CODEWITHSUNDEEP
azureapiazure-machine-learning-service
fckbo
fckbo

Reputation: 167

Cannot get Azure ML Model Management API to work in GermanyWestCentral as in EastUs for example

Considering the Azure ML API and how to manage models with it, I've got the following problem.

Here is the sequence that works:

  1. Obtain a token(OK) - > POST https://login.microsoftonline.com/{{TenantId}}/oauth2/token
  2. List Workspaces (OK) -> GET https://management.azure.com/subscriptions/{{SubscriptionId}}/providers/Microsoft.MachineLearningServices/workspaces?api-version=2018-03-01-preview I've got several workspaces, some created in EastUS and some in GermanWestCentral regions
  3. In the json returned from previous API call, there is an attribute returned for each workspace called discoveryUrl which is either 'https://germanywestcentral.api.azureml.ms/discovery' or 'https://eastus.api.azureml.ms/discovery'
  4. Invoking GET https://germanywestcentral.api.azureml.ms/discovery returns
{
    "api": "https://germanywestcentral.api.azureml.ms",
    "catalog": "https://catalog.cortanaanalytics.com",
    "experimentation": "https://germanywestcentral.api.azureml.ms",
    "gallery": "https://gallery.cortanaintelligence.com/project",
    "history": "https://germanywestcentral.api.azureml.ms",
    "hyperdrive": "https://germanywestcentral.api.azureml.ms",
    "labeling": "https://germanywestcentral.api.azureml.ms",
    "modelmanagement": "https://germanywestcentral.api.azureml.ms",
    "pipelines": "https://germanywestcentral.aether.ms",
    "studio": "https://ml.azure.com"
}
  1. Invoking GET https://eastus.api.azureml.ms/discovery returns
{
    "api": "https://eastus.api.azureml.ms",
    "catalog": "https://catalog.cortanaanalytics.com",
    "experimentation": "https://eastus.experiments.azureml.net",
    "gallery": "https://gallery.cortanaintelligence.com/project",
    "history": "https://eastus.experiments.azureml.net",
    "hyperdrive": "https://eastus.experiments.azureml.net",
    "labeling": "https://eastus.experiments.azureml.net",
    "modelmanagement": "https://eastus.modelmanagement.azureml.net",
    "pipelines": "https://eastus.aether.ms",
    "studio": "https://ml.azure.com"
}
  1. The modelmanagement url do not have the same structure in both regions

(well fine, that should not be an issue)

  1. Now invoking GET https://eastus.modelmanagement.azureml.net/api/subscriptions/{{SubscriptionId}}/resourceGroups/{{resourceGroupName}}/providers/Microsoft.MachineLearningServices/workspaces/{{workspaceName}}/services?api-version=2018-03-01-preview&count=100 do return data describing the services available under the workspace

But the problem is that invoking:

GET https://germanywestcentral.api.azureml.ms/api/subscriptions/{{SubscriptionId}}/resourceGroups/{{resourceGroupName}}/providers/Microsoft.MachineLearningServices/workspaces/{{workspaceName}}/services?api-version=2018-03-01-preview&count=100 returns a 530 ERROR with "unknown to the cluster" in the body of the response

Any idea or hints on why this and how to get around this issue ?

Upvotes: 0

Views: 200

Answers (1)

Yutong Tie
Yutong Tie

Reputation: 498

Error 530 is kind of authentication error as far as I know. Please check the input var your used for below bugged URL

https://germanywestcentral.api.azureml.ms/api/subscriptions/{{SubscriptionId}}/resourceGroups/{{resourceGroupName}}/providers/Microsoft.MachineLearningServices/workspaces/{{workspaceName}}/services?api-version=2018-03-01-preview&count=100

Upvotes: 0

Related Questions