Achieving 99.99% uptime with Open Source Observability
summary
Claim: 99.99% uptime needs end-to-end visibility, achieved by instrumenting services and reducing noisy data. Uses OpenTelemetry sdks and the OpenTelemetry Collector with receivers (>50 available), processors like kubernetes attributes processor and tail sampling processor, and Java auto instrumentation that runs in under 10 minutes. Shows concrete wins: 52 minutes downtime per year target, a customer saves ~33% on log ingest by filtering, and complex 300+ line sampling rules that keep diagnostically useful traces while dropping the rest.
AI-written summary · about
chapters (15)
- 0:00Introduction
- 3:08How to make developers happier
- 4:07Putting on a fire
- 5:35Manual correlation
- 7:17Tracing
- 9:08Why Observability is Hard
- 14:56Why Observability is a Problem
- 16:07How to Increase ROI of Observability
- 17:32How Open Source Observability Works
- 18:15Observability Receivers
- 19:00Observability Attributes Processor
- 20:23Observability Filter Processor
- 21:14Tail Sampling Processor
- 25:23Open Telemetry Recap
- 27:22Questions
related talks
12mMastering Observability in Distributed Systems
41mCustomer Centric Observability at Scale Leveraging AIOps & OpenTelemetry
11mObservability 2.0 - Much more than just logs,metrics and traces
30mPanel: Telemetry That Matters
28mScaling Observability for Real-Time Personalization Engines
25mMaking Observability Work at the Edge
13mNext-Gen Observability: reduce MTTR
19mOpenTelemetry's Path to Observability's Future