Internal error when calling Azure Deepseek API

Question

Internal error when calling Azure Deepseek API

Xiaoyi Zhang 10 Microsoft Employee

Hi,

We encountered a lot 500 errors when using Azure AI service with deepseek-v3-0324 today.

region: east us

x-ms-client-request-id: c71ca0c3-5ef2-467d-b5d9-198481bb59a3

x-ms-client-request-id: 6480870e-1f97-11f0-984a-3cecef749f50

x-ms-client-request-id: aff9fabf-b88a-4e51-a5d3-b15e29e431bd

Azar 29,520 Reputation points MVP Volunteer Moderator

2025-04-22T17:38:19.26+00:00
Hi there Xiaoyi Zhang

Thanks for using QandA platform

This usually indicate an issue on Azure's end rather than with your request.

maybe try this

Retry with exponential backoff, try with a diff region.

Check Azure status for any outages in your region: https://status.azure.com

try switching to a different model or region temporarily.

Also, if you haven’t already, raise a support ticket with Microsoft including the x-ms-client-request-id values you shared—they'll help the team trace the exact failure.

If this helps kindly accept the answer thanks much.
Ravada Shivaprasad 535 Reputation points Microsoft External Staff Moderator

2025-04-23T09:52:08.1733333+00:00

Hi Xiaoyi Zhang

We encountered 500 errors while using Azure AI service with DeepSeek-V3-0324 in the East US region, indicating server-side issues rather than request errors. To diagnose it first we need to check the Azure Status Page for any service disruptions. While awaiting resolution, implementing exponential backoff for retries can help minimize unnecessary load, allowing the system to recover efficiently. Additionally, using circuit breakers can prevent cascading failures when errors persist.

If the issue appears region-specific, temporarily switching to a different Azure region can ensure service continuity. However, monitoring the Azure Status Page is crucial to track updates. If the errors persist, submit a support request via the Azure portal, including the x-ms-client-request-id values received (c71ca0c3-5ef2-467d-b5d9-198481bb59a3, 6480870e-1f97-11f0-984a-3cecef749f50, and aff9fabf-b88a-4e51-a5d3-b15e29e431bd). The DeepSeek-V3-0324 model is a Mixture-of-Experts model, and providing multiple client request IDs can help Azure support diagnose the issue effectively.

References: Retry Guidence , Circuit-breaker Pattern , Azure Open Ai Service Overview, Support Request Creation , Azure Global Infrastructure

Thanks
Gilbert Leung 5 Reputation points

2025-04-23T10:02:43.44+00:00

We've been encountering timeouts for our R1 modelon Azure inthe last 18 hours .

No outages reported as far as we can tell.
Xiaoyi Zhang 10 Reputation points Microsoft Employee

2025-04-23T16:11:01.2766667+00:00

hey Gilbert Leung, yes the internal error also became timeout for us too! But the playground still works well, it's really weird... Thanks for letting me know you have the same issue. Keep you updated.
Gilbert Leung 5 Reputation points

2025-04-24T00:59:18.14+00:00

We filed a ticket , but they said "there was no outage."

Let's keep this thread alive in hopes of getting more attention to this matter.
Ravada Shivaprasad 535 Reputation points Microsoft External Staff Moderator

2025-04-25T16:30:01.75+00:00

Hi Xiaoyi Zhang

I attempted to reproduce the scenario in my environment, and it’s working as expected on my end.

Thanks
Ravada Shivaprasad 535 Reputation points Microsoft External Staff Moderator

2025-04-28T04:44:13.3033333+00:00

Hi Xiaoyi Zhang

Did you get any chance to check the response.

Thank you!
Xiaoyi Zhang 10 Reputation points Microsoft Employee

2025-04-28T05:03:21.0333333+00:00

hey Ravada Shivaprasad and Gilbert Leung, sorry for missing the message. After the timeout day, our error got back to "internal error". Then though debugging we find the key factor is the "tools", once I sent the "tools" param in the request payload, it will return internal error. If not sending the "tools", it will work fine.
Ravada Shivaprasad 535 Reputation points Microsoft External Staff Moderator

2025-04-28T05:07:33.78+00:00

Hi Xiaoyi Zhang

Glad to know your issue has been resolved. Since the Microsoft Q&A community has a policy that "The question author cannot accept their own answer. They can only accept answers by others "I'll repost your solution in case you'd like to accept the answer.
Thank You
Xiaoyi Zhang 10 Reputation points Microsoft Employee

2025-04-28T05:10:32.0866667+00:00

Thanks Ravada Shivaprasad for the reposting!
Ravada Shivaprasad 535 Reputation points Microsoft External Staff Moderator

2025-04-28T05:14:21.5733333+00:00

Hi Xiaoyi Zhang

I'm glad that you were able to resolve your issue and thank you for posting your solution so that others experiencing the same thing can easily reference this!

Since the Microsoft Q&A community has a policy that "The question author cannot accept their own answer. They can only accept answers by others", I'll repost your solution.

Please click "Accept" the answer as original posters help the community find answers faster by identifying the correct answer.

Issue : Internal error when calling Azure Deepseek API

Resolution : After the timeout day, our error got back to "internal error". Then though debugging we find the key factor is the "tools", once I sent the "tools" param in the request payload, it will return internal error. If not sending the "tools", it will work fine.

Please don’t forget to close the thread by clicking "Accept the answer" wherever the information provided helps you, as this can be beneficial to other community members.

Accept answers on Microsoft Q&A | Microsoft Learn

An accepted answer is the answer that the person who asked the question chooses as the one they think best solves their problem.

Accept answers on Microsoft Q&A | Microsoft Learn

An accepted answer is the answer that the person who asked the question chooses as the one they think best solves their problem.

Thanks

Your answer

Azar 29,520 Reputation points MVP Volunteer Moderator

2025-04-22T17:38:19.26+00:00

Hi there Xiaoyi Zhang

Thanks for using QandA platform

This usually indicate an issue on Azure's end rather than with your request.

maybe try this

Retry with exponential backoff, try with a diff region.

Check Azure status for any outages in your region: https://status.azure.com

try switching to a different model or region temporarily.

Also, if you haven’t already, raise a support ticket with Microsoft including the x-ms-client-request-id values you shared—they'll help the team trace the exact failure.

If this helps kindly accept the answer thanks much.
Ravada Shivaprasad 535 Reputation points Microsoft External Staff Moderator

2025-04-23T09:52:08.1733333+00:00

Hi Xiaoyi Zhang

We encountered 500 errors while using Azure AI service with DeepSeek-V3-0324 in the East US region, indicating server-side issues rather than request errors. To diagnose it first we need to check the Azure Status Page for any service disruptions. While awaiting resolution, implementing exponential backoff for retries can help minimize unnecessary load, allowing the system to recover efficiently. Additionally, using circuit breakers can prevent cascading failures when errors persist.

If the issue appears region-specific, temporarily switching to a different Azure region can ensure service continuity. However, monitoring the Azure Status Page is crucial to track updates. If the errors persist, submit a support request via the Azure portal, including the x-ms-client-request-id values received (c71ca0c3-5ef2-467d-b5d9-198481bb59a3, 6480870e-1f97-11f0-984a-3cecef749f50, and aff9fabf-b88a-4e51-a5d3-b15e29e431bd). The DeepSeek-V3-0324 model is a Mixture-of-Experts model, and providing multiple client request IDs can help Azure support diagnose the issue effectively.

References: Retry Guidence , Circuit-breaker Pattern , Azure Open Ai Service Overview, Support Request Creation , Azure Global Infrastructure

Thanks
Gilbert Leung 5 Reputation points

2025-04-23T10:02:43.44+00:00

We've been encountering timeouts for our R1 modelon Azure inthe last 18 hours .

No outages reported as far as we can tell.
Xiaoyi Zhang 10 Reputation points Microsoft Employee

2025-04-23T16:11:01.2766667+00:00

hey Gilbert Leung, yes the internal error also became timeout for us too! But the playground still works well, it's really weird... Thanks for letting me know you have the same issue. Keep you updated.
Gilbert Leung 5 Reputation points

2025-04-24T00:59:18.14+00:00

We filed a ticket , but they said "there was no outage."

Let's keep this thread alive in hopes of getting more attention to this matter.
Ravada Shivaprasad 535 Reputation points Microsoft External Staff Moderator

2025-04-25T16:30:01.75+00:00

Hi Xiaoyi Zhang

I attempted to reproduce the scenario in my environment, and it’s working as expected on my end.

Thanks
Ravada Shivaprasad 535 Reputation points Microsoft External Staff Moderator

2025-04-28T04:44:13.3033333+00:00

Hi Xiaoyi Zhang

Did you get any chance to check the response.

Thank you!
Xiaoyi Zhang 10 Reputation points Microsoft Employee

2025-04-28T05:03:21.0333333+00:00

hey Ravada Shivaprasad and Gilbert Leung, sorry for missing the message. After the timeout day, our error got back to "internal error". Then though debugging we find the key factor is the "tools", once I sent the "tools" param in the request payload, it will return internal error. If not sending the "tools", it will work fine.
Ravada Shivaprasad 535 Reputation points Microsoft External Staff Moderator

2025-04-28T05:07:33.78+00:00

Hi Xiaoyi Zhang

Glad to know your issue has been resolved. Since the Microsoft Q&A community has a policy that "The question author cannot accept their own answer. They can only accept answers by others "I'll repost your solution in case you'd like to accept the answer.
Thank You
Xiaoyi Zhang 10 Reputation points Microsoft Employee

2025-04-28T05:10:32.0866667+00:00

Thanks Ravada Shivaprasad for the reposting!
Ravada Shivaprasad 535 Reputation points Microsoft External Staff Moderator

2025-04-28T05:14:21.5733333+00:00

Hi Xiaoyi Zhang

I'm glad that you were able to resolve your issue and thank you for posting your solution so that others experiencing the same thing can easily reference this!

Since the Microsoft Q&A community has a policy that "The question author cannot accept their own answer. They can only accept answers by others", I'll repost your solution.

Please click "Accept" the answer as original posters help the community find answers faster by identifying the correct answer.

Issue : Internal error when calling Azure Deepseek API

Resolution : After the timeout day, our error got back to "internal error". Then though debugging we find the key factor is the "tools", once I sent the "tools" param in the request payload, it will return internal error. If not sending the "tools", it will work fine.

Please don’t forget to close the thread by clicking "Accept the answer" wherever the information provided helps you, as this can be beneficial to other community members.

Accept answers on Microsoft Q&A | Microsoft Learn

An accepted answer is the answer that the person who asked the question chooses as the one they think best solves their problem.

Accept answers on Microsoft Q&A | Microsoft Learn

An accepted answer is the answer that the person who asked the question chooses as the one they think best solves their problem.

Thanks

Share via

Internal error when calling Azure Deepseek API

Your answer