.Net: Bug: Semantic Kernel Agent Orchestration Time Out

[Cynthia611]

Describe the bug
I'm using Semantic Kernel Agent 1.57.0 preview library to build agentic flow. I wanted to kindly follow up on the timeout issue we’ve been encountering recently with SK:

1. We observe that timeouts occur after the app service has been running for some time. Restarting the service seems to temporarily resolve the issue, which suggests it might be related to state or caching.
2. OpenAI calls are not experiencing timeouts, so the issue appears to be specific to SK.
3. Could this be related to the preview environment?
4. Would you be able to advise on how we can access logs or error traces within SK to help with further troubleshooting?

To Reproduce
Steps to reproduce the behavior:

1. Library: Semantic Kernel Agent 1.57.0 preview

2. Call var agentOrchestration= new SequentialOrchestration<KernelArguments, string>(
   agents.XXX, agents. XXX)
   {
   ResponseCallback = (chatContent) => CaptureAgentResponse(XXX)
   };
   var result = await agentOrchestration.InvokeAsync(args, runtime, cancellationToken);

3. It worked well for local run, or sometime after app service started.

4. However, after running a while from app service, it started to face time out. No Open AI calls errors.

5. See error
   System.TimeoutException: Orchestration did not complete within the allowed duration (00:02:30).
   at Microsoft.SemanticKernel.Agents.Orchestration.OrchestrationResult1.GetValueAsync(Nullable1 timeout, CancellationToken cancellationToken)

System.TimeoutException:
at Microsoft.SemanticKernel.Agents.Orchestration.OrchestrationResult1+<GetValueAsync>d__11.MoveNext (Microsoft.SemanticKernel.Agents.Orchestration, Version=1.57.0.0, Culture=neutral, PublicKeyToken=f300afd708cefcd3) at System.Runtime.ExceptionServices.ExceptionDispatchInfo.Throw (System.Private.CoreLib, Version=6.0.0.0, Culture=neutral, PublicKeyToken=7cec85d7bea7798e) at System.Runtime.CompilerServices.TaskAwaiter.ThrowForNonSuccess (System.Private.CoreLib, Version=6.0.0.0, Culture=neutral, PublicKeyToken=7cec85d7bea7798e) at System.Runtime.CompilerServices.TaskAwaiter.HandleNonSuccessAndDebuggerNotification (System.Private.CoreLib, Version=6.0.0.0, Culture=neutral, PublicKeyToken=7cec85d7bea7798e) at System.Threading.Tasks.ValueTask1.get_Result (System.Private.CoreLib, Version=6.0.0.0, Culture=neutral, PublicKeyToken=7cec85d7bea7798e)

Expected behavior
Restarting the app service can temporarily fix this. However, would like to understand how to make it more reliable? Is it due to preview version? And how to access SK logs to get more troubleshooting?

Screenshots

Platform

• Language: C#
• Source: NuGet package version 1.57.0 preview
• AI model: AzureOpenAI:GPT-4o
• IDE: Visual Studio
• OS: Windows

Additional context
Add any other context about the problem here.

[TaoChenOSU]

Hi!

You mentioned that it worked well locally but unstable when you deployed your app. Can you confirm that it works locally 100% of the time? And could you let us know the timeout you set?

[Cynthia611]

Hi!

You mentioned that it worked well locally but unstable when you deployed your app. Can you confirm that it works locally 100% of the time? And could you let us know the timeout you set?

Yeah, when testing locally, not seeing that issue. Time out we set is 180s.

[TaoChenOSU]

Could you provide some code that we can use to reproduce the issue?

[Cynthia611]

Hi, I can share you the repo. Since it's Microsoft internal, do you mind if we can sync up offline?

[TaoChenOSU]

Yes, off course

[TaoChenOSU]

The customer agreed to find a reproduction locally. Closing for now.

[Ross-Tarrant]

I'm also seeing this issue, while using version 1.65. Was there any resolution that can be shared?