High CPU usage at startup of the bots #1654
Comments
|
Would you mind sharing your current implementation? I suppose you're using WebSocket transport right? |
|
@ken-kuro |
|
Apologies for the ambiguity. Could you share your current implementation code? Specifically, are you using FastAPI’s WebSocket class or the lower-level WebSocket server interface? The video is a bit hard to make out, but I’m guessing it’s the FastAPI WebSocket. If that’s the case, how have you applied the bot-runner pattern? From what I understand, the pipecat architecture shines when each process handles a single pipeline—scaling via multiprocessing or across multiple VMs—rather than through asynchronous or multithreaded code. |
|
@ken-kuro sorry for the late reply, I am using lower-level WebSocket server interface: WebsocketServerTransport whenever a new request comes to the bot_runner.py, I spawn a new bot with the subprocess command and in response I provide the port in which the new bot has been spawned up. So that later I can make a socket connection to this newly spawned bot. By using the subprocess command, it enables the bot to be a separate process. |
pipecat version
0.0.63
Python version
3.11.11
Operating System
Ubuntu 20.04.6 LTS
Question
I am following the bot_runner pattern to deploy pipecat bots. While testing it for scale I am seeing a spike in cpu usage and delay in startup of bots (the delay is proportional to number of bots spawned).
I am attaching a video of the same.
The docker resource configuration is as follows:
--cpus=2 --memory=4gIs there any way I can decrease the initial CPU utilisation?
Screen.Recording.2025-04-24.at.2.46.13.PM.mp4
What I've tried
No response
Context
No response
The text was updated successfully, but these errors were encountered: