OpenTelemetry in production for a year plus: the good the bad the ugly
summary
Production OpenTelemetry (deployed Feb 2023) instruments services and ships telemetry to a Prometheus + Grafana backend via an OpenTelemetry Collector; deployment runs serverless on Google Cloud Run and delivers consistent metrics. Works mostly well: stable batching, backend-agnostic exporters, and predictable scraping. Pain points are SDK fragmentation across languages, heavy boilerplate (example: 23 lines of Python just to emit metrics), and architectural limits — label cardinality explodes backends and push-and-batch latency prevents real-time guarantees.
AI-written summary · about
related talks
35mOpenTelemetry for Busy People
38mThe State of OpenTelemetry
49mAdding OpenTelemetry to Production Apps Lessons Learned
30mOpenTelemetry 101
28mOpenTelemetry Is Not a Silver Bullet
34mOpenTelemetry: The Vision, Reality, and How to Get Started
50mOpen Telemetry: The Rise of Open Standards in Observability
09mOpenTelemetry or OpenTeleMessy?