Categories

Versions

Install a standalone RapidMiner Scoring Agent

This document describes how to install a standalone RapidMiner Scoring Agent that is independent of RapidMiner AI Hub and the Web API Endpoints. While it is possible to install a standalone RapidMiner Scoring Agent as a Web API Agent, it is recommended to use the provided installation methods using Docker/Kubernetes. SeeInstall the endpoint infrastructure.

Here we discuss an alternative method ofinstalling the endpoint infrastructure, not involving Docker or Kubernetes, that may be more suitable for edge computing or other cases where resources are limited.

The starting point is one of several ZIP files, depending on the standalone operating system, that can be found on

Instructions

  1. If it is not yet installed, install Java 11 (e.g., OpenJDK 11) on the standalone operating system.

  2. Unpack the RapidMiner Scoring Agent ZIP file and notice the following files and folders:

    • README.md- contains additional detailed instructions
    • bin/migrate-home- a shell script tomigrate from version 9 to version 10
    • bin/rapidminer-scoring-agent- a shell script to start the scoring agent
    • dist/linux-service/- (Linux only) a folder containing systemd service files
    • home/config/agent.properties- a file that configures scoring agent options
    • home/deployments/- a folder for your RapidMiner deployments, in ZIP format
    • home/resources/extensions/- a folder for extensions
    • home/resources/licenses/- a folder for licenses
  3. Install / configure alicense.

    The defaultlicense.modeis

    license.mode=ALTAIR_STANDALONE

    If you wish to use a RapidMiner license, set the following property inhome/config/agent.properties

    license.mode=RAPIDMINER

    Choose one of the following:

    • ALTAIR_STANDALONE- Inhome/config/agent.properties, uncomment and set the path for the .dat node lock license file:

      #license.altair.standalone.license-file=${license.location}/altair_standalone.dat
    • ALTAIR_UNIT- see the details inhome/config/agent.properties

    • RAPIDMINER(Scoring Agent license) - Frommy.www.turtlecreekpls.com, underLicenses > Scoring Agent, you can install the Scoring Agent license as:

      home/resources/licenses/rapidminer-scoring-agent/scoring-agent.lic
    • RAPIDMINER(RapidMiner Server license) - Frommy.www.turtlecreekpls.com, underLicenses > AI Hub / Server, you can install the RapidMiner Server license as:

      home/resources/licenses/rapidminer-server/server.lic

      In this case, you also need to uncomment the following line inhome/config/agent.properties:

      #scoring-agent.enable-server-license = true
  4. (optional) Install extensions.

    If you need them, simply download extensions from RapidMiner'smarketplaceand place them into

    home/resources/extensions

    They will be picked up and initialized after a successful restart of the application.

    You can also remove extensions, please be sure to restart the application in that case as well.

  5. Install the RapidMiner ZIP files you wish to deploy.

    Follow the instructions inCreate a deployment ZIP. The result is a ZIP file that you will place inhome/deployments, for example a deployment called iris.zip:

    home/deployments/iris.zip

    You must NOT unpack this ZIP file in the部署folder, but if you do unpack it -- elsewhere! -- here is what you will see:

    repository repository/ScoreIrisData.rmp repository/IrisModel.rmmodel config.json README.md

    Notice that theconfig.jsonfile identifies thebasePath(in this case, "iris") and all the callable processes in the deployment (in this case "ScoreIrisData"). Both these values will appear in theendpoint URL.

  6. Run the shell script to start the scoring agent.

    Once you have configured the scoring agent and inserted your deployment ZIP, it's time to start the scoring agent!

    sh bin/rapidminer-scoring-agent
  7. Request results.

    In general, scoring works as described inRequest results, except that an external scoring agent does not belong to a Web API Group, and therefore the rules for assigning an endpoint URL are slightly different.

    Using our iris.zip deployment as an example, you can call the ScoreIrisData process as follows:

    curl - x \ - h”内容类型:应用cation/json" \ --data @iris.json \ http://localhost:8090/api/v1/services/iris/scoreirisdata

    with the following example fileiris.json:

    { "data":[ { "a1": 5.1, "a2": 3.5, "a3": 1.4, "a4": 0.2 } ] }

    Note that in JSON files,trailing commas are not permitted.