Share via

Document Intelligence custom models in West US 2 stopped working around 10 PM UTC

Sergii M 0 Reputation points
2026-05-27T03:29:57.64+00:00

We are using Azure AI Document Intelligence custom models in our production solution in the West US 2 region.

Today, around 3 PM Pacific Time / 10 PM UTC, our custom model stopped responding. The requests do not fail immediately, but the analysis appears to hang indefinitely.

When testing directly in Document Intelligence Studio, the status stays on:

“Running analysis”

and never returns a result.

We have verified:

  • The service/resource is still available
  • No deployment changes were made on our side
  • The issue affects existing custom models that were previously working

Is anyone else experiencing issues with Document Intelligence custom models in West US 2? Are there any known outages or degradation events in this region?

Any guidance or confirmation from Microsoft would be appreciated.

Azure Document Intelligence in Foundry Tools
0 comments

3 answers

Sort by: Most helpful
Most helpful Newest Oldest
  1. Sergii M 0 Reputation points
    2026-05-27T05:06:15.18+00:00

    Service Health and Azure Region Status both show healthy, but Run analysis is not working in either Document Intelligence Studio or through the API.

    After waiting for some time, Document Intelligence Studio returns:

    InternalServerError

    “An unexpected error occurred. Please submit support request with apim-request-id shown below to get further assistance.”

    apim-request-id: 0f4bdde1-2a7c-4d6a-a859-286e755902a4

    Can Microsoft please check this apim-request-id and confirm whether there is an active issue with Azure AI Document Intelligence custom models in West US 2?User's image

    Was this answer helpful?

    0 comments
  2. Jerald Felix 13,335 Reputation points Volunteer Moderator
    2026-05-27T04:17:56.6266667+00:00

    Hello Sergii Maksymenko,

    Greetings! Thanks for raising this question in Q&A forum.

    I completely understand how critical this is for a production system. The behavior you're describing requests hanging indefinitely without failing, and the same happening directly in Document Intelligence Studio is a strong indicator of a service-side or regional issue rather than a problem with your custom models or code. Since no changes were made on your end and existing working models are affected, this points to a platform-level disruption in the West US 2 region.

    Here's what I recommend you do right now, step by step:

    Step 1: Check Azure Service Health immediately This is your first and most important step.

    1. Go to the Azure Portal at portal.azure.com
    2. Search for "Service Health" in the top search bar and open it
    3. Filter by your subscription, region (West US 2), and service (Azure AI Document Intelligence)
    4. Look for any active incidents, degradation events, or advisories
    5. Also check the public Azure Status page at azure.status.microsoft for any broadly communicated outages in West US 2

    Go to the Azure Status page and verify whether there is any active incident for Azure Document Intelligence in the region where the resource is deployed — this confirms or rules out a service-side outage for that region. Also review Azure Service Health in the Azure portal for subscription-specific incidents or advisories affecting Document Intelligence.

    Step 2: Verify your Document Intelligence resource's network settings While this is likely a platform issue, it's worth a quick check. Under your Document Intelligence resource's Networking blade, verify that public network access is enabled if you intend to use Document Intelligence Studio over the public internet. If firewall rules are enabled, ensure your client IP is allowed. If the resource is configured for private endpoint access only, Document Intelligence Studio may not be able to connect unless proper private network routing is in place.

    Step 3: Test the same model in a different region To quickly confirm this is a regional issue and not a model issue:

    1. Create a new temporary Document Intelligence resource in a different region like East US or East US 2
    2. Copy your custom model to that resource using Document Intelligence Studio or the REST API
    3. Run the same document through the model in the new region
    4. If it works there but hangs in West US 2, that conclusively confirms a regional platform problem

    Step 4: Copy your model to another region as an immediate mitigation While waiting for Microsoft to resolve the West US 2 issue, you can restore production by copying your custom model to a healthy region and pointing your application to the new endpoint. Model copy can be done directly from Document Intelligence Studio under your resource's model management section, or via the REST API.

    Step 5: Collect diagnostic details for support Start gathering the following information now — you'll need it for the support ticket:

    • Timestamps of when the hang started (you've already noted ~3 PM PT / 10 PM UTC)
    • Your Document Intelligence resource name, subscription ID, and resource group
    • Model name and model ID
    • Any correlation IDs or request IDs from SDK logs if available
    • Confirmation that the same documents work in another region

    Step 6: Open a support ticket right away Since this is a production issue:

    1. Go to Help + Support > New Support Request in the Azure Portal
    2. Set the Issue Type to Technical, and the Service to Azure AI Document Intelligence
    3. Describe the hanging behavior, the region (West US 2), the time it started, and that Document Intelligence Studio also reproduces the issue
    4. Attach any screenshots from Studio showing the "Running analysis" state that never completes

    If no incidents are reported on the status page but the issue persists, collect details such as model name, version, subscription ID, resource ID, timestamps, and request IDs and contact the Document Intelligence team via the documented support channels so the backend can be investigated.

    The most important thing right now is to redirect your production traffic to another region using model copy while Microsoft investigates the West US 2 issue. This will restore service for your users while the root cause is being addressed.

    If this answer helps you kindly accept the answer which will help others who have similar questions.

    Best Regards,

    Jerald Felix.

    Was this answer helpful?

    0 comments

  3. AI answer

    2026-05-27T03:30:10.8633333+00:00

    There is no documented, region-specific outage for Azure AI Document Intelligence in West US 2 in the provided information. However, the described behavior (requests hanging indefinitely in both the SDK and Document Intelligence Studio, with small PDFs and no recent configuration changes) matches a pattern that strongly suggests a service-side or regional issue rather than a problem with the custom models or client code.

    Based on similar documented scenarios and current guidance, the recommended steps are:

    1. Check official service health
      • Use the Azure status page to see if there is an active incident affecting Document Intelligence or the West US 2 region.
      • In the Azure portal, open Azure Service Health to check for subscription-specific advisories or incidents.
        These are the authoritative sources for confirmed outages or degradations.
    2. Test in another supported region
      • Document Intelligence custom neural model training is only available in specific regions, but trained models can be copied and used in any region.
      • If possible, create or use a Document Intelligence resource in a different region (for example, East US or another supported region) and test the same custom model or a similar one.
      • If the same model works in another region but continues to hang in West US 2, that strongly indicates a region-specific service issue.
    3. Use model copy as a mitigation
      • As a mitigation, copy the affected custom model from West US 2 to a different region where the service is healthy and update the application to call that region.
      • Model copy can be done via the REST API or Document Intelligence Studio as described in the disaster recovery and copy guidance.
    4. Open a support request if the issue persists
      • If Azure Status and Service Health show no incident, but requests in West US 2 continue to hang while other regions work, open an Azure support request so the engineering team can investigate the backend for that region.
      • When opening the case, include timestamps, correlation IDs (from SDK logs if available), and confirmation that the same documents and models work in another region.

    These steps align with current guidance for similar symptoms where Document Intelligence requests time out or hang and the behavior is reproducible directly in Document Intelligence Studio.

    References:

    AI-generated content may be incorrect. Read our transparency notes for more information.

    Was this answer helpful?

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.