关于Microsoft Graph API返回401而非429的异常处理咨询

Let me break this down based on my experience working with Graph API and throttling scenarios:

First, let's call out the unusual response you're seeing—this definitely doesn't align with expected behavior. Here's the error you shared, formatted for clarity:

{
  "error": {
    "code": "serviceNotAvailable",
    "message": "Service unavailable",
    "retryAfterSeconds": 120
  }
}

Why this is odd #

Normally, when Graph API throttles your requests, it returns a 429 Too Many Requests status code, paired with a Retry-After header (or the retryAfterSeconds field in the response body). A 401 Unauthorized should only be for authentication failures (expired tokens, invalid permissions, etc.), so seeing serviceNotAvailable tied to a 401 is a clear mismatch.

How to handle this in your app #

You'll want to implement targeted logic for this specific edge case:

• Don't treat all 401s as throttling: Keep your existing 401 handling for actual authentication issues (like refreshing expired tokens).
• Add a special check for this error: When you get a 401, parse the response body. If it contains the serviceNotAvailable error code and a retryAfterSeconds value, handle it exactly like a 429:
  • Respect the retryAfterSeconds value for the initial delay
  • Use exponential backoff for subsequent retries (e.g., 120s → 240s → 480s, up to a reasonable maximum like 10 minutes)
• Log everything: Capture timestamps, request URLs, full response details, and retry behavior. This will help if you need to report the issue to Microsoft or debug if the problem persists.

Is this a Microsoft bug/regression? #

Yes, this almost certainly looks like a regression or service-side error in Graph API's error handling. The fact that it's returning the wrong HTTP status code for a throttling/service availability issue is not intended behavior.

To confirm and get resolution:

• Check the Microsoft 365 Admin Center Service Health dashboard to see if there's a known outage or incident affecting Graph API, SharePoint, or OneDrive.
• Search Microsoft's developer forums for similar reports—other developers might be hitting the same issue.
• If no existing reports exist, submit a support ticket or feedback via the Graph Developer portal, including your logged error details and context about your backup workload.

内容的提问来源于stack exchange，提问作者Stepan Riha