jayyanar / blr-azure-mlops-demo

[[TOC]]

Task 1 - Create Azure Subscription

Login to https://azure.microsoft.com/en-in/free/

Click --> Start free

Create New Account or Signup using your Github

---------- I signed up using GitHub ------------------------------

================ END OF TASK 1 =======================================

Task 2 - Create DevOps Subscription

Azure Portal

Once you signed up - You will be directed to https://portal.azure.com/#home

Create Resource Group

Go to --> https://portal.azure.com/#create/Microsoft.ResourceGroup

Create Resource Group

Create DevOps Project

1. Search - DevOps - from the Azure Portal Search**

2. Create Organization in Azure DevOps

Go to --> https://dev.azure.com/ --> Select "New Organization"

3. Create Project inside the Organization --> Select "New Project"
   • This will create a Repository
   • Wiki
   • Project Board

Once created you will see below the console

================ END OF TASK 2 =======================================

Task 3 - Create ML Workspace Manually (Testing)

Create ML Workspace

1. Search - "Azure Machine Learning - from the Azure Portal Search**

2. Click on "Create" - Wait for 5-10 minutes till Workspace get created.

3. Open Azure ML Workspace Studio URL

Explore Azure ML Studio

• Check Notebooks

• Data

• Jobs and Experiements

• Model

• Endpoints

• Compute

• Data Stores

Create Compute Instance

• Go to Manage - Compute

Click Create

Test Data Information

• Go to Assets --> Data --> Create --> From Local Files

• Browse and Select the files from local

• You can see the Data Discovered and Preview., Also data schema type

• Enable Checkbox of "Profile the Data Set" and select the Compute created in before steps"

================ END OF TASK 3 =======================================

Task 4 - Create Service Principal and ML Extension to Azure DevOps

Create Service Connection

• Go to https://dev.azure.com/ --> Select your project

• Select Project Settings

or https://dev.azure.com///_settings

• Click Service Connection. --> New Service Connection --> Select "Azure Resource Manager"

• Select "Service Principal (automatic) for Azure DevOps

• Select "Subscription" - We need this to be created, In Order for Azure DevOps to create the required resources inside the Resource group.

• Select "Service Principal (automatic) for Azure DevOps to Azure ML

• Select "Machine Learning Workspace " - We need this to be created, In Order for Azure DevOps to create the resources in the Azure Workspace via Azure Pipeline

Install ML Extension to Azure DevOps Project

• Install Machine Learning Extention to Azure DevOps Click on the Highlight red one --> Manage Extension --> Browser Market Place

Select Machine Learning

================ END OF TASK 4 =======================================

Task 5 - Create ML Workspace via IaaC using Azure DevOps

Create Variables

• Go to the Azure Projected created in #Task 2
• Let us create Required Variables in the Library Section Pipeline

• Create a Variable group

AZURE_RM_SVC_CONNECTION	mlops-demo-azure-devops-service-connection
BASE_NAME	mlopsblrws
LOCATION	eastus2
RESOURCE_GROUP	mlops-demo-blr-resource-group
WORKSPACE_NAME	mlops-blr-ws-aml
WORKSPACE_SVC_CONNECTION	mlops-aml-service-connection

Create Pipeline

Click on New Pipeline

• Select "Azure Repos Git"

• Select the Repo and Cli

• Select "Existing Azure Pipeline YAML File"

• Provide the Path of YAML

• Run the Pipeline
  

Validate the Results

• Select the Pipeline and go to the Run and Make sure it is GREEN

Check the Jobs for Pipeline Run logs

• Select the recent logs --> go to Jobs and check the Jobs

================ END OF TASK 5 =======================================

Task 6 - Create ML Pipeline Continuous Integration

• Follow the same procedure as above to create a Pipeline but select different path for creating MLOps Pipeline

• Create a Variables.

Name	Value
aml.sp	Get the Object ID from DevOps Service Principle
amlcompute.clusterName	mldemocluster
amlcompute.idleSecondsBeforeScaledown	300
amlcompute.maxNodes	2
amlcompute.minNodes	0
amlcompute.vmSize	STANDARD_DS2_V2
azureml.location	eastus2
azureml.resourceGroup	mlops-demo-blr-resource-group
azureml.workspaceName	mlops-blr-ws-aml

• Run the Pipeline - It takes 10 Minutes to execute.

• Select the Run and Click on "Agent Job 1" - and You can see all Tasks in Pipeline succeeded.

================ END OF TASK 6 =======================================

Task 7 - Create Deployment Pipeline

• Go to Azure devops project --> Pipelines --> Releases --> Create New Release

• You can select the Artifacts from the pipeline created before. - Click Add and Artifact

• Add a Stage to perform the deployment - The best practice is to Create a Deployment to staging environment as part of (Stage 2 )and if succeed deploy to Prod.

• For Demo Purpose, I am directly deploying to the Production environment.

• Click Add - New Agent will be created. - Rename to ML-Deployment-Agent

• Create Variables

Name	Value
aks.clusterName	Select your existing cluster
aks.vmSize	Standard_B2ms
azureml.resourceGroup	mlops-demo-blr-resource-group
azureml.workspaceName	mlops-blr-ws-aml
service.name.prod	insurance-ml-model-prod
aks.agentCount	3

• Set the Agent Job
  1. Rename the Display Name
  2. Agent Pool --> Azure Pipeline
  3. Agent Specification --> Ubuntu-18.04

• Click on Add task from the Agent

Search for Python Version

AZ CLI ML Extension

• Click on Add task from the Agent -> Search Azure CLI -> Change Display - Add ML Extension

• Inline script - Add below Value az extension add -n azure-cli-ml

AZ CLI ML Extension - Deploy to AKS

• Click on Add task from the Agent -> Search Azure CLI -> Change Display -Deploy to AKS

  • Inline script - Add below Value az ml model deploy -g $(azureml.resourceGroup) -w $(azureml.workspaceName) -n $(service.name.prod) -f ../metadata/model.json --dc aksDeploymentConfigProd.yml --ic inferenceConfig.yml --ct $(aks.clusterName) --overwrite

Create Release

• 

• Create

• Validate the Status of Pipeline

================ END OF TASK 7 =======================================

Task 8 - Validate Model Endpoint

• Go to Azure Machine learning Workspace
• Validate the Model - You can see the New Model Created

• Validate Endpoint.

• Validate with test data - Step 1 - Get the Python Code
  • Go to Assets - Endpoints -- Consume - Check Python and Take the code

• Validate with test data - Step 1 - Test in Notebook

  • Go to Author - Notebook -- Consume - Create the Notebook file. -> Paste the Python Code. -> Change the Value of the Data as mentioned below --> Run the Cell

Replace the line 8 Data with below Values

data = {'data': [[0,1,8,1,0,0,1,0,0,0,0,0,0,0,12,1,0,0,0.5,0.3,0.610327781,7,1,-1,0,-1,1,1,1,2,1,65,1,0.316227766,0.669556409,0.352136337,3.464101615,0.1,0.8,0.6,1,1,6,3,6,2,9,1,1,1,12,0,1,1,0,0,1],[4,2,5,1,0,0,0,0,1,0,0,0,0,0,5,1,0,0,0.9,0.5,0.771362431,4,1,-1,0,0,11,1,1,0,1,103,1,0.316227766,0.60632002,0.358329457,2.828427125,0.4,0.5,0.4,3,3,8,4,10,2,7,2,0,3,10,0,0,1,1,0,1]]}

• Monitor using the Application Insights using the Application Insight URL from Endpoints.

• Go to Assets - Endpoints - Details --> Application Insights URL --> Click the URL

• You can see the Application throughput here.

================ END OF LAB =======================================