How to change number of workers and threads in running process for gunicorn

I want to test performance of WEB service that running inside AWS ECS service depending on number of gunicorn workers.

Entrypoint of the container is:

WORKERS=15
THREADS=15

gunicorn \
  --reload \
  --workers "${WORKERS}" \
  --threads "${THREADS}" \
  --max-requests 10000 \
  --max-requests-jitter 200 \
  --timeout 60 \
  --access-logfile - \
  --error-logfile - \
  --bind 0.0.0.0:8000 \
  config.wsgi:application

The problem:

If I want to change number of workers / threads I have to stop gunicorn → update ECS Task definition (set new number of WORKERS and THREADS) → restart ECS container. It takes too much time if I want to test tens of configurations.

Possible workaround:

It is possible to set mock endless entrypoint like watch -n 1000 "ls -l" and login to ECS container with bash and run gunicorn with desired parameters manually. But it is little bit inconvenient and suppose to create test specific environment. So, I want to avoid this method.

The question:

Is it possible to change number of workers and threads of already running gunicorn instance? To be able test different configurations without rerunning container and stopping its entrypoint process.

Solution

You could use config file and reload the config by sending HUP signal to the gunicorn master process. See: reload configuration

Here is a simple example:

# Dockerfile
FROM python:3.9-slim

COPY . /app
RUN pip install --no-cache-dir -r /app/requirements.txt
WORKDIR /app
ENTRYPOINT ["gunicorn",  "app:app"]

# tree ./
.
├── app.py
├── Dockerfile
├── gunicorn.conf.py
└── requirements.txt

# cat gunicorn.conf.py
workers = 2
threads = 2
loglevel = 'debug'
errorlog = '-'
accesslog = '-'
pidfile = '/var/run/app.pid'

# reload the configuration by
kill -HUP $(cat /var/run/app.pid)