Limitations and known issues for IBM Streams Runner
Edit meLearn about limitations and known issues for IBM® Streams Runner for Apache Beam.
Limitations
- Streams Runner does not support the Beam Python SDK.
- Experimental Splittable ParDos are not supported.
Known issues
Issue
The Streams job for a submitted Beam application does not show the Pipeline layout in the Streams Graph in the IBM Streams Console.
Workaround
Click on the ‘Configure’ icon (the wrench) in the Streams Graph for your application, check the ‘Show raw graph’ option, and click ‘Apply’ to see the Pipeline using all primitive transforms.
Additionally, use the latest version of IBM Streams (which contains the latest Streams Console) to run your Beam application. The latest version can be obtained in the Quick Start Edition.
Issue
The local Java™ installation is the outdated Java 8.0u31 (a release from 2015) and the Streams JRE version is 8.0u93 (from 2016).
This difference in versions causes the following error to be displayed when you run a Streams job in DISTRIBUTED
mode:
java.io.InvalidClassException: org.apache.beam.runners.core.SystemReduceFn$1;
local class incompatible: stream classdesc serialVersionUID = -7628843818184175876,
local class serialVersionUID = 2067056052611054708
Workaround
Choose one of the following options:
- Export
JAVA_HOME
to the internal IBM Streams Java Runtime Environment (JRE) found at$STREAM_INSTALL/java/
. - Update your Java installation.