Observability Is Not Analytics!
summary
Observability drives a six‑month reengineering to untangle a DB‑centric, tightly coupled system and yields 300+ PRs a month, deployments under 10 minutes and 2–3 releases a day while cutting the AWS bill. Practical lessons: telemetry evolves (schema changes break period‑over‑period reporting), span events and raw high‑cardinality dumps generate noise, HTTP path vs query capture skews endpoint counts, and simple retry counters beat noisy trace events. Observability helps debugging but does not replace stable business analytics or guaranteed SLO measurement.
AI-written summary · about
chapters (37)
- 0:00Intro
- 1:48Dont do things over
- 3:27Observability worked
- 3:44The essence of engineering
- 4:29The bad news
- 4:58Continuity of knowledge
- 6:37Error budgets
- 8:22The old system
- 11:44Deployment frequency
- 12:45Stability
- 13:51Tracking Requests
- 14:55Mental Model
- 15:36Context Path
- 16:13Path
- 17:40If Nothing Else
- 18:40What You Need
- 19:04Retries
- 20:17Span Events
- 21:08Field Tracking
- 21:40High Dimensionality
- 23:49Columns
- 25:15Telemetry
- 26:11Audience Participation
- 26:32Error Rates
- 27:14Global Error Rate
- 28:50Where is this taking us
- 29:50The problem
- 32:36How it ended up
- 34:32Commitments
- 35:29Why does the CEO care
- 35:49Economic Imperatives
- 37:43Where do we land
- 38:06What is engineering
- 38:56What is SRE
- 40:49Platform Engineering
- 41:09Orbit
- 41:31Talon
related talks
36mObservability? It's a Data Analytics Problem, You Fool!
42mObservability can help you understand change
52mObservability: it's not just an ops thing
32mEssential Software Observability for Every Role: Look No Further
40mSurvivability of Observability: Taking your DIY system to success
35mBeyond Logs, Metrics and Traces. The Data Analytics Paradigm Shift in Observability — Dotan Horovits
35mObservability: Building Confidence & Answering the What, Where, How
34mThe State of SQL-Based Observability