$30 off During Our Annual Pro Sale. View Details »
Speaker Deck
Features
Speaker Deck
PRO
Sign in
Sign up for free
Search
Search
Rapid Release at Massive Scale
Search
Proko Chou
November 06, 2017
Technology
0
79
Rapid Release at Massive Scale
The study of Facebook's release architect: Rapid Release at Massive Scale
Proko Chou
November 06, 2017
Tweet
Share
More Decks by Proko Chou
See All by Proko Chou
Software Testing
prokochou
0
72
Introduce ISTQB Test
prokochou
0
74
Context Driven Testing
prokochou
0
75
Mobile App Testing
prokochou
0
60
Mining Patterns from Lyrics of Chinese Popular Music
prokochou
0
150
Other Decks in Technology
See All in Technology
Database イノベーショントークを振り返る/reinvent-2025-database-innovation-talk-recap
emiki
0
220
Haskell を武器にして挑む競技プログラミング ─ 操作的思考から意味モデル思考へ
naoya
6
1.6k
re:Invent 2025 ~何をする者であり、どこへいくのか~
tetutetu214
0
220
生成AI活用の型ハンズオン〜顧客課題起点で設計する7つのステップ
yushin_n
0
230
生成AI時代におけるグローバル戦略思考
taka_aki
0
200
ExpoのインダストリーブースでみたAWSが見せる製造業の未来
hamadakoji
0
140
GitHub Copilotを使いこなす 実例に学ぶAIコーディング活用術
74th
3
3.4k
5分で知るMicrosoft Ignite
taiponrock
PRO
0
390
ディメンショナルモデリングを支えるData Vaultについて
10xinc
1
100
シニアソフトウェアエンジニアになるためには
kworkdev
PRO
3
170
AIプラットフォームにおけるMLflowの利用について
lycorptech_jp
PRO
1
170
Reinforcement Fine-tuning 基礎〜実践まで
ch6noota
0
190
Featured
See All Featured
Practical Tips for Bootstrapping Information Extraction Pipelines
honnibal
25
1.6k
Practical Orchestrator
shlominoach
190
11k
Scaling GitHub
holman
464
140k
Visualizing Your Data: Incorporating Mongo into Loggly Infrastructure
mongodb
48
9.8k
Product Roadmaps are Hard
iamctodd
PRO
55
12k
Build The Right Thing And Hit Your Dates
maggiecrowley
38
3k
Done Done
chrislema
186
16k
The Pragmatic Product Professional
lauravandoore
37
7.1k
CoffeeScript is Beautiful & I Never Want to Write Plain JavaScript Again
sstephenson
162
16k
No one is an island. Learnings from fostering a developers community.
thoeni
21
3.6k
Code Review Best Practice
trishagee
74
19k
A Tale of Four Properties
chriscoyier
162
23k
Transcript
Rapid Release at Massive Scale 20171106 Proko
About Author
Agenda • Canary Deployment • The reasons why adopt rapid
release at massive scale • Solutions • Limitations
Canary Deployment
Canary Deployment (cont.) • Canaries were once regularly used in
coal mining as an early warning system • Distributed to a small number of users • As an early warning indicator with less impact on downtime
Canary Deployment • Deploy to one or more canary servers
• Test, or wait until satisfied • Deploy to the remaining servers
Why
Previous deployment process Three times a day Cherry-picks
Weekly Web Branch
Time Zone Issue Release duration: 14 hours every single week
London New York Menlo Park
How
Quasi-continuous “push from master” system
Tool and Process
Improve tools and process
Continuous Integration
Gatekeeper System
Test Tool Deployment (cont.)
Test Tool Deployment (cont.)
Test Tool Deployment (cont.)
Test Tool Deployment (cont.)
Test Tool Deployment
Continuous Delivery at Scale
Continuous Delivery at Scale (cont.)
Continuous Delivery at Scale: C1 (internal)
Canary Deployment Continuous Delivery at Scale: C2 (production)
Continuous Delivery at Scale: C3
Result
System
No More Hotfix
Deploy anytime, anywhere All engineers everywhere in the world can
develop and deliver their code when it makes sense for them.
Forcing Function
Make the user experience better, faster
Limitation
None
None
Android OS Version
Solution
What’s more • Use Puffin for Facebook (For Facebook) •
Add more checking point which can be executed automatically • The result can be checked by developers if it is failed
Reference 1. Article: https://code.facebook.com/posts/270314900139291/rapid-release-at-mas sive-scale/ 2. Video: https://atscaleconference.com/videos/rapid-release-at-massive-scale/ 3. Canary
Deployment: https://blog.ccjeng.com/2015/06/canary-deployment.html https://octopus.com/docs/patterns/canary-deployments
Q&A