ag2ai
[Bug]: The single-machine multi-process jupytor server always indicates that there are no available ports
Describe the bug
I’m using pyautogen: 0.2.26.
I started 10 shell sessions in my system, with each shell running a separate program for parallel inference. I deployed 5 different tool models on ports ranging from 8080 to 8085, which the agent uses.
However, I keep encountering issues like “connection reset by peer” and “no available port for Jupyter client.”
These issues don’t always happen at the start of the program — sometimes they occur after many inference tasks have already been completed. I tried assigning a specific port to each process’s Jupyter server, but it didn’t help much. After processing dozens of data samples, the agents suddenly freeze, and all agents stop running, although the program itself does not exit.
Steps to reproduce
No response
Model Used
Owen-VL-Max
Expected Behavior
No response
Screenshots and logs
No response
Additional Information
No response
@zhangzef: Thanks for reporting this issue!
To help us understand better, could you share a minimal reproducible example or a bit more about how you're setting up the Jupyter servers and ports? That would really help us investigate the problem.
Appreciate your support!
@zhangzef Let us know if you still need help with this. I just emailed. Please feel free to schedule a call.
