includes commas.|
DATEDIF function does not work for inputs that are functions returning date values.
|TD-27703||Compilation/Execution||Spark job fails with scala.MatchError|
|TD-26069||Compilation/Execution||Photon evaluates |
|TD-24121||Compilation/Execution||When publishing multi-part files, different permissions are written to the parent directory when job was executed on Hadoop or Photon.|