How many dashboards does your Grafana have?
summary
Uses Prometheus + Grafana + exporters and Alertmanager to argue for monitoring symptoms not failures, showing a Redis key leak discovered in an ugly spike and a disk full predicted by Prometheus linear prediction (alert fired ~90 minutes before crash). Recommends splitting overview vs deep‑dive dashboards, provisioning dashboards as code (reducing 106 autogenerated JSON diffs to 11), avoiding repeating panels, checking for double‑counting/physical impossibilities, and using basic anomaly detection rather than constant thresholds.
AI-written summary · about
chapters (12)
- 0:00Intro
- 1:41Why Gratana
- 3:54S startup
- 5:45Context Matters
- 15:07Question Your Data
- 20:15Clutter
- 23:25Overview at a Glance
- 29:02Don't Leave a Mess
- 34:09Version Control
- 36:52Application Monitoring
- 40:45Alerting
- 43:10Summary
related talks
50mMetrics and alerting with Prometheus and Grafana
29mParanoid Observability –Success Stories and Misfortunes observing an observability stack
31mService cockpit - developer friendly Grafana based monitoring aggregator
29mFool-Proof Kubernetes Dashboards for Sleep-Deprived Oncalls
20mThe Evolution of Grafana Alerting
37mPractical Capacity Planning using Prometheus
54mA Practical Guide to Dashboarding
39mBeyond Logs: Achieving Observability with Selenium and Grafana