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
Distributed Failure: Learning Lessons From Avia...
Search
Andrew Godwin
April 24, 2018
Programming
2
430
Distributed Failure: Learning Lessons From Aviation
A talk I first gave at Code Europe Warsaw, spring 2018.
Andrew Godwin
April 24, 2018
Tweet
Share
More Decks by Andrew Godwin
See All by Andrew Godwin
Reconciling Everything
andrewgodwin
1
230
Django Through The Years
andrewgodwin
0
130
Writing Maintainable Software At Scale
andrewgodwin
0
370
A Newcomer's Guide To Airflow's Architecture
andrewgodwin
0
280
Async, Python, and the Future
andrewgodwin
2
570
How To Break Django: With Async
andrewgodwin
1
620
Taking Django's ORM Async
andrewgodwin
0
640
The Long Road To Asynchrony
andrewgodwin
0
560
The Scientist & The Engineer
andrewgodwin
1
650
Other Decks in Programming
See All in Programming
はじめてみよう量子プログラミング
itokoichi01
0
210
Swiftコードバトル必勝法
toshi0383
0
170
Amazon BedrockでサーバレスなAIお料理ボットを作成する!!
tosuri13
0
230
Rechartsで楽にゴリゴリにカスタマイズする!
10tera
1
170
Regular Expressions, REXML, Automata Learning
makenowjust
0
220
開発を加速する共有Swift Package実践
elmetal
PRO
0
420
GraphQLの魅力を引き出すAndroidクライアント実装
morux2
3
660
Increased Performance and Developer Productivity with Jakarta EE 11
ivargrimstad
0
360
Go Code Generation at newmo / 2024-08-27 #newmo_layerx_go
genkey6
0
560
Prolog入門
qnighy
4
1k
Lessons by WebAssembly app in production on CDN Edge Computing Service
tetsuharuohzeki
0
210
Architecture Decision Record (ADR)
nearme_tech
PRO
1
690
Featured
See All Featured
Bash Introduction
62gerente
608
210k
[RailsConf 2023] Rails as a piece of cake
palkan
48
4.6k
Art, The Web, and Tiny UX
lynnandtonic
294
20k
Unsuck your backbone
ammeep
667
57k
ReactJS: Keep Simple. Everything can be a component!
pedronauck
663
120k
Understanding Cognitive Biases in Performance Measurement
bluesmoon
26
1.3k
Statistics for Hackers
jakevdp
794
220k
Visualization
eitanlees
142
15k
What’s in a name? Adding method to the madness
productmarketing
PRO
21
3k
Practical Tips for Bootstrapping Information Extraction Pipelines
honnibal
PRO
5
480
[RailsConf 2023 Opening Keynote] The Magic of Rails
eileencodes
28
8.9k
Embracing the Ebb and Flow
colly
83
4.4k
Transcript
DISTRIBUTED FAILURE Andrew Godwin @andrewgodwin Learning lessons from aviation
Hi, I’m Andrew Godwin
Content Warning Aviation accidents Road accidents Discussion of death
Software is difficult.
Distributed is even harder.
None
Not unique to distributed systems
None
Who's solved this? Aviation.
A Boeing 747 has six million parts
A Boeing 747 has six million parts
Airplane Car Walking Train 220 130 30.8 Deaths per billion
hours (UK 1990-2000) 30
People matter as much as machines
Pilot 76% Aviation Accident Causes (2005 Nall report) 9% Other
16% Mechanical
Let's look at some aviation principles
Principle #1 Hard Failure
If something is wrong it turns itself off
This only works if you have redundancy
None
These are great ways to ensure you never fix something.
No accident or outage has a single cause. Stop your
code getting into odd states.
None
Single points of failure can be good
None
Principle #2 Good Alerting
Cockpits are incredibly selective about what sets off an audio
alarm
Alert fatigue is real. Avoid at all costs.
Never, ever, put all errors in the same place
Critical Normal Background
Critical Normal Background Wakes someone up. Actionable.
Critical Normal Background Wakes someone up. Actionable. Fixed over the
next week.
Critical Normal Background Wakes someone up. Actionable. Fixed over the
next week. Metrics, not errors.
Have you been ignoring an error for weeks? Then turn
off its error reporting.
Principle #3 Find your limits
Everything will fail. You should know when.
Copyright Boeing
What's your Minimum Equipment List?
REQUIRED OPTIONAL
Did you load test? Did you fuzz test?
You don't have to perfectly scale.
Risk is fine when you're informed!
Principle #4 Build for failure
No single thing in an aircraft can fail and take
it down.
We all want this for our code, but the way
to do it is to build for failure.
Kill your application randomly Practice server network failures Develop on
unreliable connections
The majority of pilot training is handling emergencies.
None
Use checklists. Don't rely on memory.
If you practice failure, you'll be ready when the inevitable
happens.
Pilot 76% Aviation Accident Causes (2005 Nall report) 9% Other
16% Mechanical
Principle #5 Communicate well
Distributed software means separate teams.
As you grow, communication becomes exponentially harder.
None
None
None
Clear communication is vital.
Write everything down.
Have a clear chain of command.
Make decisions.
Principle #6 No blame culture
How do I know all these aviation stats?
Every incident is reported and investigated.
There is never a single cause of a problem.
Make it very difficult to do again.
None
None
Encourage reporting.
Reward maintenance as well as firefighting
None
In aviation, every rule is written in blood.
Software is not yet there. But we are getting closer.
Margaret Hamilton Her error detection code saved Apollo 11
Therac-25 Killed 3, severely injured at least 3 more
None
None
Hard failure Good alerting Find your limits Build for failure
Communicate well No blame culture
Thanks.