What's new in gRPC Observability
summary
gRPC adds concrete observability: a new disconnection_reason metric on connections, TCP-level telemetry (retransmits, RTT) for Java and C++, TLS handshaking metrics and a C++ handshake plugin, outlier-detection counters for Go, and XDS client metrics finally plumbed into Go. Call-level changes include attempt-level metrics with retry_delay, ORCA-style client↔server load signals and a backend_service label for routing dimensions, plus a single per-call custom string. Roadmap shows RPC delay observability for name resolution and load balancing and tracing API stabilization targeted for November.
AI-written summary · about
chapters (20)
- 0:00Introduction to gRPC observability
- 1:04Observability fundamentals
- 3:57Telemetry data flow in gRPC
- 4:40New metrics in gRPC Java
- 6:27C++ telemetry and TLS metrics
- 7:08Go metrics and XDS client updates
- 8:11Retry and hedging metrics
- 8:32ORCA metrics for load balancing
- 9:16Labels and backend service metrics
- 10:41Custom call metrics
- 11:24Short-term roadmap overview
- 12:48RPC delay observability
- 13:09OpenTelemetry semantic conventions
- 15:18Q&A and community resources
- 16:22OpenTelemetry streaming and semantic questions
- 18:34Streaming metrics and cost considerations
- 22:32OpenTelemetry stabilization timeline
- 24:01Cardinality and metric costs
- 25:49RTT metrics frequency and configuration
- 27:14Cost management for metrics and tracing
related talks
22mWhat's New With gRPC's Observability?
19mgRPC Observability: A Guide To Distributed Debugging and Monitoring
32mUnified Observability in GRPC: Metrics and Tracing Using OpenTelemetry Plugin
33mgRPC For Microservices: Service-mesh and Observability
19mWhat's New in gRPC? - Gina Yeh, Google
26mFeature Work In GRPC; XDS And Not
04mObservability in micro-service architectures
20mAchieving Observability into Your Application with OpenCensus