Allow to disable access logs about uvicorn-gunicorn-docker

[NixBiks]

@victornoel did you find a solution to this? I am getting info logs from uvicorn.access although I have LOG_LEVEL set to warning in gunicorn.

[NixBiks]

+1

[victornoel]

@mr-bjerre I ended up configuring a ini config file with this inside (I omitted the other things in it):

    [logger_gunicorn]
    level=INFO
    handlers=console
    propagate=0
    qualname=gunicorn

    [logger_gunicorn_access]
    level=WARNING
    handlers=console
    propagate=0
    qualname=gunicorn.access

And I pass it to gunicorn via --log-config (it means I run gunicorn by hand). I suppose #21 would be helpflu there.

[tiangolo]

You can now set (or disable) the access log with: https://github.com/tiangolo/uvicorn-gunicorn-docker#access_log 🎉

And you can also pass Gunicorn configs in an env var: https://github.com/tiangolo/uvicorn-gunicorn-docker#gunicorn_cmd_args

[joshzana]

It looks like this works when running gunicorn in start.sh, but is ignored in start-reload.sh. Would be great to get that script to run exec uvicorn --reload --no-access-log when ACCESS_LOG is unset.

[tiangolo]

@joshzana start-reload.sh is exclusively for local development, just as a simple, small, shortcut. You shouldn't use it for production. It will only use a single process and will use a lot more resources (memory, etc) watching files, etc.

But anyway, if for some reason you still need custom logic during development, you can also write a small script using start-reload.sh as template, and running that.

[github-actions]

Assuming the original issue was solved, it will be automatically closed now. But feel free to add more comments or create new issues.

[cjmcgraw]

This problem persists for me.

I am running a container. Viewing the stdout/stderr of the container.

I currently have the following gunicorn_conf.py set in my /app/gunicorn_conf.py:

# Gunicorn config variables
loglevel = "info"
workers = 2
threads = 8
bind = "0.0.0.0:80"

preload_app = False
statsd_host = "statsd:8125"
statsd_prefix = "recsys"
timeout = 180
keepalive = 120
accesslog = None
errorlog = "-"

While changing the accesslog parameter to stdout ("-") has gunicorn log. Uvicorn still is logging no matter what I do:

INFO:uvicorn.access:127.0.0.1:36678 - "GET /health/test HTTP/1.1" 200
INFO:uvicorn.access:172.17.0.1:46740 - "GET / HTTP/1.1" 200
INFO:uvicorn.access:172.17.0.1:46744 - "GET / HTTP/1.1" 200
INFO:uvicorn.access:172.17.0.1:46748 - "GET / HTTP/1.1" 200
INFO:uvicorn.access:172.17.0.1:46752 - "GET / HTTP/1.1" 200
INFO:uvicorn.access:172.17.0.1:46756 - "GET / HTTP/1.1" 200
INFO:uvicorn.access:172.17.0.1:46760 - "GET / HTTP/1.1" 200
INFO:uvicorn.access:172.17.0.1:46764 - "GET / HTTP/1.1" 200
INFO:uvicorn.access:172.17.0.1:46768 - "GET / HTTP/1.1" 200
INFO:uvicorn.access:172.17.0.1:46772 - "GET / HTTP/1.1" 200
INFO:uvicorn.access:172.17.0.1:46776 - "GET / HTTP/1.1" 200
INFO:uvicorn.access:172.17.0.1:46780 - "GET / HTTP/1.1" 200
INFO:uvicorn.access:172.17.0.1:46784 - "GET / HTTP/1.1" 200
INFO:uvicorn.access:172.17.0.1:46788 - "GET / HTTP/1.1" 200

Using the ACCESS_LOG parameter (or manually setting it in my gunicorn_conf.py) seems to have absolutely no effect on that log printing out.

I have yet to follow the advice of @victornoel and change out the logging configuration. I am wondering if we have a more direct and clear solution to this problem.

[cjmcgraw]

Following up on my previous post. If I clear out uvicorns access logger and turn off its propagation it seems to work.

import logging as log
app = ...
...

log.getLogger('uvicorn.access').handlers = []
log.getLogger('uvicorn.access').propagate = False

[tiangolo]

If anyone is still having issues, please try with a recent version of Uvicorn, a previous version was overriding the root logger.