Mirror, Mirror Tell Me If I Am Observable?
summary
The talk argues that choosing telemetry by formal risk analysis (FMEA/RPN) instead of ad hoc dashboards greatly reduces false positives and mean time to repair. It describes an implementation using OpenTelemetry, Prometheus, Jaeger, Loki and Grafana across auto‑ and manually‑instrumented services, Kafka and Kubernetes, with traces for end‑to‑end drilldown and results including richer SLIs/SLOs, fewer false alerts, lower MTTR and experiments with dynamic thresholds using simple machine learning.
AI-written summary · about
chapters (14)
- 0:00Introduction
- 1:50Monitoring definitions
- 5:19First monitoring attempt
- 7:23Dashboard failure lessons
- 9:52FMEA risk analysis
- 12:42Improving monitoring with FMEA
- 15:30Evolution to v2.0
- 17:37SLO and SLI methodology
- 20:49Methodology importance
- 22:57Microservices architecture
- 24:22Observability in modern systems
- 26:50Observability architecture demo
- 35:11Monitoring and observability complementarity
- 36:35Q&A
related talks
47mMirror, mirror, tell me if I'm observable?
38mThe Murphy's Laws of Observability
26mYou Can’t Fix What You Can’t See: Tales from the Observability Frontier
30mPanel: Telemetry That Matters
46mObservability Ecosystem in Kubernetes
48mAPI Observability by Design: Building Systems That Tell Their Own Stories
29mObservability; a gentle introduction
50mOpen Telemetry: The Rise of Open Standards in Observability