Error Invalid Model

[darcudarcu]

Hi, 

It is my first experience with TW Analytics, I am trying to upload a PMML model version 4.3 from third part.

I created a Provider using the 'AnalyticsServerConnector' but when I upload the model from file using 'Analysis Models' I get the error: 'Invalid Model' and on the ApplicationLog says: 'Unable to deploy model to TW.AnalysisServices.AnalyticsServer.AnalyticsServerConnector'. 

Do you know if I might solve that problem or if it is for sure a PMML model problem of compatibility?

Kiund regards,

Marco

[posipova]

@nsampat @jgreiner  would either of you guys be able to help here?

[nsampat]

@darcudarcu ,

My colleague @posipova brought this thread to my attention.

Can you provide the following information:

What version of ThingWorx Platform Server is being used?

What version of ThingWorx Analytics Server is being used?

Also please note, while PMML 4.3 should work with our application (8.5.1), it needs to be formatted using JPMML 1.3.3.

Have you attempted to upload a dataset into ThingWorx Analytics, then publish to Model using the Connector?

Regards,

Neel

[darcudarcu]

Hi @nsampat ,

My platform and analytics version is 8.3.1-b441, tomorrow I'll try to see if I get any error publishing a model from data. 

Thanks and kind regards,

Marco

[darcudarcu]

Hi @nsampat ,

I tried to follow the steps of the TW tutorial to create a model with analytics from an example dataset provided in the tutorial itself. I crated a dataset, a signal and a profile but I get an error when I try to create model.

ApplicationLog:
[context: com.thingworx.webservices.context.HttpExecutionContext@6ec38dba][message: Execution error in service script [CreateModelAMS] :: java.net.ConnectException: Connection refused: connect]

ScriptLog:
GetTrainingJobsFirstPageAMS: Error: JavaException: com.thingworx.common.exceptions.GenericHTTPException: java.net.ConnectException: Connection refused: connect
[message: Execution error in service script [CreateModelAMS] :: java.net.ConnectException: Connection refused: connect

Kind regards,

Marco

[nsampat]

Marco,

Can you take a look or provide at the all the logs found in /ThingWorxAnalyticsServer/data/logs ?

Does ThingWorx Platform Server Application Log or Security log provide any error messages, or just the same one you are seeing in Analytics Builder ?

Also, please double check if ThingWorx Analytics training microservice is running. Or you can restart ThingWorx Analytics and see if it connects, and all services com back online. Article on how to restart ThingWorx Analytics can be found here: https://www.ptc.com/en/support/article?n=CS265543

Can you confirm, you were able to upload a dataset, and create singles and profiles, but not a model?

Was there a power outage or any sort of violent shutdown of the ThingWorx Analytics Server?

Regards,

Neel

[darcudarcu]

Hi @nsampat, 

the applicationLog and SecurityLog do not provide any further info.

apparently ThingWorx cannot communicate with Analytics server on port 80, I already turned off the firewall etc.

The thing on composer "AnalyticsServer" is connected (property) and services that work on other ports Dataset, Signals etc. work fine.

I noticed that the file "twas-edge-ms-report.json" has the port number value empty, moreover the port 80 is  being used by java.

All the thingworx analytics services are running.

I attach hereby the logs of yesterday and today.

Kind regards,

Marco

[nsampat]

Marco,

Did this setup work previously? Port 80 is not the default port, as Analytics is looking for 8080. twas-edge-ms-report.json is just an after install report and not part of any configurations, and if the edge microservice could not connect to Platform Server, signals and profiles would have not run.

I am seeing in the training.log file the following error:

ERROR [2019-10-25 02:30:17,949] org.apache.tomcat.jdbc.pool.ConnectionPool: Unable to create initial connections of pool.
! java.io.EOFException: null
! at org.h2.mvstore.DataUtils.readFully(DataUtils.java:431)
! ... 47 common frames omitted
! Causing: java.lang.IllegalStateException: Reading from cache:nio:C:/Program Files (x86)/ThingWorxAnalytics/Thingworx_Analytics_Server/data/db/training-db.mv.db failed; file length 446464 read length 1024 at 447674 [1.4.192/1]

Typically we see this issue if there was a violent or improper shutdown of the server, causing corruption to the db file for Training (cannot be read or written to).

Please take a look at this Article on how to recover/resolve this issue with the corruption (you may lose any models you previously trained).

Regards,

Neel

[darcudarcu]

Hi @nsampat ,

After  recovering this issue with the corruption I am able to create and publish  a model from a dataset. However the model from the partner is not valid.

I assume is a problem of the model in this case, right?

Kind regards 

[nsampat · Accepted solution]

@darcudarcu ,

That is correct, I would recommend sending them the PMML of the model you create as a example of the format the partner needs to follow. Their model may be valid, but in a different format than Analytics is expecting.

Regards,

Neel

[darcudarcu]

Thank you very much!

[nsampat]

My pleasure, and thank you for marking my response as Solution Accepted!

Have a great week!

[darcudarcu]

Hi @nsampat 

I also get reported issues at the port 445 (using wireshark) when I try to create the model.

Kind regards,

marco