Observability: Metrics, Logs, Traces
15 min
Day 102: Grafana alerting
Alerting from dashboards
Grafana can alert directly on a panel's query, as an alternative or complement to Prometheus's own alerting rules — useful when the alert condition is easiest to express as "whatever this dashboard panel is already showing, crossing some threshold."
Where to put alerting logic
A common team convention: put alerts that matter for correctness/SLOs in Prometheus/Alertmanager (version-controlled, closer to the data), and use Grafana alerting more sparingly for ad hoc, dashboard-specific conditions — avoiding the same alert being defined (and drifting) in two places.
Key terms
- Grafana alerting
- Alerting configured directly on a dashboard panel's query.
Why might a team deliberately avoid defining the same alert in both Prometheus rules and Grafana alerting?
