Who Needs an API Server to Debug a Kubernetes Cluster?
summary
Shows a concrete way to debug Kubernetes nodes when the API server is down by building a single static binary called Local-Gadget that runs eBPF tracers (many borrowed from BCC) and automatically maps kernel events to containers and pods. Demo walks through the painful manual flow (ps to PID, extract mount namespace, pin hex key into an eBPF map), then shows automation via runc notify and eBPF map update on creation so tracers start before containers run. Result: live exec, socket and OOM traces enriched with Kubernetes metadata work even with the API server unavailable.
AI-written summary · about
chapters (9)
- 0:00Start
- 9:19Demo #1: What issues did we find?
- 10:45Local Gadget: What is it?
- 12:00Local Gadget: Architecture
- 14:03Data enrichment (Container-Collection)
- 16:29Filtering (Trace-Collection)
- 18:12Local-Gadget: Internal modules
- 18:27Local Gadget: Use-Cases
- 26:04The future of Local Gadget
related talks
27mStep By Step Kubernetes Observability with eBPF
01h 00mDebug like a pro on Kubernetes
28mBringing the power of eBPF tools to your k8s web UI
34mDebugging at Scale in Production - Deep into Your Containers with Kubectl Debug
50mTroubleshooting & Debugging Microservices in Kubernetes
01h 32mTutorial: Debug Your Kubernetes Apps
40mDebugging in Golang Apps Inside K8s Environment
30mObservability in Kubernetes: Unlocking the Power of eBPF