An Exemplary Path: Leveraging EBPFs and OpenTelemetry To Auto-instrument for Exemplars
Talkat KubeCon + CloudNativeCon Europe 2025 · on Wednesday, 2025-04-02· at 17:00· Level 1 | Hall Entrance N10 | Room E·
summary
Demo shows an eBPF agent (BA) attached to the kernel that auto-instruments exemplars by translating network/syscall events into OpenTelemetry metrics and traces, injecting the traceID/spanID into histogram buckets (example: a 5 ms request landed in the 10 ms bucket) and exporting to Prometheus, Jaeger and Grafana via the OpenTelemetry Collector. Setup uses K Lima, Helmfile, Cortex, seaweedFS and Grafana and works end-to-end with click-to-open traces. Caveats: kernel-only visibility limits business semantics and high-cardinality attributes need filtering.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 1:04What are exemplars
- 5:01eBPF for auto instrumentation
- 6:06Live demo: eBPF auto instrumentation
- 17:58Exemplars in observability
- 20:06Auto vs manual instrumentation
- 21:52Exemplars with machine learning
- 25:01Key takeaways
- 25:46Q&A
related talks
30mAgentless instrumentation with eBPF, Prometheus and OpenTelemetry
31mZero instrumentation observability with eBPF
34mNetwork Observability with eBPF and OpenTelemetry
45mSimplifying Instrumentation in complex systems at scale with Open Telemetry
31mEffortless, standardised homelab observability with eBPF
33mEffortless, Standardised Homelab Observability With eBPF
32mFrom Traces To Action: Auto-Instrumenting LLMs for Observability...
42mCommunity Driven Breakout Session: eBPF and Auto-Instrumentation