Peeking Under the Hood: Observability of Kubernetes gRPC Internals
summary
Instrumentation that captures kubelet→containerd gRPC calls without changing Kubernetes code uses eBPF hooks and an OpenTelemetry SDK or the cluster's native gRPC telemetry flags, sends spans to an OpenTelemetry Collector running as a DaemonSet on port 4317, and exports to Jaeger for visualization. The demo shows end-to-end traces for image pulls, container start latency and volume-mount failures, proves root causes can be pinpointed, and notes Jaeger’s default in-memory store drops traces unless replaced.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 1:27gRPC in Kubernetes
- 2:31Logging limits and debugging
- 4:15Runtime tracing need
- 4:58OpenTelemetry SDK approach
- 7:29Kind cluster setup
- 9:16Native gRPC telemetry
- 11:01OpenTelemetry collector role
- 13:08Jaeger for gRPC traces
- 14:52End-to-end visibility
related talks
28mImplementing Observability in Kubernetes
46mObservability Ecosystem in Kubernetes
43mA Guide to Observability for Your (Java) Apps in K8s
35mNavigating the Distributed Systems Execution Maze With OpenTracing
31mDistributed Tracing Integration with OpenTelemetry and Knative
46mA hitchhikers guide to observe (Java) applications in Kubernetes
30mMix Kubernetes with Linux Kernel Tracing to Boost the Observability of Your Cluster
10mExploring the Depths: Distributed Tracing with OpenTelemetry on ClusterAPI - GCP