Efficiently deploying Django

Efﬁciently deploying
Django
Dražen Lučanin

https://punkrockdev.com/

@metakermit

View Slide

So, you wrote a Django
web app…
• Followed something like the polls tutorial

• Created your own web app

• e.g. for keeping track of your family culinary recipes

• everything is running nicely on http://localhost:8000

• now how do you let your users access the app?

View Slide

… just need to deploy it
• Should be easy, right… 
• hm… What’s Apache?

View Slide

Ch. 1 Self-hosting

View Slide

OK, let's take a step back
• Core components

• web server

• domain

• database(s)

• Core actions

• start everything

• keep things running

• Security

• log errors & warnings

• encryption

• backups
Ready to jump down the rabbit hole?

View Slide

Web server
• No, we’re not gonna use Apache

• We’d go for nginx

• Now a problem is nginx (or Apache) doesn’t support Django directly

• python manage.py runserver ?

• We need to set up a production-ready application server like uWSGI
(or gunicorn) that knows how to run our Django app

• Then we need to conﬁgure nginx as a passthrough for this
server… except for static / media ﬁles… those need to be served
by nginx

View Slide

Connect the domain
• Namecheap a decent domain registrar

• A record if you have a ﬁxed IP

• CNAME if you’re forwarding to another domain

• just note that CNAME on the root domain doesn’t always
play nicely with MX rules

• check if there is ANAME / ALIAS support

• or connect your app via a subdomain like
app.mydomain.com

View Slide

Database(s)
• Postgres for your relational database

• You might need a task queue like Celery

• Redis is a good option for the backend

• Follow the tutorials and get everything set up

View Slide

Start & keep running
• So how do you start everything?

• nginx, postgres and redis will start automatically via a
systemd service

• What about uwsgi?

• You could write your custom systemd service

• You could also use something like Supervisor

View Slide

Yeah another one of those…

View Slide

Logging & monitoring
• Logging errors – https://sentry.io/

• Monitoring – https://newrelic.com/

View Slide

Security
• Oh yeah, we need to encrypt our traffic otherwise

• Let’s encrypt is a free TLS certificate issuer – decent tutorial

• but we need to set up the certbot

• update our nginx config

• set up a periodic task to auto-upgrade our certificate

View Slide

OS upgrades
• Hope you’re upgrading your OS

• Ubuntu has unattended upgrades for important security
patches – oﬀ by default

• My little Ansible script for turning this on for my servers

• https://github.com/punkrockdev/server

View Slide

Backups?

View Slide

uwsgi
nginx
namecheap letsencrypt
postgres
supervisor
systemd
ubuntu
sentry
newrelic
redis

View Slide

Summary
• We have a bunch of services

• We’re responsible for maintaining them

• They will break and change over time

• We need to keep our Django app conﬁgured to know where everything is

• Can this be automated?

• To an extent, yes – there are conﬁg management tools like Ansible 
(of Salt, Chef, Puppet)

• But there are better ways…

View Slide

Ch. 2 The twelve-
factor app

View Slide

You have a problem?
• Start throwing money at it!

• Platform-as-a-service

• Heroku the most popular example

• At $7 / month you can get a “dyno”

• a thing that knows how to run your Django app

• what’s under the hood? Someone else’s problem!

View Slide

Remember all those
services?
• Well, in Heroku they come fully integrated

• You can easily enable something like Postgres or Redis

View Slide

A whole marketplace

View Slide

Is there a catch?
• Well, there is still a fair amount of conﬁguration necessary

• Heroku integrates all the services through the twelve-
factor app methodology

• But don’t worry, you’ll see it’s a good thing!

View Slide

The twelve factors
• I. Codebase – One codebase tracked in revision control, many deploys

• II. Dependencies – Explicitly declare and isolate dependencies

• III. Config – Store config in the environment

• IV. Backing services – Treat backing services as attached resources

• V. Build, release, run – Strictly separate build and run stages

• VI. Processes – Execute the app as one or more stateless processes

• VII. Port binding – Export services via port binding

• VIII. Concurrency – Scale out via the process model

• IX. Disposability – Maximize robustness with fast startup and graceful shutdown

• X. Dev/prod parity – Keep development, staging, and production as similar as possible

• XI. Logs – Treat logs as event streams

• XII. Admin processes – Run admin/management tasks as one-off processes

View Slide

So how do we start?
• You can follow the oﬃcial Heroku tutorial

• You can also use my Yeoman generator

• https://github.com/metakermit/generator-django-rest

• Yeoman – a JavaScript-based (sssh ) code generator

View Slide

Very quick to start
npm install -g yo
npm install -g generator-django-rest
mkdir aviato
cd aviato
yo django-rest
./scripts/install.sh
./scripts/start.sh
./scripts/deploy.sh

View Slide

… and it’s live

View Slide

How does it work?

View Slide

Ch. 3 The self-hosted
twelve-factor app

View Slide

Now, Heroku can get
expensive
• Those $7 / month can easily become $14 / month if you
need periodic tasks – where you need a scheduler

• If you have more than one project running these things
add up

• If you’re not making money from those apps, that cost
could be diﬃcult to justify

• Luckily the open source world has our back!

View Slide

Dokku
• http://dokku.viewdocs.io/dokku/

• An open source Heroku alternative

• Can be self-hosted on a $5 / month Digital Ocean droplet

• Or a €3 / month Scaleway VM if you prefer something inside
the EU

• Dokku has some of the drawbacks of self-hosting, but it respects
the 12-factor methodology so it’s much easier to maintain

• Treat your servers as cattle, not as pets!

View Slide

Easy from
generator-django-rest
dokku plugin:install https://github.com/dokku/dokku-letsencrypt.git
dokku plugin:install https://github.com/dokku/dokku-postgres.git
dokku plugin:install https://github.com/dokku/dokku-redis.git
dokku apps:create aviato
dokku postgres:create aviatodb
dokku postgres:link aviatodb aviato
dokku redis:create aviatoredis
dokku redis:link aviatoredis aviato
dokku config:set --no-restart aviato \
[email protected] 
dokku letsencrypt aviato
dokku letsencrypt:auto-renew aviato
dokku letsencrypt:cron-job --add

View Slide

Aside – what about
production production?

View Slide

The “real” production
• It’s all fun and games when you only have a small number
of users (if any)

• Eventually, your app might start getting serious traﬃc

• Scaling of Django apps

• a separate topic

• lots of good talks on https://pyvideo.org/ (old PyCons)

View Slide

Some pointers
• Fully Docker-based deployments

• Amazon ECS

• Other AWS services – RDS etc.

• Hosted Kubernetes solutions

• Serverless

• Serve a Django App from AWS Lambda

• Diﬀerent architecture

• Are these the best approaches when you’re getting started

• Probably not – you’re exploring what app your users might like

• Only when you’ve found something worthwhile does it make sense to put more energy into scalability

• When you do, however, ﬁnd the right help – good DevOps engineers

View Slide

Conclusion
• Don’t go reinventing the wheel

• The twelve-factor methodology rocks!

• Heroku and Dokku rock!

• Get out and have fun working on your side-projects!

View Slide

Thanks!
Dražen Lučanin

https://punkrockdev.com/

@metakermit
Expect slides on http://metakermit.com/talks

View Slide

Efficiently deploying Django

Efficiently deploying Django

Dražen Lučanin

More Decks by Dražen Lučanin

Other Decks in Programming

Featured

Transcript