What's New in RapidMiner Radoop 8.1.0?
Released: February 06th, 2018
The following improvements are part of RapidMiner Radoop 8.1.0.
- Added support for MapR 5.x
- Local temporary files are now regularly accessed, therefore, cleanup tools like tmpwatch should no longer delete them (leading to NoClassDefFoundError errors) while the software is still open
- Avro file format is now available everywhere
- Decreased default value from 70 to 30 for Resource Allocation % for Spark in case of static settings
- Pig test is now disabled by default, as it is only required by Pig Script operator (can be re-enabled via Customize option of Full Test)
- Added a warning when using the _HOST keyword in Hive Principal which is not expected to work
- BUGFIX: Fixed user impersonation and access whitelist features when using RapidMiner Server 8.0 or later (when running inside Job Containers)
- BUGFIX: Spark Script using HiveContext no longer fails if user impersonation is used
- BUGFIX: Kerberized Hive High Availability connection no longer requires hard-coded HiveServer2 address in Hive Principal field: both connection import methods automatically add __hiveserver_hostname__ template that is dynamically replaced by chosen HiveServer2 address
- BUGFIX: Fixed potential cluster-side error for jobs involving Single Process Pushdown and SparkRM operators (NoClassDefFoundError for RDDOperationScope)
- BUGFIX: Fixed potential error in Single Process Pushdown and SparkRM operators with Server 8.0, if auxiliary jars for the user are uploded from Server first and not Studio
- BUGFIX: Hive: Failed to access metastore (harmless) error message is no longer displayed in console log for Spark 1.5 or earlier versions