CODEWITHSUNDEEP
pythonapacheflaskmod-wsgipython-multiprocessing
amone
amone

Reputation: 3862

Multi-process Flask Application on Apache

I am trying to create a simple web application with Python3/Flask and serve it on Apache. I could not figure out how can I make my application to respond multiple requests.

This is my wsgi file:

import sys
import os
sys.path.insert(0, '/var/www/html/FlaskDeploy')

from apps import app as application

This code excerpt from httpd.conf file:

<VirtualHost *:80>
        DocumentRoot /var/www/html/FlaskDeploy
        WSGIScriptAlias / /var/www/html/FlaskDeploy/app.wsgi
        WSGIDaemonProcess apps threads=1 python-path=/var/www/html/FlaskDeploy/env/bin:/var/www/html/FlaskDeploy/env/lib/python3.6/site-packages

        <Directory /var/www/html/FlaskDeploy>
        WSGIScriptReloading On
        WSGIProcessGroup apps
        WSGIApplicationGroup %{GLOBAL}
        Order deny,allow
        Allow from all
        </Directory>
</VirtualHost>

Everything works fine but the application runs the requests one by one. For example, assume that each user performs a heavy database operation that takes 3 minutes. In this case, when 3 users from different locations open the application at the same time, the last one has to wait for 9 minutes (including the others' operations to be completed).

Basically I want to create a web application that is able to handle multiple requests.

I am coming from NodeJS world and I have never encountered with this problem on NodeJS. It runs on a single thread but can handle multiple requests.

Upvotes: 0

Views: 1951

Answers (1)

Graham Dumpleton
Graham Dumpleton

Reputation: 58563

It is only capable of only handling one request at a time, because that is what you told mod_wsgi to do in using:

threads=1

Don't set that option and it will instead default to 15 threads in the daemon process group and so that is how many requests it can handle concurrently.

If your requests are I/O bound that should be fine to begin with and you can tune things later. If your requests are more CPU bound than I/O bound, start to introduce additional processes as well and distribute requests across them.

processes=3 threads=5

Even if heavily I/O bound, do not increase threads too far per process, it is better to still spread them across processes as Python doesn't work as well with high number of threads per process.

For more information read the documentation:

Upvotes: 2

Related Questions