Beginner's guide to observability
summary
Crash‑course checklist: use structured logging (no stdout, SLF4J, async appenders), strip secrets and add correlation IDs; centralize with ELK or agents (push vs pull) and prefer JSON/OTLP formats. Treat metrics as time‑series in Prometheus + Grafana: counters, gauges (store max/min), histograms for percentiles (P95/P99/P999), follow RED/USE golden signals and beware high cardinality. Implement distributed tracing with OpenTelemetry, propagate trace IDs, sample (head/tail/dynamic). For unknown unknowns use Honeycomb-style event tooling and columnar storage (ClickHouse); one real case shows a missing index yielding ~20–30x speedup.
AI-written summary · about
chapters (13)
- 0:00Introduction to observability
- 2:15Why observability matters
- 8:56Centralizing logs in modern systems
- 11:50Understanding metrics and types
- 20:07Golden signals and RED/USE models
- 23:39User experience with UP9 metric
- 27:54Percentiles and latency measurement
- 33:10Distributed tracing fundamentals
- 39:14Sampling traces in distributed systems
- 47:31OpenTelemetry overview
- 48:35Events and observability 2.0
- 50:44Live demo with Honeycomb
- 54:15Observability as iterative process
related talks
26mYou Can’t Fix What You Can’t See: Tales from the Observability Frontier
38mPractical Guide to Application Observability in 10 Steps
49mImproving Observability of Distributed Systems
42mFirst Steps with Observability: OpenTelemetry and You
33mWhat Developers Should Know About Observability
16mBuilding Observability into Cloud-Native Applications
31mObservability for Developers: How to Get There
29mObservability; a gentle introduction