Flask Cache not caching

Question

I followed a tutorial of Flask-Cache and tried to implement it myself. Given the following example, why would Flask not cache the time?

from flask import Flask
import time

app = Flask(__name__)
cache = Cache(config={'CACHE_TYPE': 'simple'})
cache.init_app(app)

@app.route('/time')
@cache.cached(timeout=50, key_prefix='test')
def test():
   return time.ctime()

Output is always the current time.

It seems like the cache is recreated every single request. What am I doing wrong?

Edit: I execute the following python-file with Python 2.7.6:

def runserver():
    port = int(os.environ.get('PORT', 5000))
    Triangle(app)
    app.run(host='0.0.0.0', port=port, processes=5)


if __name__ == '__main__':
    runserver()

Scarper

I can't reproduce this; when running with Flask with the bundled app.run() development server, the time is cached. How are you running this? – Boutte

Note that SimpleCache is a global dictionary. If you are using a WSGI server that uses multiprocessing to scale out, you'll get separate copies and they'll appear not to cache. – Boutte

Thanks Martijn, I am running it locally and not deploying it to any wsgi server. I added my runscript – Scarper

Answer 1

You are using the SimpleCache setup:

cache = Cache(config={'CACHE_TYPE': 'simple'})

This uses a single global dictionary to hold the cache, and this in turn will only work if you are using a WSGI server that uses one Python interpreter to serve all your WSGI requests. If you use a WSGI server that uses separate child processes to handle requests, you'll get a new copy of that dictionary each time and nothing is cached, effectively.

The code works fine when run with the built-in app.run() development server, given you don't use the processes argument.

Your update shows that you run the server with 5 separate processes. Each process will get its own dictionary, and the cache is not shared between them. Use a different caching backend instead, like filesystem:

cache = Cache(config={'CACHE_TYPE': 'filesystem', 'CACHE_DIR': '/tmp'})

Boutte

You are right! What do I have to change if I want to keep the processes argument? – Scarper

Use a different backend, one that can be shared between processes. The filesystem is one that doesn't require additional dependencies. – Boutte

Thank you so much for the perfect answer! I wouldn't have figured it out by myself! – Scarper

for WSGI if I am telling processes from .ini file will app.run(host='0.0.0.0', port=port, processes=5) change the case ? also cache_dir can be visible ?? – Bandicoot

@BeqaBukhradze: I'm not sure what you are asking, really. It doesn't matter how you are configuring your WSGI server, if you use separate processes you are using separate processes. – Boutte

is it better configure it from flask or from uwsgi – Bandicoot

@BeqaBukhradze: those things would be configured in the WSGI server. If you are using uwsgi as the server, then configure the host, port and concurrency settings there. Flask comes with a built-in WSGI server to ease development, but this server should never be used in production (it is not designed to handle real-world loads). – Boutte

@MartijnPieters link Here when calling app.run() can i call app.run(threaded=true) or it is not necessary ? – Bandicoot

@BeqaBukhradze: when using uWSGI, that part doesn't run. The if __name__ == '__main__': test is not true in that case. app.run() runs the Flask built-in WSGI server, not uWSGI. – Boutte