eBPF in Go: Observability for AI-Generated Services

[Cheikh Seck]

Hi gophers, 

 Wrote a hands-on piece on using eBPF with Go to gain kernel-level observability for AI-generated services. 

Covers kprobes, tracepoints, and perf-event ringbuffers - all from pure Go code. The problem: I shipped a service that looked fine in staging (P95 latency 40ms) but was slow in production with no visibility into kernel-level issues. eBPF let me trace syscalls, network events, and file I/O directly from the kernel. 

Read it here: https://cheikhhseck.medium.com/ebpf-in-go-observability-for-ai-generated-services-9aae7573b823 

 Curious if anyone else is running eBPF-based observability in production with Go - would love to compare notes on map sharing and BTF compatibility.

[Jason E. Aten]

Neat. And interesting. eBPF was cool to learn about. So, thank you Cheikh.

Two nits: you don't compare to using pprof, which would have been the first thing to pull out.

Second, eBPF is available on windows; so the article's claim there is misleading at best.

[Cheikh Seck]

Thank you Jason. Will update the article. I want to blame Lite Agent but it's my fault for failing to verify such a claim.