Unified Observability in GRPC: Metrics and Tracing Using OpenTelemetry Plugin
Talkat KubeCon + CloudNativeCon China 2025 · on Wednesday, 2025-06-11· at 11:00· Level 16 | Grand Ballroom I·
summary
gRPC adds an OpenTelemetry plugin that emits unified metrics and traces without changing app code, using interceptors plus user-provided SDK/exporters (Prometheus/OTLP). It produces three views: grpc.client.call.duration for end-to-end latency, grpc.client.attempt.duration to separate retries/hedges, and grpc.server.call.started for server concurrency and queueing, plus per-message trace events to pinpoint streaming delays. Metrics APIs are stabilized across languages; tracing remains experimental but interoperable with OpenCensus.
AI-written summary · about
chapters (16)
- 0:00gRPC and observability challenges
- 3:20Why gRPC is hard to debug
- 5:50OpenTelemetry plugin overview
- 7:57Why OpenTelemetry
- 11:10OpenTelemetry architecture
- 15:49Metrics configuration
- 16:11Tracing in gRPC
- 20:13gRPC plugin scenarios
- 21:20Client metrics breakdown
- 23:28Retries vs bottlenecks
- 25:44Client traces and instrumentation
- 27:35Monitoring retries
- 28:39Diagnosing streaming bottlenecks
- 29:41Performance optimization
- 30:23Plugin status and future
- 31:27Q&A
related talks
19mgRPC Observability: A Guide To Distributed Debugging and Monitoring
22mWhat's New With gRPC's Observability?
30mOpenTelemetry 101
46mThe State of OpenTelemetry
29mWhat's new in gRPC Observability
20mObservability and Open Telemetry
50mOpen Telemetry: The Rise of Open Standards in Observability
34mOpenTelemetry: The Vision, Reality, and How to Get Started