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, at an Unusual Scale
Search
Andrew Godwin
October 22, 2011
Programming
7
570
Deploying, at an Unusual Scale
A talk I gave at DjangoCon Europe 2011, about Epio's internal architecture at that point.
Andrew Godwin
October 22, 2011
Tweet
Share
More Decks by Andrew Godwin
See All by Andrew Godwin
Reconciling Everything
andrewgodwin
1
310
Django Through The Years
andrewgodwin
0
200
Writing Maintainable Software At Scale
andrewgodwin
0
430
A Newcomer's Guide To Airflow's Architecture
andrewgodwin
0
350
Async, Python, and the Future
andrewgodwin
2
650
How To Break Django: With Async
andrewgodwin
1
720
Taking Django's ORM Async
andrewgodwin
0
720
The Long Road To Asynchrony
andrewgodwin
0
650
The Scientist & The Engineer
andrewgodwin
1
750
Other Decks in Programming
See All in Programming
CSC307 Lecture 17
javiergs
PRO
0
110
iOSアプリ開発もLLMで自動運転する
hiragram
6
2.3k
事業戦略を理解してソフトウェアを設計する
masuda220
PRO
21
5.7k
インターフェース設計のコツとツボ
togishima
2
690
Beyond Portability: Live Migration for Evolving WebAssembly Workloads
chikuwait
0
280
ドメインモデリングにおける抽象の役割、tagless-finalによるDSL構築、そして型安全な最適化
knih
10
1.6k
Webからモバイルへ Vue.js × Capacitor 活用事例
naokihaba
0
450
eBPFを用いたAIネットワーク監視システム論文の実装 / eBPF Japan Meetup #4
yuukit
3
730
OpenNext + Hono on Cloudflare でイマドキWeb開発スタックを実現する
rokuosan
0
110
Zennの運営完全に理解した #完全に理解したTalk
wadayusuke
1
180
iOSアプリ開発で 関数型プログラミングを実現する The Composable Architectureの紹介
yimajo
2
190
RubyKaigi Hack Space in Tokyo & 函館最速 "予習" 会 / RubyKaigi Hack Space in Tokyo & The Fastest Briefing of RubyKaigi 2026 in Hakodate
moznion
1
130
Featured
See All Featured
Designing for humans not robots
tammielis
253
25k
How to train your dragon (web standard)
notwaldorf
92
6.1k
Typedesign – Prime Four
hannesfritz
42
2.7k
Testing 201, or: Great Expectations
jmmastey
42
7.5k
Java REST API Framework Comparison - PWX 2021
mraible
31
8.6k
Adopting Sorbet at Scale
ufuk
77
9.4k
RailsConf 2023
tenderlove
30
1.1k
The Psychology of Web Performance [Beyond Tellerrand 2023]
tammyeverts
47
2.8k
Docker and Python
trallard
44
3.4k
Raft: Consensus for Rubyists
vanstee
138
7k
JavaScript: Past, Present, and Future - NDC Porto 2020
reverentgeek
48
5.4k
Stop Working from a Prison Cell
hatefulcrawdad
269
20k
Transcript
Deploying, At An Unusual Scale Andrew Godwin http://www.flickr.com/photos/whiskeytango/1431343034/ @andrewgodwin
Hi, I'm Andrew. Serial Python developer Django core committer Co-founder
of ep.io
Hi, I'm Andrew. Serial Python developer Django core committer Co-founder
of ep.io Occasional fast talker
""Andrew speaks English like a machine gun speaks bullets."" Reinout
van Rees
We're ep.io Python Platform-as-a-Service Easy deployment, easy upgrades PostgreSQL, Redis,
Celery, and more
Why am I here? Our Architecture How we deploy Django
How varied Django deployments are
Our Architecture
Balancer Runner Runner Runner App 1 App 2 App 3
App 2 App 4 App 1 Databases File Storage Balancer
Oh My God, It's Full of Pairs Everything is redundant
Distributed programming is Hard
Hardware Real colo'd machines Linode EC2 (pretty unreliable) (pretty reliable)
(pretty reliable) IPv6 (as much as we can)
ØMQ We used to use Redis Everything now on ZeroMQ
Eliminates SPOF* * Single Point Of Failure. What a pointless acronym.
ØMQ Usage Redundant location-resolvers (Nexus) REQ/XREP for control messages PUSH/PULL
for stats, logs PUB/SUB for heartbeats, locking
Runners Unsurprisingly, these run the code SquashFS filesystem images Virtualenvs
per app UID & permission isolation, more coming
Logging/Stats All done asynchronously using ØMQ Logs to filesystem (chunked
files) Stats to PostgreSQL database, for now
Loadbalancers Intercept all incoming HTTP requests Look up hostname (or
suffix) HTTP 1.1 compliant
Databases Shared (only for PostgreSQL) Dedicated (uses Runner framework) PostgreSQL
9, damnit
Django in the backend We use the ORM extensively Annoying
settings fiddling in __init__
www.ep.io Runs on ep.io, just like any other app* Provides
JSON API, web UI * Well not quite - App ID 0 is special - but we're working on it
WSGI It's a standard, right?
WSGI It's a standard, right? Well, yes, and it works
fine, but it's not enough for serving a Python app
Static Files CSS, images, JavaScript, etc. Needs a URL and
a directory path
Python & Dependencies Mostly filled by pip/buildout/etc packaging apparently allows
version spec
Deploying Django It makes things consistent, right?
Settings Layouts Vanilla settings.py local_settings.py configs/HOSTNAME.py Many others...
Python Paths Project-level imports App-level imports apps/ directories
Databases If it's SQL, it's PostgreSQL Redis for key-value, MongoDB
soon Some things assume a safe network
HA (High Availability) Not terribly easy with shared DBs PostgreSQL
9's sensible warm standby Redis has SLAVEOF Possibly use DRBD for general solution
Backups High Availability is NOT a backup btrfs for consistent
snapshotting Archived remote syncs No access to backups from servers
Migrations No solution yet for migration/code sync We're working on
it...
Web serving It's not like it's important or anything
gunicorn Small and lightweight Supports long-running requests Pretty stable
nginx Even more lightweight Extremely fast Really, really stable
The Load Balancer Used to be HAProxy Rewritten to custom
Python daemon eventlet used for high throughput Can't use nginx, no HTTP 1.1 for backends
Celery See: Yesterday's Talk Slightly tricky to run many We
use Redis as the backend
Management Commands First off, run as subprocess Then, a custom
PTY module Now, run as pty-wrapping subprocesses
Some General Advice If you're crazy enough to do this
Messaging's Not Enough Having a state to check is handy
Why run one, when you can run two for twice
the price? Redundancy is good. Double redundancy is better.
Always expect the worst Hope you never have to deal
with it.
The more backups, the better. Make sure you have historical
ones, too.
Django is very flexible Sometimes a little too flexible...
Your real problems will emerge later Don't over-optimise up front
for everything
Questions? Andrew Godwin andrew@ep.io @andrewgodwin