Auto-instrumentation of Prometheus for RED Monitorin...
summary
Claims that eBPF can auto-instrument RED metrics for Prometheus by hooking kernel events, using a Cloudflare eBPF exporter to read eBPF maps and expose metrics. Demo wires exporter to Prometheus, hooks accept4 and connection syscalls, and produces rate and error counters that match handcrafted Prometheus metrics while also revealing two extra 400 errors the app missed. Implementation requires C eBPF programs, privileged containers, kernel‑specific hacks, limited stack and portability concerns, and nontrivial maintenance.
AI-written summary · about
chapters (17)
- 0:00Introduction
- 3:10What is RED Monitor
- 5:28Prometheus
- 10:34Service Mesh
- 12:39What is Auto Instrumentation
- 13:30What is DBPF
- 14:10What is EBPF
- 15:36EBPF Exporter
- 16:28Integrations
- 17:40Tradeoffs
- 19:05Development
- 19:47Linux
- 20:59Key probes
- 21:32Demo
- 24:50Golang test
- 27:07Prometheus metrics
- 31:09Feasibility
related talks
36mPrometheus as exposition format for eBPF programs ..
29mAuto-instrumentation for GPU performance using eBPF
30mAgentless instrumentation with eBPF, Prometheus and OpenTelemetry
18mFind out what your processes are capable of with ebpf
21mAuto-Instrumentation for GPU Performance using eBPF
42mCommunity Driven Breakout Session: eBPF and Auto-Instrumentation
42meBPF-Powered Observab...
35mAn Exemplary Path: Leveraging EBPFs and OpenTelemetry To Auto-instrument for Exemplars