adf-validation-rules

Azure Data Factory Validation Rules and Limitations

🚨 CRITICAL: Activity Nesting Limitations

Azure Data Factory has STRICT nesting rules for control flow activities. Violating these rules will cause pipeline failures or prevent pipeline creation.

Supported Control Flow Activities for Nesting

Four control flow activities support nested activities:

• ForEach: Iterates over collections and executes activities in a loop
• If Condition: Branches based on true/false evaluation
• Until: Implements do-until loops with timeout options
• Switch: Evaluates activities matching case conditions

✅ PERMITTED Nesting Combinations

Parent Activity	Can Contain	Notes
ForEach	If Condition	✅ Allowed
ForEach	Switch	✅ Allowed
Until	If Condition	✅ Allowed
Until	Switch	✅ Allowed

❌ PROHIBITED Nesting Combinations

Parent Activity	CANNOT Contain	Reason
If Condition	ForEach	❌ Not supported - use Execute Pipeline workaround
If Condition	Switch	❌ Not supported - use Execute Pipeline workaround
If Condition	Until	❌ Not supported - use Execute Pipeline workaround
If Condition	Another If	❌ Cannot nest If within If
Switch	ForEach	❌ Not supported - use Execute Pipeline workaround
Switch	If Condition	❌ Not supported - use Execute Pipeline workaround
Switch	Until	❌ Not supported - use Execute Pipeline workaround
Switch	Another Switch	❌ Cannot nest Switch within Switch
ForEach	Another ForEach	❌ Single level only - use Execute Pipeline workaround
Until	Another Until	❌ Single level only - use Execute Pipeline workaround
ForEach	Until	❌ Single level only - use Execute Pipeline workaround
Until	ForEach	❌ Single level only - use Execute Pipeline workaround

🚫 Special Activity Restrictions

Validation Activity:

• ❌ CANNOT be placed inside ANY nested activity
• ❌ CANNOT be used within ForEach, If, Switch, or Until activities
• ✅ Must be at pipeline root level only

🔧 Workaround: Execute Pipeline Pattern

The ONLY supported workaround for prohibited nesting combinations:

Instead of direct nesting, use the Execute Pipeline Activity to call a child pipeline:

{
  "name": "ParentPipeline_WithIfCondition",
  "activities": [
    {
      "name": "IfCondition_Parent",
      "type": "IfCondition",
      "typeProperties": {
        "expression": "@equals(pipeline().parameters.ProcessData, 'true')",
        "ifTrueActivities": [
          {
            "name": "ExecuteChildPipeline_WithForEach",
            "type": "ExecutePipeline",
            "typeProperties": {
              "pipeline": {
                "referenceName": "ChildPipeline_ForEachLoop",
                "type": "PipelineReference"
              },
              "parameters": {
                "ItemList": "@pipeline().parameters.Items"
              }
            }
          }
        ]
      }
    }
  ]
}

Child Pipeline Structure:

{
  "name": "ChildPipeline_ForEachLoop",
  "parameters": {
    "ItemList": {"type": "array"}
  },
  "activities": [
    {
      "name": "ForEach_InChildPipeline",
      "type": "ForEach",
      "typeProperties": {
        "items": "@pipeline().parameters.ItemList",
        "activities": [
          // Your ForEach logic here
        ]
      }
    }
  ]
}

Why This Works:

• Each pipeline can have ONE level of nesting
• Execute Pipeline creates a new pipeline context
• Child pipeline gets its own nesting level allowance
• Enables unlimited depth through pipeline chaining

🔢 Activity and Resource Limits

Pipeline Limits

Resource	Limit	Notes
Activities per pipeline	80	Includes inner activities for containers
Parameters per pipeline	50	-
ForEach concurrent iterations	50 (maximum)	Set via batchCount property
ForEach items	100,000	-
Lookup activity rows	5,000	Maximum rows returned
Lookup activity size	4 MB	Maximum size of returned data
Web activity timeout	1 hour	Default timeout for Web activities
Copy activity timeout	7 days	Maximum execution time

ForEach Activity Configuration

{
  "name": "ForEachActivity",
  "type": "ForEach",
  "typeProperties": {
    "items": "@pipeline().parameters.ItemList",
    "isSequential": false,  // false = parallel execution
    "batchCount": 50,       // Max 50 concurrent iterations
    "activities": [
      // Nested activities
    ]
  }
}

Critical Considerations:

• isSequential: true → Executes one item at a time (slow but predictable)
• isSequential: false → Executes up to batchCount items in parallel
• Maximum batchCount is 50 regardless of setting
• Cannot use Set Variable activity inside parallel ForEach (variable scope is pipeline-level)

Set Variable Activity Limitations

❌ CANNOT use Set Variable inside ForEach with isSequential: false

• Reason: Variables are pipeline-scoped, not ForEach-scoped
• Multiple parallel iterations would cause race conditions
• ✅ Alternative: Use Append Variable with array type, or use sequential execution

📊 Linked Services: Azure Blob Storage

Authentication Methods

1. Account Key (Basic)

{
  "type": "AzureBlobStorage",
  "typeProperties": {
    "connectionString": {
      "type": "SecureString",
      "value": "DefaultEndpointsProtocol=https;AccountName=<account>;AccountKey=<key>"
    }
  }
}

⚠️ Limitations:

• Secondary Blob service endpoints are NOT supported
• Security Risk: Account keys should be stored in Azure Key Vault

2. Shared Access Signature (SAS)

{
  "type": "AzureBlobStorage",
  "typeProperties": {
    "sasUri": {
      "type": "SecureString",
      "value": "https://<account>.blob.core.windows.net/<container>?<SAS-token>"
    }
  }
}

Critical Requirements:

• Dataset folderPath must be absolute path from container level
• SAS token expiry must extend beyond pipeline execution
• SAS URI path must align with dataset configuration

3. Service Principal

{
  "type": "AzureBlobStorage",
  "typeProperties": {
    "serviceEndpoint": "https://<account>.blob.core.windows.net",
    "accountKind": "StorageV2",  // REQUIRED for service principal
    "servicePrincipalId": "<client-id>",
    "servicePrincipalCredential": {
      "type": "SecureString",
      "value": "<client-secret>"
    },
    "tenant": "<tenant-id>"
  }
}

Critical Requirements:

• accountKind MUST be set (StorageV2, BlobStorage, or BlockBlobStorage)
• Service Principal requires Storage Blob Data Reader (source) or Storage Blob Data Contributor (sink) role
• ❌ NOT compatible with soft-deleted blob accounts in Data Flow

4. Managed Identity (Recommended)

{
  "type": "AzureBlobStorage",
  "typeProperties": {
    "serviceEndpoint": "https://<account>.blob.core.windows.net",
    "accountKind": "StorageV2"  // REQUIRED for managed identity
  },
  "connectVia": {
    "referenceName": "AutoResolveIntegrationRuntime",
    "type": "IntegrationRuntimeReference"
  }
}

Critical Requirements:

• accountKind MUST be specified (cannot be empty or "Storage")
• ❌ Empty or "Storage" account kind will cause Data Flow failures
• Managed identity must have Storage Blob Data Reader/Contributor role assigned
• For Storage firewall: Must enable "Allow trusted Microsoft services"

Common Blob Storage Pitfalls

Issue	Cause	Solution
Data Flow fails with managed identity	accountKind empty or "Storage"	Set accountKind to StorageV2
Secondary endpoint doesn't work	Using account key auth	Not supported - use different auth method
SAS token expired during run	Token expiry too short	Extend SAS token validity period
Cannot access $logs container	System container not visible in UI	Use direct path reference
Soft-deleted blobs inaccessible	Service principal/managed identity	Use account key or SAS instead
Private endpoint connection fails	Wrong endpoint for Data Flow	Ensure ADLS Gen2 private endpoint exists

📊 Linked Services: Azure SQL Database

Authentication Methods

1. SQL Authentication

{
  "type": "AzureSqlDatabase",
  "typeProperties": {
    "server": "<server-name>.database.windows.net",
    "database": "<database-name>",
    "authenticationType": "SQL",
    "userName": "<username>",
    "password": {
      "type": "SecureString",
      "value": "<password>"
    }
  }
}

Best Practice:

• Store password in Azure Key Vault
• Use connection string with Key Vault reference

2. Service Principal

{
  "type": "AzureSqlDatabase",
  "typeProperties": {
    "server": "<server-name>.database.windows.net",
    "database": "<database-name>",
    "authenticationType": "ServicePrincipal",
    "servicePrincipalId": "<client-id>",
    "servicePrincipalCredential": {
      "type": "SecureString",
      "value": "<client-secret>"
    },
    "tenant": "<tenant-id>"
  }
}

Requirements:

• Microsoft Entra admin must be configured on SQL server
• Service principal must have contained database user created
• Grant appropriate role: db_datareader, db_datawriter, etc.

3. Managed Identity

{
  "type": "AzureSqlDatabase",
  "typeProperties": {
    "server": "<server-name>.database.windows.net",
    "database": "<database-name>",
    "authenticationType": "SystemAssignedManagedIdentity"
  }
}

Requirements:

• Create contained database user for managed identity
• Grant appropriate database roles
• Configure firewall to allow Azure services (or specific IP ranges)

SQL Database Configuration Best Practices

Connection String Parameters

Server=tcp:<server>.database.windows.net,1433;
Database=<database>;
Encrypt=mandatory;          // Options: mandatory, optional, strict
TrustServerCertificate=false;
ConnectTimeout=30;
CommandTimeout=120;
Pooling=true;
ConnectRetryCount=3;
ConnectRetryInterval=10;

Critical Parameters:

• Encrypt: Default is mandatory (recommended)
• Pooling: Set to false if experiencing idle connection issues
• ConnectRetryCount: Recommended for transient fault handling
• ConnectRetryInterval: Seconds between retries

Common SQL Database Pitfalls

Issue	Cause	Solution
Serverless tier auto-paused	Pipeline doesn't wait for resume	Implement retry logic or keep-alive
Connection pool timeout	Idle connections closed	Add Pooling=false or configure retry
Firewall blocks connection	IP not whitelisted	Add Azure IR IPs or enable Azure services
Always Encrypted fails in Data Flow	Not supported for sink	Use service principal/managed identity in copy activity
Decimal precision loss	Copy supports up to 28 precision	Use string type for higher precision
Parallel copy not working	No partition configuration	Enable physical or dynamic range partitioning

Performance Optimization

Parallel Copy Configuration

{
  "source": {
    "type": "AzureSqlSource",
    "partitionOption": "PhysicalPartitionsOfTable"  // or "DynamicRange"
  },
  "parallelCopies": 8,  // Recommended: (DIU or IR nodes) × (2 to 4)
  "enableStaging": true,
  "stagingSettings": {
    "linkedServiceName": {
      "referenceName": "AzureBlobStorage",
      "type": "LinkedServiceReference"
    }
  }
}

Partition Options:

• PhysicalPartitionsOfTable: Uses SQL Server physical partitions
• DynamicRange: Creates logical partitions based on column values
• None: No partitioning (default)

Staging Best Practices:

• Always use staging for large data movements (> 1GB)
• Use PolyBase or COPY statement for best performance
• Parquet format recommended for staging files

🔍 Data Flow Limitations

General Limits

• Column name length: 128 characters maximum
• Row size: 1 MB maximum (some sinks like SQL have lower limits)
• String column size: Varies by sink (SQL: 8000 for varchar, 4000 for nvarchar)

Transformation-Specific Limits

Transformation	Limitation
Lookup	Cache size limited by cluster memory
Join	Large joins may cause memory errors
Pivot	Maximum 10,000 unique values
Window	Requires partitioning for large datasets

Performance Considerations

• Partitioning: Always partition large datasets before transformations
• Broadcast: Use broadcast hint for small dimension tables
• Sink optimization: Enable table option "Recreate" instead of "Truncate" for better performance

🛡️ Validation Checklist for Pipeline Creation

Before Creating Pipeline

• Verify activity nesting follows permitted combinations
• Check ForEach activities don't contain other ForEach/Until
• Verify If/Switch activities don't contain ForEach/Until/If/Switch
• Ensure Validation activities are at pipeline root level only
• Confirm total activities < 80 per pipeline
• Verify no Set Variable activities in parallel ForEach

Linked Service Validation

• Blob Storage: If using managed identity/service principal, accountKind is set
• SQL Database: Authentication method matches security requirements
• All services: Secrets stored in Key Vault, not hardcoded
• All services: Firewall rules configured for integration runtime IPs
• Network: Private endpoints configured if using VNet integration

Activity Configuration Validation

• ForEach: batchCount ≤ 50 if parallel execution
• Lookup: Query returns < 5000 rows and < 4 MB data
• Copy: DIU configured appropriately (2-256 for Azure IR)
• Copy: Staging enabled for large data movements
• All activities: Timeout values appropriate for expected execution time
• All activities: Retry logic configured for transient failures

Data Flow Validation

• Column names ≤ 128 characters
• Source query doesn't return > 1 MB per row
• Partitioning configured for large datasets
• Sink has appropriate schema and data type mappings
• Staging linked service configured for optimal performance

🔍 Automated Validation Script

CRITICAL: Always run automated validation before committing or deploying ADF pipelines!

The adf-master plugin includes a comprehensive PowerShell validation script that checks for ALL the rules and limitations documented above.

Using the Validation Script

Location: ${CLAUDE_PLUGIN_ROOT}/scripts/validate-adf-pipelines.ps1

Basic usage:

# From the root of your ADF repository
pwsh -File validate-adf-pipelines.ps1

With custom paths:

pwsh -File validate-adf-pipelines.ps1 `
    -PipelinePath "path/to/pipeline" `
    -DatasetPath "path/to/dataset"

With strict mode (additional warnings):

pwsh -File validate-adf-pipelines.ps1 -Strict

What the Script Validates

The automated validation script checks for issues that Microsoft's official @microsoft/azure-data-factory-utilities package does NOT validate:

1. Activity Nesting Violations:

   • ForEach → ForEach, Until, Validation
   • Until → Until, ForEach, Validation
   • IfCondition → ForEach, If, IfCondition, Switch, Until, Validation
   • Switch → ForEach, If, IfCondition, Switch, Until, Validation

2. Resource Limits:

   • Pipeline activity count (max 120, warn at 100)
   • Pipeline parameter count (max 50)
   • Pipeline variable count (max 50)
   • ForEach batchCount limit (max 50, warn at 30 in strict mode)

3. Variable Scope Violations:

   • SetVariable in parallel ForEach (causes race conditions)
   • Proper AppendVariable vs SetVariable usage

4. Dataset Configuration Issues:

   • Missing fileName or wildcardFileName for file-based datasets
   • AzureBlobFSLocation missing required fileSystem property
   • Missing required properties for DelimitedText, Json, Parquet types

5. Copy Activity Validations:

   • Source/sink type compatibility with dataset types
   • Lookup activity firstRowOnly=false warnings (5000 row/4MB limits)
   • Blob file dependencies (additionalColumns logging pattern)

Integration with CI/CD

GitHub Actions example:

- name: Validate ADF Pipelines
  run: |
    pwsh -File validate-adf-pipelines.ps1 -PipelinePath pipeline -DatasetPath dataset
  shell: pwsh

Azure DevOps example:

- task: PowerShell@2
  displayName: 'Validate ADF Pipelines'
  inputs:
    filePath: 'validate-adf-pipelines.ps1'
    arguments: '-PipelinePath pipeline -DatasetPath dataset'
    pwsh: true

Command Reference

Use the /adf-validate command to run the validation script with proper guidance:

/adf-validate

This command will:

1. Detect your ADF repository structure
2. Run the validation script with appropriate paths
3. Parse and explain any errors or warnings found
4. Provide specific solutions for each violation
5. Recommend next actions based on results
6. Suggest CI/CD integration patterns

Exit Codes

• 0: Validation passed (no errors)
• 1: Validation failed (errors found - DO NOT DEPLOY)

Best Practices

1. Run validation before every commit to catch issues early
2. Add validation to CI/CD pipeline to prevent invalid deployments
3. Use strict mode during development for additional warnings
4. Re-validate after bulk changes or generated pipelines
5. Document validation exceptions if you must bypass a warning
6. Share validation results with team to prevent repeated mistakes

🚨 CRITICAL: Enforcement Protocol

When creating or modifying ADF pipelines:

1. ALWAYS validate activity nesting against the permitted/prohibited table
2. REJECT any attempt to create prohibited nesting combinations
3. SUGGEST Execute Pipeline workaround for complex nesting needs
4. VALIDATE linked service authentication matches the connector type
5. CHECK all limits (activities, parameters, ForEach iterations, etc.)
6. VERIFY required properties are set (e.g., accountKind for managed identity)
7. WARN about common pitfalls specific to the connector being used

Example Validation Response:

❌ INVALID PIPELINE STRUCTURE DETECTED:

Issue: ForEach activity contains another ForEach activity
Location: Pipeline "PL_DataProcessing" → ForEach "OuterLoop" → ForEach "InnerLoop"

This violates Azure Data Factory nesting rules:
- ForEach activities support only a SINGLE level of nesting
- You CANNOT nest ForEach within ForEach

✅ RECOMMENDED SOLUTION:
Use the Execute Pipeline pattern:
1. Create a child pipeline with the inner ForEach logic
2. Replace the inner ForEach with an Execute Pipeline activity
3. Pass required parameters to the child pipeline

Would you like me to generate the refactored pipeline structure?

📚 Reference Documentation

Official Microsoft Learn Resources:

• Activity nesting: https://learn.microsoft.com/en-us/azure/data-factory/concepts-nested-activities
• Blob Storage connector: https://learn.microsoft.com/en-us/azure/data-factory/connector-azure-blob-storage
• SQL Database connector: https://learn.microsoft.com/en-us/azure/data-factory/connector-azure-sql-database
• Pipeline limits: https://learn.microsoft.com/en-us/azure/azure-resource-manager/management/azure-subscription-service-limits#data-factory-limits

Last Updated: 2025-01-24 (Based on official Microsoft documentation)

This validation rules skill MUST be consulted before creating or modifying ANY Azure Data Factory pipeline to ensure compliance with platform limitations and best practices.

Progressive Disclosure References

For detailed validation matrices and resource limits, see:

• Nesting Rules: references/nesting-rules.md - Complete matrix of permitted and prohibited activity nesting combinations with workaround patterns
• Resource Limits: references/resource-limits.md - Complete reference for all ADF limits (pipeline, activity, trigger, data flow, integration runtime, expression, API)