Grafana (LGTM)
Grafana Dashboards & Data Sources
Use Grafana as the visualization layer over many backends, not as a store itself.
Grafana is the open-source visualization and dashboarding layer that sits in front of other systems. It stores almost no telemetry of its own; instead it connects to data sources — Prometheus, Loki, Tempo, Mimir, Elasticsearch, SQL databases, cloud monitors — and queries them live. A dashboard is a set of panels, each running a query against a data source and rendering it as a time series, table, gauge, or heatmap.
A universal TV remote: it does not broadcast any channel itself, but it controls every device and shows them all on one screen with the same set of buttons.
Key Concepts
1
Two features make dashboards reusable: template variables and mixed data sources. Variables turn hard-coded values (a service, a namespace) into dropdowns so one dashboard works for every service. Mixed-source panels and data links let a single view pull metrics from Prometheus and logs from Loki side by side, and jump from a spike to the related logs or traces. Provisioning dashboards as JSON in version control keeps them reproducible across environments.
2
What the interviewer is probing.
What the interviewer is probing.
3
1. "What is Grafana actually responsible for?" *Probing:* the positioning. *Stalls:*
"Monitoring." *Moves up:* visualisation and alerting over other people's data — it stores no
telemetry itself.
1. "What is Grafana actually responsible for?"
4
2. "What is a template variable for?" *Probing:* the practical feature. *Stalls:* "Filtering."
*Moves up:* one dashboard serving every service or environment through a dropdown, instead of fifty
near-identical copies.
2. "What is a template variable for?"
5
3. "How should dashboards be managed at scale?" *Probing:* the governance. *Stalls:* "In the
UI." *Moves up:* as code through provisioning, so they are reviewed, versioned and identical across
environments.
3. "How should dashboards be managed at scale?"
6
4. "What makes a dashboard useful on call?" *Probing:* the design. *Stalls:* "More panels."
*Moves up:* the few signals that answer is it broken and where, with detail a click away — a wall of
graphs is unreadable at 3am.
4. "What makes a dashboard useful on call?"