Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Features
Speaker Deck
PRO
Sign in
Sign up for free
Search
Search
Deploying Flask (WSGI) Applications
Search
Matt Wright
June 26, 2014
Technology
3
1.6k
Deploying Flask (WSGI) Applications
Supporting files:
https://github.com/mattupstate/flask-nyc-deploying
Matt Wright
June 26, 2014
Tweet
Share
More Decks by Matt Wright
See All by Matt Wright
Python Apps and Docker
mattupstate
1
1.2k
Testing Flask Applications
mattupstate
5
1.5k
Other Decks in Technology
See All in Technology
Alpine.js を活用した Laravel MPA フロントエンド最適化戦略 / Alpine.js MPA
tzmfreedom
1
680
モンテカルロ木探索のパフォーマンスを予測する Kaggleコンペ解説 〜生成AIによる未知のゲーム生成〜
rist
3
820
コミュニティとおかねの話 / Community and Money
kgsi
0
100
DIってなんだか難しい? 依存という概念を「使う・使われる」 という言葉で整理しよう
akinoriakatsuka
1
730
モジュラーモノリスでスケーラブルなシステムを作る - BASE のリアーキテクチャのいま
panda_program
7
1.7k
Go の analysis パッケージで自作するリファクタリングツール
kworkdev
PRO
1
340
Redefine_Possible
upsider_tech
0
160
非エンジニアにも伝えるメールセキュリティ / Email security for non-engineers
ykanoh
13
3.6k
VISITS_20250311_こねくとあいとりおす.pdf
iotcomjpadmin
0
220
移行できそうでやりきれなかった 10年超えのシステムを葬るための戦略 / phper-kaigi-2025-ryu
carta_engineering
0
670
大規模プロジェクトにおける 品質管理の要点と実践 / 20250327 Suguru Ishii
shift_evolve
0
190
DevinはクラウドエンジニアAIになれるのか!? 実践的なガードレール設計/devin-can-become-a-cloud-engineer-ai-practical-guardrail-design
tomoki10
2
890
Featured
See All Featured
Intergalactic Javascript Robots from Outer Space
tanoku
270
27k
The World Runs on Bad Software
bkeepers
PRO
67
11k
CoffeeScript is Beautiful & I Never Want to Write Plain JavaScript Again
sstephenson
160
15k
Optimising Largest Contentful Paint
csswizardry
34
3.1k
Scaling GitHub
holman
459
140k
Facilitating Awesome Meetings
lara
53
6.3k
For a Future-Friendly Web
brad_frost
176
9.6k
Large-scale JavaScript Application Architecture
addyosmani
511
110k
The Web Performance Landscape in 2024 [PerfNow 2024]
tammyeverts
4
490
Art, The Web, and Tiny UX
lynnandtonic
298
20k
[Rails World 2023 - Day 1 Closing Keynote] - The Magic of Rails
eileencodes
33
2.1k
Performance Is Good for Brains [We Love Speed 2024]
tammyeverts
8
690
Transcript
Deploying Flask (WSGI) Applications
Matt Wright @mattupstate Engineer @ChatID • Python, Ruby, CoffeeScript •
Flask, Chef, AngularJS Open Source • Flask-Security • Flask-Social • Flask-Mail + a few others
WSGI (Web Server Gateway Interface)
PEP 333 http://legacy.python.org/dev/peps/pep-0333/
# ~/hello.py def application(environ, start_response): start_response('200 OK', [('Content-Type','text/html')]) return ["Hello
World"]
# ~/app.py from flask import Flask app = Flask(__name__) @app.route('/')
@app.route('/<path:path>') def catch_all(path=''): return 'Hello World'
Browser WSGI Gateway WSGI App 1 2 4 3 Request
Flow
Browser WSGI Gateway WSGI App 1 1. Browser sends an
HTTP request to the gateway server Request Flow
Browser WSGI Gateway WSGI App 2 2. Gateway server prepares
and sends the request to the WSGI compatible application 1 Request Flow
Browser WSGI Gateway WSGI App 1 2 3 3. Application
processes the request and returns a response Request Flow
Browser WSGI Gateway WSGI App 1 2 4 3 4.
The gateway server returns the response to the browser Request Flow
Browser WSGI Gateway WSGI Middle- ware WSGI App Middleware
Browser WSGI Gateway WSGI Middle- ware WSGI App WSGI Middle-
ware WSGI Middle- ware WSGI Middle- ware WSGI Middle- ware MOAR MIDDLEWARE!
Gateway Servers (sometimes called “containers”)
http://www.tornadoweb.org/ Tornado
1. Written in Python! 2. Web application framework 3. Asynchronous
networking library 4. WSGI support Tornado
# ~/tor.py from tornado.wsgi import WSGIContainer from tornado.httpserver import HTTPServer
from tornado.ioloop import IOLoop from app import app http_server = HTTPServer(WSGIContainer(app)) http_server.listen(8000) IOLoop.instance().start()
matt @ pc in ~/ $ mkvirtualenv wsgi matt @
pc in ~/ workon:wsgi $ pip install flask tornado matt @ pc in ~/ workon:wsgi $ python tor.py
matt @ pc in ~/ $ curl localhost:8000 Hello World
matt @ pc in ~/ workon:wsgi $ pip install tornado matt @ pc in ~/ workon:wsgi $ python tor.py
http://www.gevent.org/ gevent
1. Written in Python! 2. Coroutine networking library 3. 3rd
party module support gevent
# ~/ge.py from gevent.wsgi import WSGIServer from app import app
server = WSGIServer(('', 8000), app) server.serve_forever()
matt @ pc in ~/ workon:wsgi $ pip install gevent
matt @ pc in ~/ workon:wsgi $ python ge.py
matt @ pc in ~/ $ curl localhost:8000 Hello World
matt @ pc in ~/ workon:wsgi $ pip install gevent matt @ pc in ~/ workon:wsgi $ python ge.py 127.0.0.1 - - [2014-06-24...
http://gunicorn.org/ Gunicorn
Gunicorn 1. Written in Python! 2. Easy configuration 3. Broadly
compatible 4. Extensible 5. Fairly speedy
matt @ pc in ~/ workon:wsgi $ pip install gunicorn
matt @ pc in ~/ workon:wsgi $ gunicorn app:app
matt @ pc in ~/ $ curl localhost:8000 Hello World
matt @ pc in ~/ $ mkvirtualenv wsgi matt @ pc in ~/ workon:wsgi $ pip install gunicorn flask matt @ pc in ~/ workon:wsgi $ gunicorn app:app
https://uwsgi-docs.readthedocs.org/ uWSGI
uWSGI 1. Support for many platforms and languages 2. Multiple
configuration formats 3. Flexible logging 4. Extensible 5. Speaks HTTP and uwsgi
matt @ pc in ~/ workon:wsgi $ pip install uwsgi
matt @ pc in ~/ workon:wsgi $ uwsgi --http :8000 -w app:app ... spawned uWSGI worker 1 ...
matt @ pc in ~/ $ curl localhost:8000 Hello World
matt @ pc in ~/ workon:wsgi $ pip install uwsgi matt @ pc in ~/ workon:wsgi $ uwsgi --http :8000 -w app:app ... spawned uWSGI worker 1 ... [pid: 42176|app: 0|req: ...
(how to start your app) Process Management
http://upstart.ubuntu.com/ upstart
# /etc/init/hello-world.conf description "Hello World app" start on runlevel [2345]
stop on runlevel [06] chdir /srv/hello-world exec /usr/local/bin/uwsgi \ --die-on-term \ --http :8000 \ -w app:app
http://www.freedesktop.org/wiki/Software/systemd/ systemd
# /etc/systemd/system/hello-world.service [Unit] Description=Hello World app [Service] ExecStart=/usr/local/bin/uwsgi \ --die-on-term
\ --http :8000 \ -w app:app [Install] WantedBy=multi-user.target
http://supervisord.org/ supervisord
# /etc/supervisor.d/hello-world.conf [program:hello-world] directory=/srv/hello-world command=/usr/local/bin/uwsgi \ --die-on-term \ --http :8000
\ -w app:app
(some things are better than others) Reverse Proxies
Why? • Static files • URL rewrites • Routing •
Header mods • App aggregation • and many more! • Network security • Authentication • Load balancing • SSL termination • Compression • Caching
http://www.nginx.org/ Apache
# /etc/apache2/sites-enabled/hello-world.conf <VirtualHost *:80> ServerName hello-world.com ProxyPreserveHost On ProxyPass /static
! Alias /static "/srv/hello-world/static" ProxyPass / http://127.0.0.1:8000/ ProxyPassReverse / http://127.0.0.1:8000/ </VirtualHost>
http://www.nginx.org/ Nginx
# /etc/nginx/sites-enabled/hello-world.conf server { listen 80; server_name hello-world.com; location /
{ proxy_pass http://127.0.0.1:8000; } location ^~ /static/ { root /srv/hello-world/static; } }
(is hard) Concurrency
Concurrency Models callback/generator (auto) lightweight threads (auto) processor/thread (config) processor/thread
(config) Tornado gevent Gunicorn uWSGI
Increasing Concurrency processes processes workers + processes workers + processes
Tornado gevent Gunicorn uWSGI
Increasing Workers Gunicorn uWSGI --workers <integer> --workers <integer>
DISCLAIMER!
(when one server isn’t enough) Load Balancing
Machine Level App Server 192.168.1.101 App Server 192.168.1.102 App Server
192.168.1.102 Primary Load Balancer 192.168.1.100
# /etc/nginx/sites-enabled/load-balancer.conf upstream app_upstream { least_conn; server 192.168.1.101:8000 server 192.168.1.102:8000
server 192.168.1.103:8000 } server { listen 80; server_name hello-world.com; location / { proxy_pass http://app_upstream; } }
Machine + Process Level App Server 192.168.1.101 App Server 192.168.1.102
App Server 192.168.1.102 Primary Load Balancer 192.168.1.100 t 127.0.0.1:8001 127.0.0.1:8002 127.0.0.1:8000 127.0.0.1:8000 127.0.0.1:8001 127.0.0.1:8002 127.0.0.1:8000 127.0.0.1:8001 127.0.0.1:8002
# /etc/supervisor.d/hello-world.conf [program:hello-world] numprocs=3 process_name=hello-world%(process_num) directory=/srv/hello-world command=/usr/local/bin/uwsgi \ --die-on-term \
--http :80%(process_num)02d \ -w app:app
# /etc/nginx/sites-enabled/hello-world.conf upstream local_upstream { least_conn; server 127.0.0.1:8000 server 127.0.0.1:8001
server 127.0.0.1:8002 } server { listen 80; server_name hello-world.com; location / { proxy_pass http://local_upstream; } }
(I’ve failed at this a lot) Recommendations
1. Use virtualenv and pip 2. Don’t use virtualenvwrapper 3.
Use a `requirements.txt` file Python
1. Provide default settings 2. Enable settings to be overridden
a. Environment Variable(s) b. Default file locations 3. Raise an error at startup when required settings are missing 4. Generate a `SECRET_KEY` with `os. urandom(16)` Configuration
1. Prefer the uwsgi protocol over HTTP 2. Prefer TCP
sockets over Unix sockets 3. Lots of useful options a. virtualenv + pythonpath b. enable-threads + lazy-apps c. need-app d. stats Use uWSGI
1. Pass custom headers 2. Lots of useful modules a.
ngx_http_geoip_module b. ngx_http_gzip_module c. ngx_http_memcached_module d. ngx_http_status_module Use Nginx
(putting it all together) Automation
http://www.ansible.com/ Ansible
(because what could go wrong?) Demo
Thank You! mattupstate.com gittip.com/mattupstate github.com/mattupstate twitter.com/mattupstate