Possible problems

Possible cause: Missing device name

The metrics collected by AI Model Monitor Agent are labeled with the name of the device (device_name), where the collection occurs. This label is used to identify the source of the metrics.

On startup, the Agent application tries to acquire the device's system information. After several unsuccessful attempts, the application will be terminated.

Please download the application logs from to device to get additional information on the cause of the problem.

Missing pipeline metric

If you do not see pipeline metrics appearing, there might be a problem with pipeline configurations in AI Inference Server. Make sure that the following points are fulfilled:

Pipeline is running in the AI Inference Server (Overview → Status should be "Running")

Pipeline variables are correctly configured (Pipeline visualization → Variables)

Data connection, Username and Password are correctly configured. (Data connection → edit button (Pencil icon))

If none of the solutions work, please take a look in the AI Inference Server manual.

Missig monitoring extension or custom metric

Monitoring nodes and custom metrics are collected from the Databus. If you find that these metrics are missing from AI Model Monitor or its Agent, make sure that the Databus credentials set in AI Inference Server, AI Model Monitor Agent applications and Data connections on IEM match.

If the above credentials are correct and the expected metrics are still missing, try restarting the applications.

Whenever the Databus user permissions used by AI Model Monitor Agent are changed on IEM, restart the Agent on the device so that the applied modifications can take effect.

AI Model Monitor Agent forwards the collected metric data to the AI Model Monitor application using a TLS-enabled, secure connection. To ensure the security of the data, AI Model Monitor Agent uses a certificate file provided by the Edge platform. In case metric data from the Agent do not appear in the AI Model Monitor application after a few minutes, the following check points are suggested for troubleshooting:

• Make sure that the AI Model Monitor's address was correctly configured at Agent app install in the configuration form.

• If you uploaded a custom certificate file, make sure it is the correct public pair of the key configured for the Edge device hosting the AI Model Monitor application.

• If the logs indicate an issue with the default device root certificate, try uploading the Device root certificate by hand with an app configuration update. The Device root certificate can be downloaded from IEM.