Designing a System for Chaos
summary
Chaos engineering experiments on production-like failures to uncover real weaknesses: Netflix’s Simian Army (including Chaos Monkey and Chaos Kong) intentionally kills instances or zones, companies like Uber and Dropbox learned hard lessons, and tools such as Chaos Toolkit and Gremlin automate attacks. A typical test is to terminate DB master (Kubernetes pod) then probe for HTTP 200 and check logs; monitoring (uptime, system, user, service metrics) is mandatory. Start small—disable SSH or reboot a dev box—then fold tests into CI.
AI-written summary · about
chapters (16)
- 0:00Introduction and speaker intro
- 1:13Complexity of modern systems
- 4:46Failures in distributed systems
- 6:54What is chaos engineering
- 9:51Injecting chaos into system layers
- 11:38Chaos engineering principles
- 13:47Types of monitoring for chaos
- 19:13Netflix and the Simian Army
- 25:11System failure inevitability
- 25:54Chaos Toolkit demo
- 28:48Gremlin tool overview
- 30:14Getting started with chaos
- 34:10Interactive chaos learning
- 35:36Chaos engineering resources
- 37:27Chaos in company culture
- 40:19Q&A and closing remarks
related talks
36mChaos Engineering
18mChaos Engineering: Break It On Purpose
15mChaos - Does It Need to Be Included in Test Strategy?
52mGetting started with Chaos Engineering
39mChaos: Breaking your Systems to Make them Unbreakable
38mChaos Engineering
43mBuilding a culture from Chaos
52mChaos engineering: Breaking things on purpose in production