From PIDs to Pods: the life cycle of an eBPF-autoinstrumented Kubernetes app
summary
eBPF-based Beyla auto-attaches at the kernel level to produce HTTP request duration and traces enriched with path and method without touching app code, mapping kernel PIDs to Kubernetes Pods by reconciling /proc container IDs and PID namespaces. Deployment options are per-pod sidecar or DaemonSet with hostPID (requires privileged + hostPID), and a centralized Kubernetes cache replaces per-node informers to avoid API overload in thousand-node clusters. Performance tests show minimal overhead; metrics and traces work today, logs do not.
AI-written summary · about
chapters (9)
- 0:00eBPF auto-instrumentation intro
- 3:55What is BA and how it works
- 4:58eBPF overview and use cases
- 7:51Instrumenting K8s apps with BA
- 10:22PIDs to Kubernetes pods
- 13:38BA at scale in K8s
- 15:31Live demo of BA
- 17:16OpenTelemetry and BA future
- 18:00Q&A
related talks
24mFrom PIDs to Pods: The Life Cycle of an eBPF-Autoinstrumented..
41mFrom PIDs to Pods: our journey to connect eBPF and Kubernetes
30mObservability in Kubernetes: Unlocking the Power of eBPF
27mLow-Overhead Tracing Using eBPF for Observability into Kubernetes Apps and Services
27mStep By Step Kubernetes Observability with eBPF
35mAn Introduction to eBPF - Superpowers for Networking, Observability & Security
42mCommunity Driven Breakout Session: eBPF and Auto-Instrumentation
24mMaking the Kernel Speak Kubernetes: Unlocking eBPF’s Power for Observability