What's New With gRPC's Observability?
summary
gRPC ships concrete observability work: gRPC OpenTelemetry plugins implemented (Java, C++, Go) and marked experimental with planned interop tests; C++ on Linux adds TCP packet-level tracing that records four kernel timestamps plus delivery, retransmit and congestion stats used internally at Google to isolate network vs app latency; new metrics include retries/hedges, WR and XDS in core/Java/Go, subchannel metrics exposing disconnect errors, Dropbox-contributed outlier detection and an optional gcb_backend_service label; features also include binary logging with filtered binary RPC replay, grpcurl, channelz (v2 in progress), grpc_debug CLI and a Perfetto-friendly latency profiler.
AI-written summary · about
chapters (17)
- 0:00gRPC observability intro
- 1:06OpenTelemetry integration
- 2:53OpenTelemetry tracing
- 4:01TCP tracing in C++
- 5:24Why metrics matter
- 6:47Client metrics per attempt
- 8:14New metrics for retries
- 9:43Outlier detection labels
- 11:10TCP metrics proposal
- 12:37Binary logging for debug
- 14:43gRPCurl CLI testing
- 15:47Channelz insights
- 17:35gRPC debug queries
- 18:43Roadmap future plans
- 20:08Bidirectional streaming observability
- 21:30Binary logging for streams
- 22:13Q&A
related talks
19mgRPC Observability: A Guide To Distributed Debugging and Monitoring
29mWhat's new in gRPC Observability
32mUnified Observability in GRPC: Metrics and Tracing Using OpenTelemetry Plugin
24mWhat's New in gRPC
19mWhat's New in gRPC? - Gina Yeh, Google
14mBringing Clarity to gRPC through GoFr: The Role of Observability
16mPeeking Under the Hood: Observability of Kubernetes gRPC Internals
33mgRPC For Microservices: Service-mesh and Observability