Budgets keep shrinking while observability demands explode—so how do you capture L7 insight across 25+ Kubernetes clusters without paying a sidecar tax? At UW Madison, we paired Cilium’s eBPF networking with the Prometheus Agent’s remote write, streaming rich flow telemetry to a central Grafana stack at a fraction of the usual footprint. Live on stage, we’ll fire test traffic with curl from a single app pod and watch Cilium Flow Metrics spike and Alertmanager light up within seconds—no sidecars, no context switching. You’ll leave with a field-tested playbook: scrape-target pruning, discovery patterns that survive air-gapped research zones, and the real story of what broke first—and how we fixed it. This case study demonstrates light weight mesh and light weight observability for a university budget.