Skip to content

SetupAzureDevops.md

Latest commit

 

History

History
35 lines (28 loc) · 2.69 KB

SetupAzureDevops.md

File metadata and controls

35 lines (28 loc) · 2.69 KB

Setup Devops Project

  1. Login to Azure Devops -> Enable preview feature called Multi Stage Pipeline. Instructions here.

  2. Create a project from the devops portal (top right of the portal). If you have trouble then refer to docs

  3. Create Azure Resource Manager Service connection. This is needed for azure devops to connect to your subscription and create/manage resources.

    Go to project settings in bottom left of devops portal & select Service Connections and setup a Resource Manager connection. You have few options:

    • Option 1: If you have Contributor or Owner access to the Subscription or a Resource Group
      • Select Service Principal (Automatic)
      • Select the scope of your choice (ideally select Subscription as scope and specific Resource group)
      • Name of this Connection should be AzureResourceManagerConnection. Leave this checked Allow all pipelines to use this connection.
    • Option 2: If you would like to restrict access to a ML workspace level OR if you are in a tighter access control environment:
      • Create Service Principal & look up Subscription id. Instructions here. If you already created one part of prerequisites, you can skip this.
      • Select Service Principal (Manual)
      • Select the scope of your choice (in this case select Machine Learning Workspace)
      • Fill in all the values
      • Name of this Connection should be AzureResourceManagerConnection. Leave this checked Allow all pipelines to use this connection.

  4. The following step is needed for additional security for the prediction service that we will deploy. Inorder to treat the service endpoint URI and API key as secret in the devops pipeline, create a variable group:

    1. In Azure Devops leftnav, navigate to Pipeline -> Library. Create a new Variable group by clicking + Variable. Name it MLOPSVG
    2. Open the group and select Allow access to all pipelines
    3. Add two new variables TMP_API_KEY and TMP_SCORING_URI. For the values enter any value e.g. dummy. Click the Lock icon in the value to mark it Secret.
      Save the changes to the Variable group

  5. [Optional] For Batch Inference you need to add the following variables using the values you got in the prerequsite step:

    1. WORKSPACE
    2. RESOURCE_GROUP
    3. SP_APP_ID
    4. SP_APP_SECRET
    5. SUBSCRIPTION_ID
    6. TENANT_ID
      Save the changes to the Variable group