Profiling is for Dev, Monitoring is for Ops – no more!
summary
Claims that Prometheus-style whitebox monitoring and always-on Go profiling let developers instrument production binaries, attach pprof endpoints via an anonymous import, and profile live traffic. Demonstrates Prometheus (written in Go) used at SoundCloud to debug and optimize itself, uses metrics at /metrics and on-the-fly CPU profiles, and shows why ~100 probes are needed to estimate a one‑minute 99th‑percentile latency. The tools are concrete and the approach demonstrably works in production.
AI-written summary · about
chapters (11)
- 0:00Start
- 3:09Site Reliability Engineering
- 5:06SRE vs. DevOps
- 9:11Symptom-based alerting
- 11:06Black-box monitoring FTW?
- 14:121. Your probe is not real user traffic.
- 21:11What makes white-box monitoring a dev concern?
- 27:46Profiling in Go Anything new here?
- 31:36What makes profiling an Ops concern?
- 34:52Next step: Distributed Tracing always and everywhere...
- 36:362. Play with Prometheus.
related talks
51mProduction Profiling: What, Why and How
06mContinuous Profiling
44mGo Profiling and Observability from Scratch
40mContinuous Profiling, the missing piece in your observability puzzle!
28mConprof - Profiling in the cloud-native era
24mCommunity Driven Breakout Session: Continuous Profiling
33mBuilding a Go Profiler Using Go
56mContinuous Profiling the missing piece in your observability puzzle