Endpoint Unhealthy error when trying to test deployed endpoint for inference pipeline

Dhruv Bala 16

Hello, I am following the module from Microsoft Learn on regression using Azure's Machine Learning Designer, when trying to Test the service

When trying to test the predict-auto-price endpoint, I get the following screen

Does anyone know how I can address this? I am fairly new to Azure, so I apologize if I'm missing something obvious

Ramr-msft 17,616 Reputation points

2022-11-29T07:46:52.23+00:00

@Dhruv Bala Thanks for the question. Can you please add more details about Pipeline Endpoint Status?
Dave Chennault 41 Reputation points

2022-12-07T23:29:12.62+00:00

Bottom line - wait a bit and keep checking to see if it deploys

I had the exact same issue in this lab

I am trying to deploy to an Azure Compute instance

The only option available when I first tried to deploy was a Kubernetes compute target option so I manually Typed in the Azure Compute Instance in that dropdown and then it was able to deploy - You should update the docs to reflect you need to manually type in the Azure Compute Instance into that dropdown - I just guessed how to do it and it worked

The endpoint was unhealthy for around 15 minutes with a null end point then the state changed to transitioning but endpoint was still null

After a few more minutes it deployed and now has a healthy status and a REST endpoint

Probably took around 30 minutes (or more) to deploy so you may want to note that in the documentation

BTW - I am working on hooking this endpoint up to a Flow and Power App and storing the inputs and predictions into a list in SharePoint
The Power App is an embedded form in that SharePoint list - :)

:)

2 answers