Cloud Native Network Observability: Treating Routers like Kubernetes Resources
summary
Replaces poll‑based SNMP with gNMI streaming telemetry at 5 second samples across a five‑router spine/leaf lab and streams into Prometheus and Grafana, while a Kubernetes stack — Kubernetes Operators and CRDs plus Flux — treats each router as a resource and reconciles targets from a CI renderer. InfraHub serves as the typed, branchable source of truth; the renderer writes git artifacts and Flux applies them. Demo shows cascade deletion and live metrics working, with practical fixes for webhook debouncing, Git provider dispatch differences, tokens and CI runner tooling.
AI-written summary · about
chapters (14)
- 0:00Introduction to talk
- 1:32SNMP vs modern telemetry
- 4:53Problems with manual configuration
- 7:56Telemetry as derived data
- 10:24Kubernetes operator approach
- 12:12Moving to GitOps with Flux
- 14:22GitOps as delivery mechanism
- 15:29InfraHub as source of truth
- 18:10Automated device list generation
- 20:46Safety in reconciliation
- 21:27Full pipeline demonstration
- 25:24Removing devices from monitoring
- 25:48Debugging the pipeline
- 28:18Key takeaways and conclusion
related talks
29m“Are These Things Talking to Each Other?” - Observing Kubernetes Networking...
26mFlux + Observability: Featuring Prometheus Operator and Pixie
36mNetwork Machinery: A United-Front for Network Troubleshooting with CRDs
31mCloud Native Observability
33mCloud Native and Kubernetes Observability Panel: The State of Union
01h 29mTutorial: A Hitchhiker's Guide to CNCF/OSS Observability So...
37mCisco's Devnet Goes Cloud Native: A Case Study on Transforming to Cloud Native [I]
39mThe State of the Cloud Native Network: Deep-dive from TAG-Network