To use the ORC reader or writer operators in DataFlow, the MapReduce jar files must be available in the classpath of containers launched on a Hadoop cluster. The classpath for containers (processes) launched by YARN is specified by the yarn.application.classpath property. The value for this property varies for each Hadoop distribution.

To ensure that the required MapReduce jar files are included in the yarn.application.classpath property, follow the instructions provided in To set yarn.application.classpath.