Azure Functions

Azure Functions Dev/Test Workflow

This skill helps tracer engineers test changes to the Datadog.AzureFunctions package: build a local dev version, deploy it to a test Azure Function App, trigger the function, and analyze traces/logs to verify instrumentation behavior.

Prerequisites

This skill requires the following tools (assume they are installed and only troubleshoot if errors occur):

• PowerShell: pwsh (PowerShell 7+) preferred, or powershell.exe (PowerShell 5.1+ on Windows)
  • Always prefer pwsh over powershell.exe when available
  • Scripts use PowerShell-specific cmdlets like Expand-Archive
• Azure CLI: az (must be authenticated)
• Azure Functions Core Tools: func
• .NET SDK: Matching target framework of sample app

Only if a tool fails, provide installation links:

• PowerShell: https://learn.microsoft.com/en-us/powershell/scripting/install/installing-powershell
• Azure CLI: https://learn.microsoft.com/en-us/cli/azure/install-azure-cli
• Azure Functions Core Tools: https://learn.microsoft.com/en-us/azure/azure-functions/functions-run-local
• .NET SDK: https://dotnet.microsoft.com/download

Commands

When invoked with an argument, perform the corresponding workflow:

• /azure-functions build-nuget - Build the Datadog.AzureFunctions NuGet package
• /azure-functions deploy [app-name] - Deploy to Azure Function App
• /azure-functions configure [app-name] - Configure environment variables for Datadog instrumentation
• /azure-functions test [app-name] - Trigger and verify function execution
• /azure-functions logs [app-name] - Download and analyze logs

If no argument is provided, guide the user through the full workflow interactively.

Context

Current repository: This skill assumes you are working from the root of the dd-trace-dotnet repository.

Prerequisites: Users provide their own Azure Function App name (-AppName), resource group (-ResourceGroup), and sample app path (-SampleAppPath). The sample app must:

1. Reference the Datadog.AzureFunctions NuGet package
2. Have a nuget.config file (in the app directory or a parent directory) that defines a local NuGet feed pointing to a directory on disk, for example:

   <?xml version="1.0" encoding="utf-8"?>
   <configuration>
     <packageSources>
       <clear />
       <add key="local" value="nuget/local-source" />
       <add key="nuget" value="https://api.nuget.org/v3/index.json" />
     </packageSources>
   </configuration>
   

   The -CopyTo parameter of Build-AzureFunctionsNuget.ps1 must point to the same directory as the local feed (e.g. if the feed value is nuget/local-source relative to the nuget.config location, then -CopyTo should be the absolute path to that directory).

Workflow Steps

1. Build NuGet Package

CRITICAL: Before building, verify that tracer/src/Datadog.AzureFunctions/Datadog.AzureFunctions.csproj uses PackageReference (not ProjectReference) for Datadog.Trace and Datadog.Trace.Annotations. This ensures the locally-built package references the latest releases from nuget.org instead of building them from source.

Check the .csproj — it should contain:

<PackageReference Include="Datadog.Trace" Version="*"/>
<PackageReference Include="Datadog.Trace.Annotations" Version="*" />

If instead it contains ProjectReference lines like these, replace them with the PackageReference lines above:

<!-- These are the production references — replace for local testing: -->
<ProjectReference Include="$(MSBuildThisFileDirectory)..\Datadog.Trace.Manual\Datadog.Trace.Manual.csproj" />
<ProjectReference Include="$(MSBuildThisFileDirectory)..\Datadog.Trace.Annotations\Datadog.Trace.Annotations.csproj" />

IMPORTANT: The PackageReference change is for local testing only. Do NOT commit it. If it's already using PackageReference, no change is needed.

Build the Datadog.AzureFunctions NuGet package with your changes:

./tracer/tools/Build-AzureFunctionsNuget.ps1 -CopyTo <output-dir>

What this does:

1. (If -BuildId specified and files not already downloaded) Downloads bundle from Azure DevOps build once
2. Generates a unique prerelease version from a timestamp (e.g. 3.38.0-dev20260209143022)
3. Cleans previous builds
4. Builds Datadog.Trace (net6.0)
5. Publishes to bundle folder
6. Packages Datadog.AzureFunctions.nupkg with the generated version (referencing latest nuget.org releases)
7. Copies to the directory specified by -CopyTo

Versioning: Each build gets a unique version, so NuGet caching is never an issue. The sample app should use a floating version like 3.38.0-dev.* in its package reference (or Directory.Packages.props) to always resolve the latest local dev build.

Options:

• -CopyTo <output-dir> - Copy the built package to the specified directory (typically your local NuGet feed)
• -Version '3.38.0-dev.custom' - Use a specific version instead of auto-generating
• -BuildId 12345 - One-time download of bundle files from Azure DevOps build (only needed once per dd-trace-dotnet release, then reused for subsequent local builds)

Examples:

# Typical local build (after bundle files already downloaded)
./tracer/tools/Build-AzureFunctionsNuget.ps1 -CopyTo <output-dir>

# First build after new dd-trace-dotnet release (download bundle files once)
./tracer/tools/Build-AzureFunctionsNuget.ps1 -BuildId 12345 -CopyTo <output-dir>

2. Deploy Function

IMPORTANT: Before deploying, verify prerequisites:

Verify nuget.config exists:

$nugetConfig = ./.claude/skills/azure-functions/Find-NuGetConfig.ps1 -StartPath "<path-to-sample-app>"
if (-not $nugetConfig) {
    Write-Error "nuget.config not found in sample app directory or parent directories"
    exit 1
}
Write-Host "Found nuget.config at: $nugetConfig"

Verify environment variables are configured (skip if already done on a previous deploy):

$envCheck = ./.claude/skills/azure-functions/Test-EnvVars.ps1 -AppName "<app-name>" -ResourceGroup "<resource-group>"
if (-not $envCheck.AllRequiredPresent) {
    Write-Warning "Required environment variables are missing. Run '/azure-functions configure' first, or proceed if you plan to configure after deploying."
}

Use the Deploy-AzureFunction.ps1 script to automate deployment, wait, and trigger:

./tracer/tools/Deploy-AzureFunction.ps1 `
  -AppName "<app-name>" `
  -ResourceGroup "<resource-group>" `
  -SampleAppPath "<path-to-sample-app>"

What this does:

1. Runs dotnet restore in the sample app directory
2. Publishes to Azure with func azure functionapp publish
3. Waits 60 seconds (default) for worker process to restart
4. Triggers the HTTP endpoint and captures execution timestamp
5. Outputs a result object for pipeline usage

Options:

• -SkipBuild - Skip dotnet restore
• -SkipWait - Skip 2-minute wait (not recommended)
• -WaitSeconds 60 - Custom wait duration
• -SkipTrigger - Skip HTTP trigger
• -TriggerUrl "https://..." - Custom trigger URL

Note: -AppName and -ResourceGroup are required parameters.

Pipeline usage (save output for log analysis):

$deploy = ./tracer/tools/Deploy-AzureFunction.ps1 `
  -AppName "<app-name>" `
  -ResourceGroup "<resource-group>" `
  -SampleAppPath "<path-to-sample-app>"

3. Configure Environment Variables

Configure Datadog instrumentation environment variables for an Azure Function App using Set-EnvVars.ps1:

./.claude/skills/azure-functions/Set-EnvVars.ps1 `
  -AppName "<app-name>" `
  -ResourceGroup "<resource-group>" `
  -ApiKey "<api-key>" `
  -Tier recommended `
  -Env "dev-lucas"

What this does:

1. Detects the OS/platform (Linux or Windows)
2. Builds the correct set of variables for the chosen tier
3. Sets platform-specific profiler paths automatically
4. Applies all settings via Azure CLI
5. Restarts the function app (unless -SkipRestart)

Tiers:

• required - Minimum variables for instrumentation (CORECLR_*, DD_API_KEY, DD_DOTNET_TRACER_HOME, DOTNET_STARTUP_HOOKS)
• recommended - required + feature disables (AppSec, CI Visibility, RCM, Agent Feature Polling, Process)
• debug - recommended + debug logging (DD_TRACE_DEBUG, DD_LOG_LEVEL, DD_TRACE_LOG_SINKS, direct log submission)

Options:

• -Tier required|recommended|debug - Configuration tier (default: required)
• -Env "dev-lucas" - Set DD_ENV
• -Service "my-service" - Set DD_SERVICE
• -Version "1.0.0" - Set DD_VERSION
• -SamplingRules '<json>' - Set DD_TRACE_SAMPLING_RULES
• -ExtraSettings @{"KEY"="value"} - Set additional variables
• -SkipRestart - Don't restart the app after applying
• -WhatIf - Preview changes without applying

Preview before applying:

./.claude/skills/azure-functions/Set-EnvVars.ps1 `
  -AppName "<app-name>" `
  -ResourceGroup "<resource-group>" `
  -ApiKey "<api-key>" `
  -Tier debug `
  -WhatIf

Complete reference: See environment-variables.md for all available variables.

4. Test Function

Trigger an already-deployed function and capture the execution timestamp. Useful for re-testing after a deploy, or testing an app that was deployed earlier.

Discover available triggers:

# List HTTP-triggered functions and their URLs
func azure functionapp list-functions <app-name> --show-keys

Or via Azure CLI:

az functionapp function list --name <app-name> --resource-group <resource-group> --query "[].{name:name, href:invokeUrlTemplate}" -o table

Trigger and capture timestamp:

$timestamp = (Get-Date).ToUniversalTime().ToString("yyyy-MM-dd HH:mm:ss")
$response = Invoke-WebRequest -Uri "<trigger-url>" -UseBasicParsing
Write-Host "HTTP Status: $($response.StatusCode)"
Write-Host "Execution timestamp (UTC): $timestamp"

Save the timestamp — you'll need it for log filtering in the next step.

Note: The Deploy script (step 2) already triggers and captures a timestamp. Use this step when you want to re-test without redeploying.

5. Download and Analyze Logs

Use the Get-AzureFunctionLogs.ps1 script to download, extract, and analyze logs:

./tracer/tools/Get-AzureFunctionLogs.ps1 `
  -AppName "<app-name>" `
  -ResourceGroup "<resource-group>" `
  -OutputPath $env:TEMP `
  -ExecutionTimestamp "<YYYY-MM-DD HH:MM:SS>" `
  -All

IMPORTANT: Always specify -OutputPath $env:TEMP to save logs to the system temp folder instead of cluttering the repository directory.

What this does:

1. Downloads logs from Azure to a timestamped zip file
2. Extracts the archive
3. Identifies host and worker log files
4. Analyzes tracer version, span count, and trace parenting

Analysis options:

• -ShowVersion - Display Datadog tracer version from worker logs
• -ShowSpans - Count spans at execution timestamp (split by host/worker)
• -CheckParenting - Validate trace parenting (detect root span duplication)
• -All - Enable all analysis (recommended)

Pipeline usage (with Deploy script):

$deploy = ./tracer/tools/Deploy-AzureFunction.ps1 `
  -AppName "<app-name>" `
  -ResourceGroup "<resource-group>" `
  -SampleAppPath "<path-to-sample-app>"

./tracer/tools/Get-AzureFunctionLogs.ps1 `
  -AppName $deploy.AppName `
  -ResourceGroup "<resource-group>" `
  -OutputPath $env:TEMP `
  -ExecutionTimestamp $deploy.ExecutionTimestamp `
  -All

Log file patterns:

• Host process: dotnet-tracer-managed-Microsoft.Azure.WebJobs.Script.WebHost-{pid}.log
• Worker process: dotnet-tracer-managed-dotnet-{pid}.log

CRITICAL: The script automatically filters logs by execution timestamp. Raw log files are append-only and contain entries from multiple deployments/restarts.

Parenting analysis:

• Span fields: s_id (span ID), p_id (parent ID), t_id (trace ID)
• Healthy trace: Worker spans have same t_id as host, p_id matching host span IDs
• Broken trace: Worker spans have different t_id or p_id: null (orphaned root)

Verification Checklist

After deployment and testing:

• Function responds successfully (HTTP 200)
• Worker loaded correct tracer version (check "Assembly metadata")
• Host and worker spans share same trace ID
• Span parent-child relationships are correct (check p_id → s_id links)
• Process tags are correct (aas.function.process:host or worker)
• No error logs at execution timestamp
• AsyncLocal context flows correctly (if debugging context issues)

Common Troubleshooting

General guidance: For environment variable configuration issues, see environment-variables.md for complete reference on required, recommended, and debugging variables.

Function Not Responding

First, check if the app is running:

$envCheck = ./.claude/skills/azure-functions/Test-EnvVars.ps1 -AppName "<app-name>" -ResourceGroup "<resource-group>"
if ($envCheck.State -ne "Running") {
    Write-Host "App is '$($envCheck.State)' — starting it..."
    az functionapp start --name <app-name> --resource-group <resource-group>
}

If the app is running but not responding:

# Restart function app
az functionapp restart --name <app-name> --resource-group <resource-group>

Deployment Fails (func azure functionapp publish)

If the publish command fails, diagnose with:

# Check recent deployment status
az functionapp deployment list \
  --name <app-name> \
  --resource-group <resource-group> \
  --query "[0].{status:status, message:message, startTime:startTime}" -o table

# Stream live logs to see startup errors
func azure functionapp logstream <app-name>

Common causes:

• Auth expired: Run az login and retry
• App not running: Start it first with az functionapp start --name <app-name> --resource-group <resource-group>
• Build errors: Check dotnet restore output in the sample app directory

Wrong Tracer Version After Deployment

# Check all worker initializations
grep "Assembly metadata" LogFiles/datadog/dotnet-tracer-managed-dotnet-*.log

# If old version, rebuild from the dd-trace-dotnet repo root (each build gets a unique version, no cache issues)
./tracer/tools/Build-AzureFunctionsNuget.ps1 -CopyTo <output-dir> -Verbose
# Then restore and redeploy the sample app

Traces Not Appearing in Datadog

Verify all required environment variables (including DD_API_KEY, profiler paths, etc.):

./.claude/skills/azure-functions/Test-EnvVars.ps1 -AppName "<app-name>" -ResourceGroup "<resource-group>" -IncludeRecommended

If all env vars pass, check worker initialization in logs:

grep "Datadog Tracer initialized" LogFiles/datadog/dotnet-tracer-managed-dotnet-*.log

Complete reference: See environment-variables.md for all available variables.

Separate Traces (Parenting Issue)

1. Get trace ID from host logs at execution timestamp
2. Search worker logs for same trace ID
3. If not found → worker created separate trace
4. Look for worker spans with p_id: null instead of parent IDs matching host spans
5. Enable debug logging: az functionapp config appsettings set --name <app> --resource-group <resource-group> --settings DD_TRACE_DEBUG=1
6. Re-test and analyze debug messages about AsyncLocal context flow

Query Datadog API

When traces or logs have reached Datadog (e.g., verifying spans look correct, correlating logs with a trace ID), use these scripts. Both require DD_API_KEY and DD_APPLICATION_KEY environment variables.

Retrieve all spans for a trace ID:

# Table view (default)
./tracer/tools/Get-DatadogTrace.ps1 -TraceId "<trace-id>"

# Hierarchy view — shows span parent-child tree with process tags
./tracer/tools/Get-DatadogTrace.ps1 -TraceId "<trace-id>" -OutputFormat hierarchy

# Search further back in time (default: 2h)
./tracer/tools/Get-DatadogTrace.ps1 -TraceId "<trace-id>" -TimeRange "1d"

Query logs from Datadog:

./tracer/tools/Get-DatadogLogs.ps1 -Query "service:<app-name>"
./tracer/tools/Get-DatadogLogs.ps1 -Query "service:<app-name> error" -TimeRange "2h" -Limit 100

See scripts-reference.md for full parameter reference.

Additional Resources

• Log analysis: log-analysis-guide.md - Manual log investigation patterns (when the automated -All analysis isn't sufficient)
• Scripts: scripts-reference.md - Reusable PowerShell scripts and one-liners
• Environment variables: environment-variables.md - Complete reference for Azure Functions configuration

References

• Detailed docs: docs/development/AzureFunctions.md
• Architecture deep dive: docs/development/for-ai/AzureFunctions-Architecture.md