adf-master（ADF 主控器）

Azure Data Factory Master Knowledge Base

Deprecated Features

Apache Airflow Workflow Orchestration Manager - DEPRECATED

Status: Deprecated since early 2025. Available only for existing customers. Retirement Date: Not yet announced, but no new deployments permitted. Impact: New customers cannot provision Apache Airflow in Azure Data Factory.

Deprecation Details:

• Apache Airflow Workflow Orchestration Manager is deprecated with no retirement date set
• Only existing deployments can continue using this feature
• No new Airflow integrations can be created in ADF

Migration Path:

• Recommended: Migrate to Fabric Data Factory with native Airflow support
• Alternative: Use standalone Apache Airflow deployments (Azure Container Instances, AKS, or VM-based)
• Alternative: Migrate orchestration logic to native ADF pipelines with control flow activities

Why Deprecated:

• Microsoft focus shifted to Fabric Data Factory as the unified data integration platform
• Fabric provides modern orchestration capabilities superseding Airflow integration
• Limited adoption and maintenance burden for standalone Airflow feature in ADF

Action Required:

• If using Airflow in ADF: Migrate to Fabric Data Factory, standalone Airflow, or native ADF patterns
• For new projects: Do NOT use Airflow in ADF
• Monitor Microsoft announcements for official retirement timeline

Reference:

• Microsoft Roadmap: https://www.directionsonmicrosoft.com/roadmaps/ref/azure-data-factory-roadmap/

Feature Updates (2025-2026)

Microsoft Fabric Integration (GA)

ADF Mounting in Fabric:

• Bring existing ADF pipelines into Fabric workspaces without rebuilding
• Generally Available since June 2025
• Seamless integration enables hybrid ADF + Fabric workflows

Cross-Workspace Pipeline Orchestration:

• New Invoke Pipeline activity supports cross-platform calls
• Invoke pipelines across Fabric, Azure Data Factory, and Synapse
• Managed VNet support for secure cross-workspace communication

Variable Libraries:

• Environment-specific variables for CI/CD automation
• Automatic value substitution during workspace promotion
• Eliminates separate parameter files per environment

Connector Enhancements:

• ServiceNow V2 (V1 End of Support)
• Enhanced PostgreSQL and Snowflake connectors
• Native OneLake connectivity for zero-copy integration

Node.js 20.x Requirement for CI/CD

CRITICAL: As of 2025, npm package @microsoft/azure-data-factory-utilities requires Node.js 20.x

Breaking Change:

• Older Node.js versions (14.x, 16.x, 18.x) may cause package incompatibility errors
• Update CI/CD pipelines to use Node.js 20.x or compatible versions

GitHub Actions:

- name: Setup Node.js
  uses: actions/setup-node@v4
  with:
    node-version: '20.x'

Azure DevOps:

- task: UseNode@1
  inputs:
    version: '20.x'

Official Documentation Sources

Primary Microsoft Learn Resources

Main Documentation Hub:

• URL: https://learn.microsoft.com/en-us/azure/data-factory/
• Last Updated: February 2025
• Coverage: Complete ADF documentation including tutorials, concepts, how-to guides, and reference materials
• Key Topics: Pipelines, datasets, triggers, linked services, data flows, integration runtimes, monitoring

Introduction to Azure Data Factory:

• URL: https://learn.microsoft.com/en-us/azure/data-factory/introduction
• Summary: Managed cloud service for complex hybrid ETL, ELT, and data integration projects
• Key Features: 90+ built-in connectors, serverless architecture, code-free UI, single-pane monitoring

Context7 Library Documentation

Library ID: /websites/learn_microsoft_en-us_azure_data-factory

• Trust Score: 7.5
• Code Snippets: 10,839
• Topics: CI/CD, ARM templates, pipeline patterns, data flows, monitoring, troubleshooting

How to Access:

Use Context7 MCP tool to fetch latest documentation:
mcp__context7__get-library-docs:
  - context7CompatibleLibraryID: /websites/learn_microsoft_en-us_azure_data-factory
  - topic: "CI/CD continuous integration deployment pipelines ARM templates"
  - tokens: 8000

CI/CD Deployment Methods

Modern Automated Approach (Recommended)

npm Package: @microsoft/azure-data-factory-utilities

• Latest Version: 1.0.3+ (check npm for current version)
• npm URL: https://www.npmjs.com/package/@microsoft/azure-data-factory-utilities
• Node.js Requirement: Version 20.x or compatible

Key Features:

• Validates ADF resources independently of service
• Generates ARM templates programmatically
• Enables true CI/CD without manual publish button
• Supports preview mode for selective trigger management

package.json Configuration:

{
  "scripts": {
    "build": "node node_modules/@microsoft/azure-data-factory-utilities/lib/index",
    "build-preview": "node node_modules/@microsoft/azure-data-factory-utilities/lib/index --preview"
  },
  "dependencies": {
    "@microsoft/azure-data-factory-utilities": "^1.0.3"
  }
}

Commands:

# Validate resources
npm run build validate <rootFolder> <factoryId>

# Generate ARM templates
npm run build export <rootFolder> <factoryId> [outputFolder]

# Preview mode (only stop/start modified triggers)
npm run build-preview export <rootFolder> <factoryId> [outputFolder]

Official Documentation:

• URL: https://learn.microsoft.com/en-us/azure/data-factory/continuous-integration-delivery-improvements
• Last Updated: January 2025
• Topics: Setup, configuration, build commands, CI/CD integration

Traditional Manual Approach (Legacy)

Method: Git integration + Publish button

Process:

1. Configure Git integration in ADF UI (Dev environment only)
2. Make changes in ADF Studio
3. Click "Publish" button to generate ARM templates
4. Templates saved to adf_publish branch
5. Release pipelines deploy from adf_publish branch

When to Use:

• Migrating from existing setup
• No build pipeline infrastructure
• Simple deployments without validation

Limitations:

• Requires manual publish action
• No validation until publish
• Not true CI/CD (manual step required)
• Can't validate on pull requests

Migration Path: Modern approach recommended for new implementations

ARM Template Deployment

PowerShell Deployment

Primary Command: New-AzResourceGroupDeployment

Syntax:

New-AzResourceGroupDeployment `
  -ResourceGroupName "<resource-group-name>" `
  -TemplateFile "ARMTemplateForFactory.json" `
  -TemplateParameterFile "ARMTemplateParametersForFactory.<environment>.json" `
  -factoryName "<factory-name>" `
  -Mode Incremental `
  -Verbose

Validation:

Test-AzResourceGroupDeployment `
  -ResourceGroupName "<resource-group-name>" `
  -TemplateFile "ARMTemplateForFactory.json" `
  -TemplateParameterFile "ARMTemplateParametersForFactory.<environment>.json" `
  -factoryName "<factory-name>"

What-If Analysis:

New-AzResourceGroupDeployment `
  -ResourceGroupName "<resource-group-name>" `
  -TemplateFile "ARMTemplateForFactory.json" `
  -TemplateParameterFile "ARMTemplateParametersForFactory.<environment>.json" `
  -factoryName "<factory-name>" `
  -WhatIf

Azure CLI Deployment

Primary Command: az deployment group create

Syntax:

az deployment group create \
  --resource-group <resource-group-name> \
  --template-file ARMTemplateForFactory.json \
  --parameters ARMTemplateParametersForFactory.<environment>.json \
  --parameters factoryName=<factory-name> \
  --mode Incremental

Validation:

az deployment group validate \
  --resource-group <resource-group-name> \
  --template-file ARMTemplateForFactory.json \
  --parameters ARMTemplateParametersForFactory.<environment>.json \
  --parameters factoryName=<factory-name>

What-If Analysis:

az deployment group what-if \
  --resource-group <resource-group-name> \
  --template-file ARMTemplateForFactory.json \
  --parameters ARMTemplateParametersForFactory.<environment>.json \
  --parameters factoryName=<factory-name>

PrePostDeploymentScript

Current Version: Ver2

Location: https://github.com/Azure/Azure-DataFactory/blob/main/SamplesV2/ContinuousIntegrationAndDelivery/PrePostDeploymentScript.Ver2.ps1

Key Improvement in Ver2:

• Turns off/on ONLY triggers that have been modified
• Ver1 stopped/started ALL triggers (slower, more disruptive)
• Compares trigger payloads to determine changes

Download Command:

# Linux/macOS/Git Bash
curl -o PrePostDeploymentScript.Ver2.ps1 https://raw.githubusercontent.com/Azure/Azure-DataFactory/main/SamplesV2/ContinuousIntegrationAndDelivery/PrePostDeploymentScript.Ver2.ps1

# PowerShell
Invoke-WebRequest -Uri "https://raw.githubusercontent.com/Azure/Azure-DataFactory/main/SamplesV2/ContinuousIntegrationAndDelivery/PrePostDeploymentScript.Ver2.ps1" -OutFile "PrePostDeploymentScript.Ver2.ps1"

Parameters

Pre-Deployment (Stop Triggers):

./PrePostDeploymentScript.Ver2.ps1 `
  -armTemplate "<path-to-ARMTemplateForFactory.json>" `
  -ResourceGroupName "<resource-group-name>" `
  -DataFactoryName "<factory-name>" `
  -predeployment $true `
  -deleteDeployment $false

Post-Deployment (Start Triggers & Cleanup):

./PrePostDeploymentScript.Ver2.ps1 `
  -armTemplate "<path-to-ARMTemplateForFactory.json>" `
  -ResourceGroupName "<resource-group-name>" `
  -DataFactoryName "<factory-name>" `
  -predeployment $false `
  -deleteDeployment $true

PowerShell Requirements

Version: PowerShell Core (7.0+) recommended

• Azure DevOps: Use pwsh: true in AzurePowerShell@5 task
• Locally: Use pwsh command, not powershell

Modules Required:

• Az.DataFactory
• Az.Resources

Official Documentation:

• URL: https://learn.microsoft.com/en-us/azure/data-factory/continuous-integration-delivery-sample-script
• Last Updated: January 2025

GitHub Actions CI/CD

Official Resources

Medium Article (Recent 2025):

• URL: https://medium.com/microsoftazure/azure-data-factory-build-and-deploy-with-new-ci-cd-flow-using-github-actions-cd46c95054e0
• Author: Jared Zagelbaum (Microsoft Azure)
• Topics: Modern CI/CD flow, npm package usage, GitHub Actions setup

Microsoft Community Hub:

• URL: https://techcommunity.microsoft.com/blog/fasttrackforazureblog/azure-data-factory-cicd-with-github-actions/3768493
• Topics: End-to-end GitHub Actions setup, workload identity federation

Community Blog (February 2025):

• URL: https://linusdata.blog/2025/03/14/automating-azure-data-factory-deployments-with-github-actions/
• Topics: Practical implementation guide, troubleshooting tips

Key GitHub Actions

Essential Actions:

• actions/checkout@v4 - Checkout repository
• actions/setup-node@v4 - Setup Node.js
• actions/upload-artifact@v4 - Publish ARM templates
• actions/download-artifact@v4 - Download ARM templates in deploy workflow
• azure/login@v2 - Authenticate to Azure
• azure/arm-deploy@v2 - Deploy ARM templates
• azure/powershell@v2 - Run PrePostDeploymentScript

Authentication Methods

Service Principal (JSON credentials):

{
  "clientId": "<GUID>",
  "clientSecret": "<STRING>",
  "subscriptionId": "<GUID>",
  "tenantId": "<GUID>"
}

Store in GitHub secret: AZURE_CREDENTIALS

Workload Identity Federation (More secure):

• No secrets stored
• Uses OIDC (OpenID Connect)
• Recommended for production
• Setup: https://learn.microsoft.com/en-us/azure/developer/github/connect-from-azure

Azure DevOps CI/CD

Official Resources

Microsoft Learn:

• URL: https://learn.microsoft.com/en-us/azure/data-factory/continuous-integration-delivery-automate-azure-pipelines
• Topics: Build pipeline, release pipeline, service connections, variable groups

Community Guides:

• Adam Marczak Blog: https://marczak.io/posts/2023/02/quick-cicd-for-data-factory/
• Topics: Quick setup, best practices, folder structure

Towards Data Science:

• URL: https://towardsdatascience.com/azure-data-factory-ci-cd-made-simple-building-and-deploying-your-arm-templates-with-azure-devops-30c30595afa5
• Topics: ARM template build and deployment workflow

Key Azure DevOps Tasks

Build Pipeline Tasks:

• UseNode@1 - Install Node.js
• Npm@1 - Install packages, run build commands
• PublishPipelineArtifact@1 - Publish ARM templates

Release Pipeline Tasks:

• DownloadPipelineArtifact@2 - Download ARM templates
• AzurePowerShell@5 - Run PrePostDeploymentScript
• AzureResourceManagerTemplateDeployment@3 - Deploy ARM template

Service Connection Requirements

Permissions Needed:

• Data Factory Contributor (on all Data Factories)
• Contributor (on Resource Groups)
• Key Vault access policies (if using secrets)

Configuration:

• Project Settings → Service connections → New service connection
• Type: Azure Resource Manager
• Authentication: Service Principal (recommended) or Managed Identity

Troubleshooting Resources

Official Troubleshooting Guide

URL: https://learn.microsoft.com/en-us/azure/data-factory/ci-cd-github-troubleshoot-guide Last Updated: January 2025

Common Issues Covered:

1. Template parameter validation errors
2. Integration Runtime type cannot be changed
3. ARM template size exceeds 4MB limit
4. Git connection problems
5. Authentication failures
6. Deployment errors

Diagnostic Logs

Enable Diagnostic Settings:

Azure Portal → Data Factory → Diagnostic settings → Add diagnostic setting
Send to: Log Analytics workspace

Logs to Enable:
- PipelineRuns
- TriggerRuns
- ActivityRuns
- SandboxPipelineRuns
- SandboxActivityRuns

Kusto Queries for Troubleshooting:

// Failed pipeline runs in last 24 hours
ADFPipelineRun
| where Status == "Failed"
| where TimeGenerated > ago(24h)
| project TimeGenerated, PipelineName, RunId, Status, ErrorMessage, Parameters
| order by TimeGenerated desc

// Failed CI/CD deployments
ADFActivityRun
| where ActivityType == "ExecutePipeline"
| where Status == "Failed"
| where TimeGenerated > ago(7d)
| project TimeGenerated, PipelineName, ActivityName, ErrorCode, ErrorMessage
| order by TimeGenerated desc

// Performance analysis
ADFActivityRun
| where TimeGenerated > ago(7d)
| extend DurationMinutes = datetime_diff('minute', End, Start)
| summarize AvgDuration = avg(DurationMinutes) by ActivityType, ActivityName
| where AvgDuration > 10
| order by AvgDuration desc

Common Error Patterns

Error: "Template parameters are not valid"

• Cause: Deleted triggers still referenced in parameters
• Solution: Regenerate ARM template or use PrePostDeploymentScript cleanup

Error: "Updating property type is not supported"

• Cause: Trying to change Integration Runtime type
• Solution: Delete and recreate IR (not in-place update)

Error: "Operation timed out"

• Cause: Network connectivity, large data volume, insufficient compute
• Solution: Increase timeout, optimize query, increase DIUs

Error: "Authentication failed"

• Cause: Service principal expired, missing permissions, wrong credentials
• Solution: Verify credentials, check role assignments, renew if expired

Best Practices

Repository Structure

Recommended Folder Layout:

repository-root/
├── adf-resources/          # ADF JSON files (if using npm approach)
│   ├── dataset/
│   ├── pipeline/
│   ├── trigger/
│   ├── linkedService/
│   └── integrationRuntime/
├── .github/
│   └── workflows/          # GitHub Actions workflows
│       ├── adf-build.yml
│       └── adf-deploy.yml
├── azure-pipelines/        # Azure DevOps pipelines
│   ├── build.yml
│   └── release.yml
├── parameters/             # Environment-specific parameters
│   ├── ARMTemplateParametersForFactory.dev.json
│   ├── ARMTemplateParametersForFactory.test.json
│   └── ARMTemplateParametersForFactory.prod.json
├── package.json            # npm configuration
└── README.md

Git Configuration

Only Configure Git on Development ADF:

• Development: Git-integrated for source control
• Test: CI/CD deployment only (no Git)
• Production: CI/CD deployment only (no Git)

Rationale: Prevents accidental manual changes in higher environments

Multi-Environment Strategy

Environment Flow:
Dev (Git) → Build → Test → Approval → Production
            ↓
        ARM Templates

Parameter Management:

• Separate parameter file per environment
• Store secrets in Azure Key Vault
• Reference Key Vault in parameter files
• Never commit secrets to source control

Monitoring and Alerting

Set up alerts for:

• Build pipeline failures
• Deployment failures
• Pipeline run failures
• Performance degradation
• Cost anomalies

Recommended Tools:

• Azure Monitor (Metrics and Alerts)
• Log Analytics (Kusto queries)
• Application Insights (for custom logging)
• Azure Advisor (optimization recommendations)

Additional Resources

GitHub Repositories

Official Azure Data Factory Samples:

• URL: https://github.com/Azure/Azure-DataFactory
• Path: SamplesV2/ContinuousIntegrationAndDelivery/
• Contents: PrePostDeploymentScript.Ver2.ps1, example pipelines, documentation

Community Examples:

• Search GitHub for "azure-data-factory-cicd" for real-world examples
• Many organizations publish their CI/CD patterns as reference

Community Support

Microsoft Q&A:

• URL: https://learn.microsoft.com/en-us/answers/tags/130/azure-data-factory
• Active community, Microsoft employees respond

Stack Overflow:

• Tag: azure-data-factory
• Large knowledge base of resolved issues

Azure Status:

• URL: https://status.azure.com
• Check for service outages and incidents

When to Fetch Latest Information

Situations requiring current documentation:

1. npm package version updates
2. New ADF features or activities
3. Changes to ARM template schema
4. Updates to PrePostDeploymentScript
5. New GitHub Actions or Azure DevOps tasks
6. Breaking changes or deprecations

How to Fetch:

• Use WebFetch for Microsoft Learn articles
• Check npm for latest package version
• Use Context7 for comprehensive topic coverage
• Review Azure Data Factory GitHub for script updates

This knowledge base should be your starting point for all Azure Data Factory questions. Always verify critical information with the latest official documentation when making production decisions.

Progressive Disclosure References

For detailed JSON schemas and complete reference materials, see:

• Activity Types: references/activity-types.md - Complete JSON schemas for all activity types (Copy, ForEach, IfCondition, Switch, Until, Lookup, ExecutePipeline, WebActivity, DatabricksJob, SetVariable, AppendVariable, Wait, Fail, GetMetadata)
• Expression Functions: references/expression-functions.md - Complete reference for all ADF expression functions (string, collection, logical, conversion, math, date/time, pipeline/activity references)
• Linked Services: references/linked-services.md - Complete JSON configurations for all connector types (Blob Storage, ADLS Gen2, Azure SQL, Synapse, Fabric Lakehouse/Warehouse, Databricks, Key Vault, REST, SFTP, Snowflake, PostgreSQL)
• Triggers: references/triggers.md - Complete JSON schemas for schedule, tumbling window, and event triggers
• Datasets: references/datasets.md - Complete JSON schemas for all dataset types with parameterization patterns

For machine learning and analytics patterns, see the dedicated skill:

• ML & Analytics: adf-master:adf-ml-analytics - Azure ML pipelines, batch endpoints, Azure AI Services, Databricks ML/MLflow, SQL-to-Storage archival, feature engineering with Data Flows