NGINX - Gunicorn - HTTP 502 upstream prematurely closed - docker

Question

I face a very strange behavior into a docker container.
This running was running a while until it crashed, I then mounted the docker container into another infrastructure and since it crashes.
The logs :
[INFO] Booting worker with pid: 315
"GET /oidc/authenticate/ HTTP/1.1" 302 0 "https://build.contoso.com/" "<user agent>"
"GET /xxx/yyy/ HTTP/1.1" 302 0 "-" "<user agent>"
"GET /oidc/callback/?state=state&session_state=session_state HTTP/1.1" 302 0 "-" "<user agent>"
[CRITICAL] WORKER TIMEOUT (pid:45)
[error] 306#306: *54 upstream prematurely closed connection while reading response header from upstream, client: 192.168.3.110, server: app, request: "GET /xxx/yyy/ HTTP/1.1", upstream: "http://unix:/run/gunicorn.sock:/xxx/yyy/", host: "build.contoso.com"
"GET /xxx/yyy/ HTTP/1.1" 502 559 "-" "<user agent>"
[INFO] Worker exiting (pid: 45)
My gunicorn is launched using :
$VENV/bin/gunicorn app.wsgi:application --daemon --bind unix:/run/gunicorn.sock --workers $workers --capture-output --enable-stdio-inheritance
and is up and running :
p# ps aux | grep gunicorn
root 34 0.0 0.0 30016 20832 ? S 16:41 0:01 /app/venv/bin/python /app/venv/bin/gunicorn app.wsgi:application --daemon --bind unix:/run/gunicorn.sock --workers 17 --capture-output --enable-stdio-inheritance
root 37 0.0 0.0 66068 52104 ? S 16:41 0:00 /app/venv/bin/python /app/venv/bin/gunicorn app.wsgi:application --daemon --bind unix:/run/gunicorn.sock --workers 17 --capture-output --enable-stdio-inheritance
[...]
root 413 0.0 0.0 5196 528 pts/0 S+ 21:25 0:00 grep gunicorn
My nginx configuration:
# cat /etc/nginx/sites-enabled/gunicorn
upstream gunicorn {
server unix:/run/gunicorn.sock;
}
server {
listen 80;
server_name app;
location /static/ {
alias /app/static/;
}
location / {
include proxy_params;
proxy_redirect off;
proxy_pass http://gunicorn;
}
}
Still it crashes, any help welcomed.