Azure IaaS VM垂直自动缩放：除自动化账户外的实现方式问询

Great question! Vertical auto-scaling for Azure IaaS VMs isn't limited to Automation Accounts and Runbooks—there are a couple of practical alternatives that fit well with your daily 9:00-13:00 high CPU use case. Here are the top options:

1. Azure Monitor Alerts + Logic Apps #

This is a low-code approach that leverages Azure's native monitoring and workflow tools:

• Setup steps:
  • Create an Azure Monitor alert rule targeting your VM's CPU usage, or use a schedule-based alert (perfect for your fixed time window) to trigger at 8:45 AM (to scale up before peak load) and 13:15 PM (to scale down after).
  • Link the alert to a Logic App, which will execute the resize action. In the Logic App, use the built-in "Azure Virtual Machines" connector to call the Change VM Size operation.
• Pros: No need to write complex scripts—use a visual designer to build your workflow. You can easily add extra checks (e.g., verify VM status before resizing) or notifications (e.g., send an email when scaling happens).

2. Azure Virtual Machine Scale Sets (VMSS) with Vertical Scaling Policies #

If you can migrate your VM to a VMSS, this is the most native Azure solution for vertical scaling:

• Setup steps:
  • Create a VMSS (or convert your existing VM to one) with your current VM size.
  • In the VMSS scaling settings, select Vertical scaling instead of horizontal.
  • Add either:
    • A schedule-based rule: Configure it to scale up to a larger VM size at 9:00 AM, and scale down to a smaller size at 13:00 PM.
    • A metric-based rule: Trigger scaling when CPU usage exceeds your threshold (e.g., 70% for 10 minutes) during peak hours.
• Important note: Vertical scaling for VMSS requires instances to restart during the resize process. Plan this for just before your peak window to minimize downtime, or use multiple instances with rolling updates if availability is critical.

3. Azure Monitor Alerts + Azure Functions #

For scenarios where you need custom logic (e.g., conditional scaling based on multiple metrics), Azure Functions are a flexible choice:

• Setup steps:
  • Create an Azure Function (using Python, .NET, or your preferred language) that uses the Azure SDK to call the VM resize API.
  • Link an Azure Monitor alert (schedule or metric-based) to trigger the Function when scaling is needed.
• Pros: Full control over scaling logic—you can add checks like verifying current VM size, ensuring no maintenance is in progress, or combining CPU metrics with memory usage to make smarter scaling decisions.

Key Considerations for All Approaches #

• VM Restarts: Almost all vertical scaling operations for Azure IaaS VMs require a restart. Coordinate scaling times to avoid disrupting peak user activity (e.g., scale up 15-30 minutes before 9 AM).
• Permissions: Ensure the service principal or managed identity used for scaling has the Virtual Machine Contributor role (or a custom role with Microsoft.Compute/virtualMachines/write permission) on your VM/VMSS.
• Testing: Always validate scaling workflows in a non-production environment first to catch issues like incompatible VM sizes or permission errors.

内容的提问来源于stack exchange，提问作者Puneet