azure-troubleshoot

majiayu000 的头像作者 majiayu000

使用工具优先访问对 Azure 进行故障排除,必要时回退到 Azure CLI。重点关注虚拟机、AKS、Azure 容器注册表、存储帐户和日志分析。

📖 中文导读:使用工具优先访问对 Azure 进行故障排除,必要时回退到 Azure CLI。重点关注虚拟机、AKS、Azure 容器注册表、存储帐户和日志分析。 以下正文为该技能的原始 SKILL.md 内容(英文),上方安装命令可直接使用。

name: azure-troubleshoot description: "Troubleshoot Azure using tool-first access, falling back to Azure CLI when necessary. Focus on Virtual Machines, AKS, Azure Container Registry, Storage Accounts, and Log Analytics."

Azure Troubleshooting Skill

General Guidance

Always use tool-based queries first to fetch logs, metrics, and diagnostic data.
Only fall back to Azure CLI for deeper or unsupported inspection.

Investigations should:

  1. Use Log Analytics Kusto queries with proper scoping
  2. Use Activity Logs to identify failures
  3. Use metrics when diagnosing performance issues
  4. Provide minimal, targeted remediation advice

Core Services Covered

Virtual Machines

Common issues:

  • Boot failures
  • OS/disk failures
  • NIC/IP misconfiguration

Investigations:

  • Inspect boot diagnostics logs
  • Query Heartbeat table for VM status
  • Check Activity Logs for failed start/stop operations

AKSS

Common issues:

  • Pod scheduling failures
  • Node pressure
  • Image pull errors (ACR auth)
  • Container crashes

Investigations:

  • Query KubeEvents
  • Query KubePodInventory
  • Inspect ContainerLog

Azure Container Registry (ACR))

Common issues:

  • Permission denied (RBAC)
  • Token expiration

Investigations:

  • Query Activity Logs for push/write or pull/read failures
  • Check repository event logs

Storage Accountss

Common issues:

  • Firewall-restricted access
  • SAS token expiration
  • Object not found

Investigations:

  • Query StorageBlobLogs
  • Validate configuration + permissions

Log Analytics

Best practices:

  • Always filter by _ResourceId
  • Narrow time range
  • Query only the tables relevant to the service

Workflow

  1. Identify target service
  2. Query Log Analytics with scoped KQL
  3. Query Activity Logs
  4. Review metrics
  5. Interpret patterns
  6. Recommend targeted fixes