| name | grafana |
| description | Build and operate Grafana dashboards for operational, observability, and time-series workflows. Use for panels, variables, alerting, datasource setup, dashboard provisioning, incident views, and self-hosted OSS delivery when the selected path is Grafana. |
| license | MIT |
| argument-hint | [operational-use-case] [datasource-scope] |
| metadata | {"author":"data-visualization-kit","version":"1.0.0"} |
Grafana Skill
Production-ready Grafana delivery for operational dashboards, time-series metrics, observability workflows, alert-aware panel design, and self-hosted dashboard operations.
When to Use
- Building dashboards for operational monitoring, observability, or near-real-time status tracking
- Working with time-series, logs, traces, or infrastructure/application metrics
- Wiring Grafana to Prometheus, Loki, Tempo, PostgreSQL, MySQL, Elasticsearch, or similar datasources
- Designing panels, variables, dashboard drill paths, alert views, or incident-oriented layouts
- Provisioning or maintaining self-hosted Grafana in OSS environments
Grafana Selection Guide
Choose Grafana when:
- the primary user workflow is operational monitoring or incident response
- the data is time-series, event-stream, or telemetry heavy
- dashboards need templating, time ranges, annotations, or alert context
- the stack already lives in observability tooling
Do not force Grafana when:
- the job is general BI, stakeholder self-service analytics, or lightweight business dashboarding
- the project would be simpler and clearer in Metabase
- the team mainly needs governed SQL questions instead of operational monitoring surfaces
See: references/grafana-core.md
Grafana Mindset
The 10 Commandments of Grafana Delivery:
- Design for operators under pressure
- Time range is part of the question
- Panels must explain action, not just display signal
- Variables should reduce noise, not hide it
- Alert context belongs near the dashboard
- Datasource contracts matter more than panel cosmetics
- One dashboard should serve one operational job
- Provision what must be reproducible
- Logs, metrics, and traces should connect cleanly
- If the ask is BI-first, move it to Metabase
See: references/grafana-operations.md
Reference Navigation
Core Domain References:
grafana-core.md - datasource choices, panels, variables, time range design, alert-aware dashboard construction
grafana-operations.md - provisioning, alert context, incident views, performance, governance, deployment heuristics
Key Best Practices
Dashboard Design:
- Start from operator tasks: detect, triage, isolate, confirm
- Make time range, environment, and service selection obvious
- Prefer a few decisive panels over huge galleries of weak metrics
Datasource and Querying:
- Validate sampling interval, aggregation window, and retention assumptions
- Align legends, units, thresholds, and null handling across related panels
- Treat datasource schema changes as dashboard contract changes
Operational Readability:
- Use thresholds, annotations, and labels to accelerate triage
- Keep dashboard sections aligned to system layers: traffic, errors, latency, saturation, infra, logs
- Connect related logs or traces when the workflow benefits from it
Provisioning and Ops:
- Provision datasources and dashboards when reproducibility matters
- Keep secrets out of dashboard JSON and repo content
- Document datasource ownership, environment mapping, and alert dependencies
Quick Decision Matrix
| Need | Choose |
|---|
| Operational monitoring | Grafana |
| Time-series metrics and service health | Grafana |
| Logs, traces, metrics correlation | Grafana |
| General stakeholder BI dashboard | Metabase |
| Legacy SQL-lab-heavy analytics estate | Apache Superset |
| DV operational portfolio path | Grafana |
Implementation Checklist
Intake:
- Confirm the dashboard is operational, observability, or time-series first
- Identify users, incident workflows, SLIs/SLOs, and required datasources
- Confirm deployment model and access boundaries
Datasource Layer:
- Validate datasource connectivity and query semantics
- Normalize metric names, units, tags, and environment labels
- Confirm retention windows and acceptable refresh cost
Dashboard Layer:
- Build dashboards around triage flow, not datasource structure
- Add variables only where they reduce repeated dashboards
- Use thresholds, annotations, and panel descriptions where needed
Operational Layer:
- Review alert context, links, incident handoff, and runbook entry points
- Provision dashboards/datasources if environment consistency matters
Quality:
- Verify queries against live or representative data
- Check empty-state behavior, time-range behavior, and variable interactions
- Confirm the dashboard still makes sense under pressure
Common Pitfalls to Avoid
- Using Grafana for BI work that should live in Metabase
- Building huge dashboards with no clear operator journey
- Hiding weak metrics behind too many variables
- Mixing unrelated services into one dashboard
- Ignoring units, nulls, or time aggregation side effects
- Treating provisioning as optional when multiple environments exist
Resources
Official Documentation:
Data Visualization Kit Context:
- Grafana is the preferred Data Visualization Kit path for operational and time-series dashboards
- Prefer Metabase when the ask is general BI or stakeholder analytics