Job Containers

Job Containers are the executing entities, they communicate with RapidMiner Server and connect to various persistent storages and endpoints if needed by the process itself. Job Containers are ephemeral, they are spawned by Job Agents and either stop normally at the end of the execution or will be stopped by the spawning agents. Before the execution the agent spawning the container creates a context directory and places necessary data into it. The path to this directory is then passed along to the container so it can utilize the information needed for execution.

Context directory

The directory is created under Job Agent's {baseDir}/home/data/jobs/{jobId}/ using the job's ID as the name, where baseDir is the root directory of the unpackaged agent distribution. The agent puts everything from {baseDir}/lib/container as is into the context directory (See Installation > Job Agents), so before starting the container the directory contains the following:

  • process.xml: the RapidMiner process itself
  • contains a JWT token to authenticate with RapidMiner Server endpoint
  • properties which are going to be added as system properties to the execution
  • jdbc_properties.xml: JDBC drivers and their location _ .RapidMiner: context directory used by RapidMiner Core

The container itself writes its output to the {baseDir}/home/data/jobs/{jobId}/ directory:

  • execution.log: the log of the container
  • error.log: log file containing the error message, created only in case of error
  • hs_err.log: JVM error log, created only in case JVM crashes