Lost Art of Troubleshooting
summary
Cloud’s “destroy and rebuild” fails for production incidents so the talk argues for concrete troubleshooting: replicate isolate understand fix rather than vague examples. It shows tools and results: logging tuned for actionable entries, monitoring tied to business metrics, profiling with CPU per URL via DTrace to find the hot endpoint. Concrete numbers and tradeoffs appear — 200ms to 600ms to >1s degradation, a widget driving 27% revenue from recommendations, caching that sacrifices 12% of pages to keep the rest live.
AI-written summary · about
chapters (15)
- 0:00Introduction
- 2:05Why is troubleshooting important
- 3:56What is the first step
- 7:33Understanding the problem
- 9:33Understand the business
- 11:41Understand the impact
- 14:59Incremental improvements
- 17:01Dont assume
- 17:44Dont trust the errors
- 19:01Troubleshooting is stressful
- 19:48Everything is broken
- 22:00Tools
- 22:43Logs
- 24:17Monitoring
- 28:43Troubleshooting
related talks
01h 05mThe art and methodology of troubleshooting
39mHow to Use Common Investigation Practices in Troubleshooting
40mProduction Detectives: Solve the Case Before the Crash
31mLow-Level Computing in a Cloud Native World
43mSoftware Operations and Human Cognition
52mPerformance & Stability Anti patterns
21mLive Q&A Session #1
43mObservability for Emerging Infra: What Got You Here Won't Get You There