My colleagues and I are working on a system that should interoperate with Apache Spark. The idea behind this interoperability is to send Spark Job Server a JAR package containing classes that, whenever invoked from the system, should deal with Spark. The issue is, it sometimes happens that a new update to this JAR package doesn’t update it at all! You probably are already seeing where this is leading: the JAR package is somewhat retained by Spark Job Server, which then uses the old classes. It is necessary to erase the old files, then. In a Unix-like system, this can be accomplished by stopping Spark Job Server, then firing the following command:

  rm -rf /tmp/spark-jobserver/*
  

Please note that your instance of Spark Job Server may save temporary files in a different location. Moreover we use a rather old version of Spark Job Server: newer version may require different commands (or not require them at all!)