Only one CPU is being used on 4 core CPU

[Laxmi Kant]

Hi,
I am hosting a web application on 4 cores CPU. When I make concurrent requests, only one CPU is being used and other 3 CPU's are using only 3 to 4%. How can I get high parallelism and concurrency?

my docker-compose file is as follows

version: '3'
services:
  web:
    build:
      context: .
      
    volumes:
      - ./app:/app
    ports:
      - "80:80"
    environment:
      - WORKERS=16

    command: bash -c "uvicorn main:app --reload --host 0.0.0.0 --port 80"

I have tried to set WORKERS but it is not getting reflected. Only one CPU is being used regardless of WORKERS value.
How can I correctly set these values?

[Thibault Lemaire]

I think the problem is that you're overriding the command:. Take a look at the start.sh script from the base image.

[Sebastián Ramírez]

Thanks for the help here @ThibaultLemaire ! 👏 🙇

Thanks for closing the issue @laxmimerit 👍

[Sourav Kumar]

hi @ThibaultLemaire
My dockerfile is as follows, but it seems like it's not utilising even 10% of the CPU and the latency on more than 20 users is going above 2-3 seconds.
Can you please take a look on this docker image?

# Python Base Image
FROM tiangolo/uvicorn-gunicorn-fastapi:python3.10

RUN apt update
RUN apt upgrade -y
# install zbar
RUN apt install -y zbar-tools ffmpeg libsm6 libxext6 libgl1-mesa-glx

# Create the working directory


# get the requirements 
COPY ./requirements.txt /app/requirements.txt

# install dependencies
RUN pip install --no-cache-dir --upgrade -r /app/requirements.txt

# get the working app 
COPY ./server /app


# create dirs under /tmp
RUN mkdir -p /tmp/data && mkdir -p /tmp/media


# Execute the server
CMD ["uvicorn", "main:app", "--proxy-headers", "--host", "0.0.0.0", "--port", "80"]

[Thibault Lemaire]

hi @souravjamwal77

At first glance, it looks like you too are overriding the CMD. I'm sorry but I will not help you any further.

I consider it good practice on Github (and other collaborative platforms) to not tag people on issues 3 years after they have been closed.

If you think your problem is related, maintainers will usually ask that you open a new issue referencing this one. Thank you.