The Ticking Timebomb of Observability Expectations
summary
Argues that “monitor everything” leads to millions of low‑value signals, costly vendor bills and cognitive overload: expect 10k metrics per cluster, tracing and logs that nobody uses, and situations where observability costs eat 20% of cloud bill. Recommends using an OpenTelemetry collector to filter hot noise, designing separate status vs diagnostic dashboards (red/yellow/green entry points) and prioritizing a few customer‑impact metrics over turnkey metric sprawl.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 1:32The problem with observability
- 11:34Dead ends and bad ideas
- 14:30Constructing meaningful data
- 17:38Fighting cognitive overload
- 21:07Prioritizing important signals
- 24:41Managing costs and assumptions
- 29:42Evaluation and conclusion
- 35:09Q&A and closing
related talks
38mThe Ticking Time Bomb of Observability Expectations
38mThe Murphy's Laws of Observability
18mHow not to practice observability
30mPanel: Telemetry That Matters
16mShrinking the Observability Bill
26mDriving O11y x to Nx: Probing Bullet Holes and not Breaking the Bank
10mWhy your observability is costing you too much?
22mObservability is Too Expensive!