Glossary — Speak the Language
The five friends may use car dashboard words, but in job interviews and documentation people use these. This page has everything you need to talk about Grafana with confidence: a quick-reference table, a set of interview-ready answers, and a Classic Interview Q&A section.
Quick Reference
| Term | Car dashboard meaning | Plain meaning |
|---|---|---|
| Grafana | The dashboard system | An open-source visualization and analytics platform for monitoring data |
| Dashboard | The instrument cluster | A collection of panels arranged to display monitoring data |
| Panel | A single gauge | A visualization (graph, gauge, stat, table) showing one or more queries |
| Data Source | The engine sensor connection | An external system (Prometheus, Loki, etc.) that Grafana reads data from |
| Query | The sensor reading | An expression (PromQL, SQL, etc.) that fetches data from a data source |
| Visualization | How the gauge looks | The display type: time series, gauge, stat, table, bar chart, etc. |
| Variable | Dashboard control knob | A dropdown that filters all panels on a dashboard |
| Row | Dashboard section | A horizontal group of panels that can be collapsed |
| Threshold | Warning color band | Color-coded value ranges (green/yellow/red) on panels |
| Annotation | Event marker on the timeline | A vertical line on graphs highlighting specific events |
| Alert | Warning light | A notification triggered when a metric exceeds a threshold |
| Notification Channel | Where the warning goes | The destination for alerts (email, Slack, PagerDuty) |
| Contact Point | Who gets the alarm | A configured notification endpoint in Grafana |
| Notification Policy | Alarm routing rules | Rules that route alerts to the right contact point based on labels |
| Silence | Muting the alarm | Temporarily suppressing alerts during maintenance |
| Provisioning | Pre-configured dashboard | Automatically setting up data sources, dashboards, and alerts from files |
| Organization | Different car brands | Multi-tenant separation within a single Grafana instance |
| Folder | Dashboard category | A logical grouping for dashboards (e.g., "Infrastructure", "Application") |
| Time Range | How far back to look | The period displayed on the dashboard (last 1h, 6h, 24h, etc.) |
| Refresh Rate | How often the gauge updates | How often Grafana re-queries the data source (default: 30s) |
| Explore | Diagnostic mode | A raw query interface for ad-hoc investigation without saving panels |
| Snapshot | Dashboard photo | A saved, static copy of a dashboard at a point in time |
| Dashboard JSON | The dashboard blueprint | The JSON representation of a dashboard — can be exported and imported |
| Legend | Gauge label | Text labels identifying each line/series on a graph |
| Transform | Data reshaping | Modify data before display (merge, filter, calculate, rename) |
| Stat panel | Odometer | Shows a single current value with optional color thresholds |
| Gauge panel | Fuel gauge | Shows a value within a min/max range with color bands |
| Time series panel | Speedometer | Shows values changing over time — the most common panel type |
| Table panel | Diagnostic report | Shows structured data in rows and columns |
| Bar gauge panel | Comparison chart | Compares values across categories |
Interview-Ready Answers
Core Concepts
Grafana Grafana is an open-source visualization and analytics platform. It reads data from external sources (Prometheus, Loki, Elasticsearch, etc.) and displays it as interactive dashboards. Grafana doesn't store metrics — it's purely a visualization layer. The gotcha: Grafana is only as good as its data source. If Prometheus has no data, Grafana shows nothing.
Dashboard A dashboard is a collection of panels arranged in a grid. Each panel has its own query and visualization type. Dashboards can include variables (dropdowns), annotations (event markers), and alerts. The gotcha: a dashboard with 50 panels is slow and overwhelming. Keep dashboards focused — one dashboard per service or signal type.
Panel A panel is a single visualization — a graph, gauge, stat, or table. Each panel has a query, a visualization type, optional thresholds, and optional alerting. Panels are the atomic unit of Grafana dashboards. The gotcha: one panel should answer one question. If a panel shows too many lines, split it into multiple panels.
Data Source
A data source is a connection to an external system. Grafana reads from data sources but never writes to them. Common data sources: Prometheus (metrics), Loki (logs), Tempo (traces), PostgreSQL (relational data). The gotcha: the data source URL must be reachable from the Grafana container. Inside Docker, use service names, not localhost.
Visualization
Variables
Variables are dropdown menus that filter all panels on a dashboard. They're defined with queries that return label values (e.g., label_values(up, instance)). Variables make dashboards reusable — one dashboard works for every server. The gotcha: variables must be used in panel queries with $variable_name syntax.
Thresholds Thresholds are color-coded value ranges on panels. They change panel colors based on values (green < 70%, yellow 70-90%, red > 90%). Thresholds make problems visible at a glance. The gotcha: threshold values depend on your system's baseline. A 90% CPU threshold might be normal for a batch processing server.
Annotations Annotations are vertical markers on time series graphs that highlight specific events (deploys, incidents, maintenance windows). They help correlate metrics with events. The gotcha: annotations require a data source query or manual input. Without annotations, you're guessing which deploy caused the spike.
Operations
Provisioning Provisioning is the practice of defining Grafana resources (data sources, dashboards, alerts) as files and loading them automatically on startup. This makes dashboards version-controlled and reproducible. The gotcha: provisioned resources can be modified in the UI, but changes are lost on restart. Edit the files, not the UI.
Alerting in Grafana Grafana alerting evaluates conditions and sends notifications. It works with any data source (not just Prometheus). Alerts can be created from panels or defined as code. The gotcha: Grafana alerting is separate from Prometheus alerting. Use Grafana for dashboard-centric alerts, Prometheus for application-level monitoring.
Classic Interview Q&A
Q1: What is the difference between Grafana and Prometheus?
Answer: Prometheus collects and stores metrics. Grafana visualizes them. Prometheus has a basic built-in graph UI; Grafana provides rich, interactive dashboards. In practice, they're used together — Prometheus for data collection and alerting, Grafana for visualization. The gotcha: Grafana is not a replacement for Prometheus. You need both.
Q2: How do you make a Grafana dashboard production-ready?
Answer: Five steps: (1) Use variables for reusability. (2) Set thresholds for visual health indicators. (3) Add annotations for deploy correlation. (4) Provision dashboards as code (version-controlled). (5) Keep dashboards focused — one per service or signal type. The gotcha: a production dashboard should answer three questions: "Is everything healthy?" → "Which service has a problem?" → "What's the root cause?"
Q3: What is the difference between Stat, Gauge, and Time Series panels?
Answer: Stat shows a single current value (like an odometer). Gauge shows a value within a range with color bands (like a fuel gauge). Time series shows values changing over time (like a speedometer). Use Stat for key metrics at a glance. Use Gauge for percentages. Use Time series for trend analysis. The gotcha: don't mix panel types randomly. Organize logically: gauges at top, time series below.
Q4: How do you handle multi-tenancy in Grafana?
Answer: Grafana supports organizations for multi-tenancy. Each organization has its own dashboards, data sources, and users. A single Grafana instance can serve multiple teams without data overlap. The gotcha: organizations provide logical separation, not security isolation. For strict isolation, use separate Grafana instances.
Q5: What is the difference between Grafana alerting and Prometheus alerting?
Answer: Grafana alerting evaluates conditions in Grafana and sends notifications through Grafana's contact points. Prometheus alerting evaluates rules in Prometheus and sends alerts through Alertmanager. Grafana works with any data source. Prometheus works only with PromQL. In practice, use Prometheus alerting for application health and Grafana alerting for dashboard-centric alerts. The gotcha: running both can create duplicate alerts. Choose one and stick with it.
Q6: How do you optimize Grafana dashboard performance?
Answer: (1) Reduce the number of panels. (2) Increase the refresh interval (60s instead of 30s). (3) Narrow the default time range. (4) Simplify queries (aggregate early, use recording rules). (5) Use template variables to avoid loading all data. The gotcha: each panel runs a query on every refresh. 20 panels × 30s = 40 queries per minute. This adds up.
Q7: What is provisioning and why is it important?
Answer: Provisioning defines Grafana resources (data sources, dashboards, alerts) as YAML/JSON files that are loaded automatically on startup. This makes dashboards version-controlled, reproducible, and reviewable. Without provisioning, dashboards exist only in the Grafana UI and can be lost. The gotcha: provisioned resources can be modified in the UI, but changes are overwritten on restart. Always edit the files.