debug-crash-dump

作者: microsoft

從 CI 失敗中下載並分析崩潰轉儲,使用 CDB/WinDbg。當被要求除錯崩潰、分析轉儲檔案、設定 WinDbg/CDB 時,使用此技能。

npx skills add https://github.com/microsoft/ebpf-for-windows --skill debug-crash-dump

Crash Dump Debugging

Download and analyze crash dumps from CI workflow failures (especially fault_injection_full).

See docs/CrashDumpDebugging.md for the full human-readable guide.

When to Use

  • User asks to debug a CI crash or test failure with a crash dump
  • User asks to download crash dump artifacts from a workflow run
  • User asks to set up CDB, WinDbg, or mcp-windbg
  • User asks to analyze a .dmp file
  • User asks about fault injection leak detection failures

Agent Prerequisites

  1. Windows Debugging Tools (CDB/WinDbg) — Install the Debugging Tools feature from the Windows SDK:

    # Download the SDK online installer
    Invoke-WebRequest -Uri "https://go.microsoft.com/fwlink/?linkid=2272610" -OutFile winsdksetup.exe
    # Install only the debugging tools (elevated)
    Start-Process -FilePath .\winsdksetup.exe -ArgumentList "/features","OptionId.WindowsDesktopDebuggers","/quiet","/norestart" -Verb RunAs -Wait
    

    CDB installs to: C:\Program Files (x86)\Windows Kits\10\Debuggers\x64\cdb.exe

  2. mcp-windbg (MCP server for AI-assisted dump analysis):

    # Install Python 3.12 via NuGet (if not already available)
    nuget install python -Version 3.12.8 -OutputDirectory C:\Users\$env:USERNAME\tools
    $pyExe = (Get-ChildItem "C:\Users\$env:USERNAME\tools\python.3.12.8" -Recurse -Filter "python.exe" | Select-Object -First 1).FullName
    & $pyExe -m ensurepip --upgrade
    & $pyExe -m pip install mcp-windbg
    

Downloading Artifacts

# Download crash dumps and build artifacts (PDBs) from a CI run
gh run download <run_id> -R <owner>/ebpf-for-windows -n "Crash-Dumps-<test>-x64-<config>" -D crash-dumps
gh run download <run_id> -R <owner>/ebpf-for-windows -n "Build-x64-<config>" -D build
# Extract inner build zip for PDBs (produces build\<config>\<config>\ with PDBs)
Expand-Archive build\build-<config>.zip -DestinationPath build\<config>

Analyzing Dumps with CDB

$cdbPath = "C:\Program Files (x86)\Windows Kits\10\Debuggers\x64\cdb.exe"
$symPath = "build\<config>\<config>;SRV*C:\Symbols*https://msdl.microsoft.com/download/symbols"

# Quick triage: exception record + stack trace
& $cdbPath -z crash-dumps\unit_tests.exe.XXXX.dmp -y $symPath -lines -c ".ecxr;kP;q"

# Full automated analysis
& $cdbPath -z crash-dumps\unit_tests.exe.XXXX.dmp -y $symPath -c "!analyze -v;q"

Running mcp-windbg Server

$pyExe = (Get-ChildItem "C:\Users\$env:USERNAME\tools\python.3.12.8" -Recurse -Filter "python.exe" | Select-Object -First 1).FullName
$cdbPath = "C:\Program Files (x86)\Windows Kits\10\Debuggers\x64\cdb.exe"
$symPath = "build\<config>\<config>;SRV*C:\Symbols*https://msdl.microsoft.com/download/symbols"

& $pyExe -m mcp_windbg --cdb-path $cdbPath --symbols-path $symPath --transport streamable-http --port 8765
# MCP endpoint: http://127.0.0.1:8765/mcp

Common Patterns

  • Fault injection crashes: Usually assert _allocations.empty() in leak_detector.cpp — a memory leak detected during teardown. Check the leak detector's _in_memory_log for the allocation call stack:
    $$ Get exception context and inspect the stack to find the leak_detector frame:
    .ecxr
    kP
    $$ Identify the frame index where 'this' is the leak_detector instance
    .frame <N>
    dx this->_in_memory_log
    
  • Crash dump artifacts: Named Crash-Dumps-<test_name>-<platform>-<config> (e.g., Crash-Dumps-fault_injection_full-x64-Debug)
  • Build artifacts with PDBs: Named <build_artifact>-<config> (for the default build_artifact=Build-x64, this is Build-x64-Debug), and contain an inner zip build-<config>.zip

來自 microsoft 的更多技能

oss-growth
microsoft
開源增長駭客角色
agent-framework-azure-ai-py
microsoft
使用Microsoft Agent Framework Python SDK(agent-framework-azure-ai)构建Azure AI Foundry代理。适用于使用AzureAIAgentsProvider创建持久化代理、使用托管工具(代码解释器、文件搜索、网络搜索)、集成MCP服务器、管理对话线程或实现流式响应。涵盖函数工具、结构化输出和多工具代理。
development
airunway-aks-setup
microsoft
在AKS上設定AI Runway——從裸叢集到執行模型。涵蓋叢集驗證、控制器安裝、GPU評估、供應商設定及首次部署。時機:「設定AI Runway」、「上線AKS叢集」、「安裝AI Runway」、「airunway設定」、「部署模型至AKS」、「在AKS上進行GPU推論」、「在AKS上設定KAITO」、「在AKS上執行LLM」、「在AKS上使用vLLM」、「在AKS上設定模型服務」、「AI Runway控制器」。
devops
appinsights-instrumentation
microsoft
使用Azure Application Insights檢測Web應用程式的指南。提供遙測模式、SDK設定與組態參考。適用時機:如何檢測應用程式、App Insights SDK、遙測模式、什麼是App Insights、Application Insights指南、檢測範例、APM最佳實踐。
devops
applicationinsights-web-ts
microsoft
使用Application Insights JavaScript SDK(@microsoft/applicationinsights-web)為瀏覽器/Web應用程式進行檢測。適用於真實使用者監控(RUM)——頁面檢視、點擊、AJAX/fetch依賴、例外、自訂事件,以及與後端OpenTelemetry追蹤關聯的瀏覽器端GenAI代理追蹤。涵蓋SDK載入器指令碼與npm設定、框架擴充(React、React Native、Angular)、點擊分析、遙測初始化器,以及從瀏覽器發出的代理/工具/模型span的OTel GenAI語意慣例。
devops
azure-ai-anomalydetector-java
microsoft
使用適用於 Java 的 Azure AI 異常偵測器 SDK 建置異常偵測應用程式。在實作單變量/多變量異常偵測、時間序列分析或 AI 驅動監控時使用。
development
azure-ai-language-conversations-py
microsoft
使用 azure-ai-language-conversations Python SDK 實作對話語言理解(CLU)。當使用 ConversationAnalysisClient 分析對話意圖與實體、建置 NLP 功能,或將語言理解整合至應用程式時使用。
development
azure-ai-ml-py
microsoft
Azure Machine Learning SDK v2 for Python。用於機器學習工作區、作業、模型、資料集、計算資源與管線。 觸發詞:「azure-ai-ml」、「MLClient」、「workspace」、「model registry」、「training jobs」、「datasets」。
development