What's New in RapidMiner Radoop 9.1.0?

Released: December 13th, 2018

The following improvements are part of RapidMiner Radoop 9.1.0.

Enhancement

  • User can select Hortonworks Data Platform(HDP) 3.x as a Hadoop Distribution to connect to.
  • User can select Cloudera(CDH) 6 as a Hadoop Distribution to connect to, but is limited in features Radoop is compatible with.
  • Extended support to Spark Scripting operator for various Spark 2.3.x versions including Apache 2.3.0, Apache 2.3.1+ and version included in HDP3 default installation.
  • Allowing user to import from a Cluster Manager that provides an untrusted certificate.
  • Introduced additional validation of Advanced Parameter settings for new connections.
  • Improved robustness of setting up client classpath for MapR connections.
  • Additional to the existing built-in defaults users can now specify their PySpark and SparkR libraries in their Radoop Connections to be used with the Spark Scripting operator.

Bug Fix

  • Duplicate property keys received by Cluster Import won't stop wizard. Deduplication uses Hadoop equivalent priorities between the sources of the duplicated properties.
  • MapR connection - Fixed a malformed logging statement during classpath setup.

Additional Note

  • Radoop Proxy feature is not yet compatible with HDP3
  • Rounding errors may be observed after 6th digit with 'integer' arithmetic's which are limited to DECIMAL(9,6) precision in HiveServer2 3.0. Recommendation is to use 'real' instead for calculations which require higher precision.
  • Apply Model does not accept DECIMAL columns in input (same for previous versions, but HiveServer2 3.0 introduces lot of implicit DECIMAL casts), which results in UDFArgumentTypeExceptions. The workaround is to cast 'integer' values to 'real' with Type Conversion operator.
  • Password based Kerberos authentication is no longer supported because of incompatibilities with key renewals and HDP3 configuration. Recommendation is to use keytab based Kerberos authentication instead.