Azure AI Foundry stuck at running

Question

Azure AI Foundry stuck at running

Sangmin Lee 0

I have a project with several prompt flows. Sometimes they work. It definitely worked fine in the beginning. After a while it's stuck at running and even if I leave it running for around an hour, it's still running in the LLM process.

When it works, it doesn't take more than 30 seconds.

Each flow is one LLM and one Python script to return the output in JSON format.

I've tried creating new projects and it's the same result.

What can I do?

Ravada Shivaprasad 40 Reputation points Microsoft External Staff

2025-05-02T13:48:54.7666667+00:00

Hi Sangmin Lee

If your Azure AI Foundry projects are stuck in the running stage and taking longer than expected, several factors could be contributing to the issue. One common cause is concurrency limits, where old pipeline runs may still be in progress, preventing new runs from executing smoothly. To check this, go to the Azure AI Foundry authoring canvas, select your pipeline, and verify if a concurrency property is assigned. Ensure no runs from the past 45 days are still active—if they are, cancel them to free up resources.
Another potential issue involves transient failures such as network disruptions, credential errors, or service outages. In such cases, retrying the pipeline runs can be effective since Azure Data Factory has an internal recovery process that monitors and attempts to restart failed runs automatically. If your pipeline remains stuck beyond an hour, consider opening a support case.

Optimizing data flow activities is another important troubleshooting step, especially if ADF Copy and Data Flow activities experience long start-up times. You may want to implement the Time to Live feature for Data Flow activities and review Microsoft’s guidance on Copy Performance Improvement and Data Flow Integration Runtime Optimization (Troubleshoot deploy and monitor). Effective resource management and monitoring can also prevent delays. This includes setting up timeouts, tracking resource usage, and implementing cleanup procedures after each pipeline run. Additionally, optimizing prompt flows is useful in identifying potential issues with complexity or input formats, particularly if performance has degraded over time. If needed, Microsoft provides troubleshooting steps for prompt flow (Prompt Flow Troubleshooting).

If you've followed these troubleshooting steps but are still facing issues, please let me know.

Thanks
Pavankumar Purilla 7,090 Reputation points Microsoft External Staff Moderator

2025-05-05T22:27:02.0533333+00:00

Hi Sangmin Lee,
Did you get any chance to check the response. Thank you!
SriLakshmi C 4,875 Reputation points Microsoft External Staff Moderator

2025-05-07T16:04:51.1433333+00:00

Hi @Sangmin Lee,

We haven’t heard from you on the last response and was just checking back to see if you have a resolution yet. In case if you have any resolution please do share that same with the community as it can be helpful to others. Otherwise, will respond with more details and we will try to help.

Thank you!

1 answer

Your answer

Ravada Shivaprasad 40 Reputation points Microsoft External Staff

2025-05-02T13:48:54.7666667+00:00

Hi Sangmin Lee

If your Azure AI Foundry projects are stuck in the running stage and taking longer than expected, several factors could be contributing to the issue. One common cause is concurrency limits, where old pipeline runs may still be in progress, preventing new runs from executing smoothly. To check this, go to the Azure AI Foundry authoring canvas, select your pipeline, and verify if a concurrency property is assigned. Ensure no runs from the past 45 days are still active—if they are, cancel them to free up resources.
Another potential issue involves transient failures such as network disruptions, credential errors, or service outages. In such cases, retrying the pipeline runs can be effective since Azure Data Factory has an internal recovery process that monitors and attempts to restart failed runs automatically. If your pipeline remains stuck beyond an hour, consider opening a support case.

Optimizing data flow activities is another important troubleshooting step, especially if ADF Copy and Data Flow activities experience long start-up times. You may want to implement the Time to Live feature for Data Flow activities and review Microsoft’s guidance on Copy Performance Improvement and Data Flow Integration Runtime Optimization (Troubleshoot deploy and monitor). Effective resource management and monitoring can also prevent delays. This includes setting up timeouts, tracking resource usage, and implementing cleanup procedures after each pipeline run. Additionally, optimizing prompt flows is useful in identifying potential issues with complexity or input formats, particularly if performance has degraded over time. If needed, Microsoft provides troubleshooting steps for prompt flow (Prompt Flow Troubleshooting).

If you've followed these troubleshooting steps but are still facing issues, please let me know.

Thanks
Pavankumar Purilla 7,090 Reputation points Microsoft External Staff Moderator

2025-05-05T22:27:02.0533333+00:00

Hi Sangmin Lee,
Did you get any chance to check the response. Thank you!
SriLakshmi C 4,875 Reputation points Microsoft External Staff Moderator

2025-05-07T16:04:51.1433333+00:00

Hi @Sangmin Lee,

We haven’t heard from you on the last response and was just checking back to see if you have a resolution yet. In case if you have any resolution please do share that same with the community as it can be helpful to others. Otherwise, will respond with more details and we will try to help.

Thank you!

Answer 1

Hello Sangmin Lee,

Thank you for posting your question in the Microsoft Q&A forum.

Azure AI Foundry’s prompt flows occasionally encountering prolonged execution times stuck indefinitely in the "Running" state suggests an underlying systemic issue rather than an isolated code problem. Since these flows consist of just one LLM call and a Python script, and since they previously worked reliably, the degradation points to potential throttling, resource allocation constraints, or backend service instability.

First, check Azure’s status dashboard for regional outages or degraded performance in AI services like OpenAI or Azure Machine Learning. Microsoft occasionally imposes rate limits on LLM APIs, which could cause queuing delays. If the status page shows no issues, review your application logs in Azure Monitor to identify failed or throttled requests.

Next, validate your Python script’s error handling. If the LLM response is malformed or times out, the script might hang while waiting for a resolution. Implement explicit timeouts and retry logic to prevent indefinite stalls. Additionally, test with simplified flows remove the Python step temporarily to isolate whether delays originate from the LLM or the script.

If the above answer helped, please do not forget to "Accept Answer" as this may help other community members to refer the info if facing a similar issue. Your contribution to the Microsoft Q&A community is highly appreciated.

Deleted

This comment has been deleted due to a violation of our Code of Conduct. The comment was manually reported or identified through automated detection before action was taken. Please refer to our Code of Conduct for more information.

Share via

Azure AI Foundry stuck at running

1 answer

Your answer