Debugging a Flask app running in Gunicorn

Question:

I’ve been working on a new dev platform using nginx/gunicorn and Flask for my application.

Ops-wise, everything works fine – the issue I’m having is with debugging the Flask layer. When there’s an error in my code, I just get a straight 500 error returned to the browser and nothing shows up on the console or in my logs.

I’ve tried many different configs/options.. I guess I must be missing something obvious.

My gunicorn.conf:

import os

bind = '127.0.0.1:8002'
workers = 3
backlog = 2048
worker_class = "sync"
debug = True
proc_name = 'gunicorn.proc'
pidfile = '/tmp/gunicorn.pid'
logfile = '/var/log/gunicorn/debug.log'
loglevel = 'debug'

An example of some Flask code that borks- testserver.py:

from flask import Flask
from flask import render_template_string
from werkzeug.contrib.fixers import ProxyFix

app = Flask(__name__)

@app.route('/')
def index():
    n = 1/0
    return "DIV/0 worked!"

And finally, the command to run the flask app in gunicorn:

gunicorn -c gunicorn.conf.py testserver:app

Thanks y’all

Asked By: mafrosis

||

Answers:

The Flask config is entirely separate from gunicorn’s. Following the Flask documentation on config files, a good solution would be change my source to this:

app = Flask(__name__)
app.config.from_pyfile('config.py')

And in config.py:

DEBUG = True
Answered By: mafrosis

Try setting the debug flag on the run command like so

gunicorn -c gunicorn.conf.py --debug testserver:app

and keep the DEBUG = True in your Flask application. There must be a reason why your debug option is not being applied from the config file but for now the above note should get you going.

Answered By: Fuchida

The accepted solution doesn’t work for me.

Gunicorn is a pre-forking environment and apparently the Flask debugger doesn’t work in a forking environment.

Attention

Even though the interactive debugger does not work in
forking environments (which makes it nearly impossible to use on
production servers) […]

Even if you set app.debug = True, you will still only get an empty page with the message Internal Server Error if you run with gunicorn testserver:app. The best you can do with gunicorn is to run it with gunicorn --debug testserver:app. That gives you the trace in addition to the Internal Server Error message. However, this is just the same text trace that you see in the terminal and not the Flask debugger.

Adding the if __name__ ... section to the testserver.py and running python testserver.py to start the server in development gets you the Flask debugger. In other words, don’t use gunicorn in development if you want the Flask debugger.

app = Flask(__name__)
app.config['DEBUG'] = True

if __name__ == '__main__':
    app.run()

## Tip for Heroku users:
Personally I still like to use `foreman start`, instead of `python testserver.py` since [it sets up all the env variables for me](https://devcenter.heroku.com/articles/config-vars#using-foreman). To get this to work:

Contents of Procfile

web: bin/web

Contents of bin/web, file is relative to project root

#!/bin/sh

if [ "$FLASK_ENV" == "development" ]; then
        python app.py
else
        gunicorn app:app -w 3
fi

In development, create a .env file relative to project root with the following contents (docs here)

FLASK_ENV=development
DEBUG=True

Also, don’t forget to change the app.config['DEBUG']... line in testserver.py to something that won’t run Flask in debug mode in production.

app.config['DEBUG'] = os.environ.get('DEBUG', False)
Answered By: Nick Zalutskiy

For Heroku users, there is a simpler solution than creating a bin/web script like suggested by Nick.

Instead of foreman start, just use foreman run python app.py if you want to debug your application in development.

Answered By: aristidesfl

I had similiar problem when running flask under gunicorn I didn’t see stacktraces in browser (had to look at logs every time). Setting DEBUG, FLASK_DEBUG, or anything mentioned on this page didn’t work. Finally I did this:

app = Flask(__name__)
app.config.from_object(settings_map[environment])
if environment == 'development':
    from werkzeug.debug import DebuggedApplication
    app_runtime = DebuggedApplication(app, evalex=False)
else:
    app_runtime = app

Note evalex is disabled because interactive debbugging won’t work with forking (gunicorn).

Answered By: jazgot

I used this:

gunicorn "swagger_server.__main__:app" -w 4 -b 0.0.0.0:8080
Answered By: Steven McConnon

You cannot really run it with gunicorn and for example use the flask reload option upon code changes.

I’ve used following snippets in my api launchpoint:

app = Flask(__name__)

try:
    if os.environ["yourapp_environment"] == "local":
        run_as_local = True
        # some other local configs e.g. paths
        app.logger.info('Running server in local development mode!')
except KeyError as err:
    if "yourapp_environment" in err.args:
        run_as_local = False
        # some other production configs e.g. paths
        app.logger.info('No "yourapp_environment env" given so app running server in production mode!')
    else:
        raise

...
...
...

if __name__ == '__main__':
    if run_as_local:
        app.run(host='127.0.0.1', port='8058', debug=True)
    else:
        app.run(host='0.0.0.0')

For above solution you need to give export yourapp_environment = "local" in the console.

now I can run my local as python api.py and prod gunicorn --bind 0.0.0.0:8058 api:app

The else statement app.run() is not actually needed, but I keep it for reminding me about host, port etc.

Answered By: eemilk
Categories: questions Tags: , ,
Answers are sorted by their score. The answer accepted by the question owner as the best is marked with
at the top-right corner.