OTel Sucks (But Also Rocks!)
Talkat KubeCon + CloudNativeCon Europe 2025 · on Thursday, 2025-04-03· at 11:00· Level 1 | Hall Entrance N10 | Room G·
summary
Config SIG now simplifies SDK configuration but auto instrumentation for .NET and Java sometimes creates extremely high cardinality spans that overload collectors and vendor ingestion until tuned; the collector can suffer from tail‑sampling complexity and Prometheus scrape/performance issues while also proving that OTLP in one binary handles traces/metrics/logs with strong throughput; semantic conventions (HTTP) repeatedly change before stabilizing, downstream distributions break on internal collector interface updates, and the community size drives rapid iteration and real-world fixes.
AI-written summary · about
chapters (12)
- 0:00OpenTelemetry introduction
- 1:44What makes OpenTelemetry complex
- 5:11Auto-instrumentation challenges
- 7:42Semantic conventions problems
- 9:51Collector issues
- 11:14Why OTel rocks
- 12:40Auto-instrumentation benefits
- 14:28Semantic conventions stability
- 16:39Collector power
- 18:40The OpenTelemetry community
- 21:08Resources and closing
- 21:59Q&A
related talks
26mSimplifying OpenTelemetry with Configuration
32mOpenTelemetry feat. logs, metrics and traces
37mShould I Use OTel (collectors), or Is Prometheus Good Enough?
24mOTel-y Oops: Learning From Our Observability Blunders
13mManaging Observability Data at the Edge with the OpenTelemetry Collector and OTTL
32mOpenTelemetry Semantic Conventions and How to Avoid Broken Observability
09mOpenTelemetry or OpenTeleMessy?
53mOpenTelemetry Metrics: observability’s holy grail