From PIDs to Pods: The Life Cycle of an eBPF-Autoinstrumented..
summary
Automatic instrumentation that attaches eBPF probes to live processes and exports Prometheus or OpenTelemetry metrics and traces, implemented as Grafana Beyla. Runs as a DaemonSet with privileged + hostPID so a sidecar pod sees host processes, probes read /proc and use the eBPF helper get_current_task with pid namespace level to map host PIDs to Kubernetes pod/container IDs, then decorates samples with pod metadata. Demo shows per-endpoint latency, dependency graphs, it works at cluster scale but needs reduced privileges and a centralized pod-metadata cache.
AI-written summary · about
chapters (8)
- 0:00eBPF auto-instrumentation intro
- 1:25Manual instrumentation challenges
- 5:25What is eBPF and how it works
- 9:23Baya in Kubernetes
- 11:55PID to Pod mapping
- 17:01Live demo of Baya
- 21:27Baya future roadmap
- 23:18Summary
related talks
22mFrom PIDs to Pods: the life cycle of an eBPF-autoinstrumented Kubernetes app
41mFrom PIDs to Pods: our journey to connect eBPF and Kubernetes
18mFind out what your processes are capable of with ebpf
27mStep By Step Kubernetes Observability with eBPF
05mBeyla: eBPF automatic instrumentation
35mAn Exemplary Path: Leveraging EBPFs and OpenTelemetry To Auto-instrument for Exemplars
42meBPF-Powered Observab...
31mZero instrumentation observability with eBPF