Good day,
I am getting the error message: "Model deployment failed on connector TW.AnalysisServices.ThingPredictorConnector" as a pop-up message and in the Application monitor: "No client thing found to deploy model".
From support case CS269545 it seems that the issue is the deployment of Thingpredictor. I have gone through the deployment guide (https://support.ptc.com/help/thingworx_hc/thingworx_analytics_8/index.html#page/thingworx_analytics_8/thingpredictor/thingpredictor-deployment.html) and I am slightly confused because I was under the impression that all the required microservices except Thingwatcher are deployed by default. I checked the predictive-standalone.yml file and the steps to change in the Thingpredictor deployment guide but the defaults look to be proficient. Is there a setting I should change that I have missed?
From topic https://community.ptc.com/t5/ThingWorx-Developers/Model-deployement-failed-on-connector-TW-AnalysisServices/m-p/543898 which seems like a similar issue, it was the default ports (8080 and 8081) that was the issue. I have tested different ports and it does not seem to be my issue.
Any advice?
Thank you.
Johan Bester
Solved! Go to Solution.
Hi
To clarify maybe some confusion you had, ThingPredictor is not deployed by default.
You do need to download the ThingPredictor media and update the system-environment-variables.properties file (do not touch the predictive-standalone.yml file) with relevant data for your site (be aware of https://www.ptc.com/en/support/article?n=CS281505 if using Windows)
Then start ThingPredictor with a command similar to the below one , executed from a terminal in the directory where thingpredictor.jar is located
java "-Dvariables.file=file:///C:/ThingPredictor/system-environment-variable.properties" -jar thingpredictor.jar server predictive-standalone.yml
If you still have issue, it will be helpful that you upload:
- predicitve-standalone.yml
-system-environment-variables.properties
- thingpredictor.log file
- ThingWorx Application.log
- the OS you run ThingPredictor and ThignWorx on
- which deployment did you use: Trial edition, docker or native deployment
Hope this helps
Kind regards
Christophe
Hi
To clarify maybe some confusion you had, ThingPredictor is not deployed by default.
You do need to download the ThingPredictor media and update the system-environment-variables.properties file (do not touch the predictive-standalone.yml file) with relevant data for your site (be aware of https://www.ptc.com/en/support/article?n=CS281505 if using Windows)
Then start ThingPredictor with a command similar to the below one , executed from a terminal in the directory where thingpredictor.jar is located
java "-Dvariables.file=file:///C:/ThingPredictor/system-environment-variable.properties" -jar thingpredictor.jar server predictive-standalone.yml
If you still have issue, it will be helpful that you upload:
- predicitve-standalone.yml
-system-environment-variables.properties
- thingpredictor.log file
- ThingWorx Application.log
- the OS you run ThingPredictor and ThignWorx on
- which deployment did you use: Trial edition, docker or native deployment
Hope this helps
Kind regards
Christophe
Hi @cmorfin,
Thank you for the reply, I have got it working.
Just out of curiosity, when I look at the connected remote things there is a PredictionThing connected by default as in the screenshot:
And in the default standalone_jars file on the analytics server, you already have the predictive-standalone.yml file as well as a jar file called predictive-ms.jar.
What is the role of the default "PredictionThing" that is running and the "predictive-ms.jar" file?
Kind regards,
Johan
Hi
The PredictionThing that you see is one of the service things created by ThingWorx Analytics Server.
ThingPredictor is an independent standalone scoring service.
It actually does use a similar engine but is packaged in a standalone way to allow access to the scoring functionality without the need of a full ThingWorx Analytics Server.
Note that this is true as of release 8.2. It is likely to change in future release where we probably will have only one scoring agent.
Hope this answer your question
Kind regards
Christophe