Microsoft Foundry: High Peak Access Error During Peak Hours

Hello @唐新惠

Thank you for reaching out to Microsoft Q&A.

We understand that you are experiencing request failures or delays when using Microsoft Foundry, particularly during peak hours, and that you are seeing the following error:

“ERR_BAD_REQUEST – Your resource has been temporarily blocked because we detected unusual behavior.”

This message indicates that the resource has been temporarily blocked by the service after unusual activity was detected. However, based on the information currently available, we cannot confirm that peak-hour traffic or quota usage is the specific cause of the block.

High request volumes, rapid request bursts, excessive retries, or other request patterns can contribute to service-side protection mechanisms. Therefore, we recommend checking the request pattern around the time the block occurs.

Recommended steps

Stop rapid or continuous retry attempts

• If automatic retries are configured, ensure that they use an appropriate delay and exponential backoff.
• Avoid continuously retrying requests while the resource is blocked.

Temporarily reduce request rate and concurrency

• Reduce the number of parallel requests during peak periods.
• Avoid sudden traffic spikes where possible.

Allow the temporary block to clear

• Since the message indicates a temporary resource block, avoid repeatedly sending requests while the block is active.

Review quota and usage

• Check the applicable quota/usage metrics for the resource and deployment in Azure Portal.
• Determine whether request volume or throughput increases significantly when the issue occurs.

Review failed requests immediately before the block

• Check whether there are repeated 400, 429, or other failed requests before the resource becomes blocked.
• Review whether an application retry loop may be amplifying the request volume.

Validate the request configuration

• Confirm that the correct endpoint and API version are being used.
• Verify the required authentication and Content-Type headers.
• Confirm that the request body matches the API schema and does not contain invalid or unsupported parameters.
• Verify that the model deployment is active and available.

If the issue occurs consistently during peak hours, please compare the request rate, concurrency, retry behavior, and error responses immediately before and during the block. This will help determine whether the behavior is associated with request volume/throttling or requires further service-side investigation.

Thank you!