Share via

Getting a ton of Server Errors

Nitin Kumar 0 Reputation points
2026-03-31T21:05:48.57+00:00

I am seeing a lot of Server Errors (5xx). It looks like that it started at 3/30/3036 4 AM for my Document Intelligence resources. There are more than 13.1 K errors in past 10 hours..

resource 1

User's image

resource 2

User's image

Can you please advice on this?..

Azure Document Intelligence in Foundry Tools

1 answer

Sort by: Most helpful
  1. Manas Mohanty 16,190 Reputation points Microsoft External Staff Moderator
    2026-04-14T07:38:24.9466667+00:00

    Hi Nitin Kumar

    A lot of 5xx points towards quota exhaustion, frequent rate limit, unhealthy resources etc.

    Quoting them all below for troubleshooting.

    1. Check service health

    2. Review your resource’s health

    • In the Azure portal, go to your Document Intelligence resource → Resource health.
      • Look for any recent incidents or outages in that blade.

    4. Verify quotas and throttling

    • In the portal’s Usage + quotas page for your resource, confirm you haven’t exceeded any request rate or throughput limits.
      • If you are close to your quota, consider adding retry logic with exponential back-off or reducing request concurrency and

    Please check the file size /request limit here . https://dotnet.territoriali.olinfo.it/en-us/azure/foundry/agents/concepts/limits-quotas-regions#limit-error-reference

    5. Consider transient/back-off logic

    • Many server-side errors are transient. Implement a retry policy (3 – 5 retries with exponential back-off) to smooth out temporary spikes.

    6. Check for CMK-related issues (if applicable)

    • If you’re using customer-managed keys (CMK) for encryption, stale key caches can cause ServiceUnavailable errors. You can refresh the CMK cache or switch to Microsoft-managed keys temporarily.
    1. Opt for Batch request/PTU (Provisioned throughput unit- for dedicated SLA compared to Pay as you go quota)

    8. Load balance with multi- region endpoint to reduc

    9. Reduce total size of your payload

    If you’ve gone through these steps and still see a high volume of 5xx errors, please let us know:

    • Which region(s) your resources are in
    • Whether you’re using any custom models or customer-managed keys
    • Rough request volume or pattern (e.g., bursts vs. steady stream)
    • Exact error codes you see in the responses (500 vs. 503, etc.)

    Hope this helps stabilize things!

    Reference list

    Please help fill the details in support ticket link shared in private message if the issue persists even loading reducing payload and balancing with multi-region endpoints.

    Thank you.


Your answer

Answers can be marked as 'Accepted' by the question author and 'Recommended' by moderators, which helps users know the answer solved the author's problem.