Azure ML "Real-time Endpoint Deploy" not working

user_8937663 0
Viewed 92

I am trying to deploy as Kubernetes service (AKS) on Azure ML studio. I get a notification stating "Preparing to deploy", but Nothing shows up after that. I checked the endpoints list after a while and it is not present there either.

After that "Preparing to deploy" notification I don't receive any success or failure notification.

If anyone has any solution to this issue, the help is much appreciated.

I have attached screenshots of the notification and ml pipeline I tried to deploy for context.

Facing the issues when using container instances as well...

The notification right after saving endpoint details and clicking create:

image1

The pipeline I tried to deploy:

image2

UPDATE: I think the problem was with the pipeline... I removed the web services, ran the pipeline, then created an inference pipeline from that, and then I was able to deploy using AKS. Although would be good to know the reason behind this... and why this workaround made all the difference

azure-machine-learning-studio azure-machine-learning-service azure
1 Answers

There are some limitations applied to attaching an AKS to the machine learning studio.

  1. AKS requires a public IP for the egress traffic. If your AZURE policy restricts the creation of public IP, the creation of the AKS cluster will fail.

  2. To attach an AKS cluster, whoever performing the operation must be assigned an Owner or contributor access control (Azure RBAC) role on the Azure resource group for the cluster

  3. The name of the cluster should be unique within your workspace

much more are there, you can refer to Microsoft AZURE documentations here

I would recommend you create the cluster first and then attach it to the ML workspace.

user_50393560
Related