Grafana · Monitoring · Indoor farming · 15 October 2026
Grafana Dashboards for Real-Time Indoor Farming Monitoring
Twenty-seven sensor nodes, actuator states, and a local time-series database become an operational view that helps researchers see what is happening, where it is happening, and whether the climate system is responding.
Albert Sagala · Toba Smart Farming
Monitoring architecture
The dashboard is the last layer of a local-first system. Sensing and control continue at the facility even when remote access is unavailable.
27 sensor nodes
Temperature and humidity are measured across horizontal and vertical zones.
Jetson Nano
Fresh observations support local climate-control decisions.
Local database
Sensor histories and actuator states are retained as time-series data.
Grafana
Panels combine current conditions, trends, zones, equipment, and alarms.
Cloudflare Tunnel
Authorized users reach the dashboard without exposing the local database.
Start with operational questions—not decorative charts
A dashboard is useful only when it shortens the path from a problem to a decision. The first screen should answer whether the climate is safe, which zone is abnormal, whether the AC or humidifier is operating, how many sensors are online, and when the latest data arrived.
Use colour consistently: normal, approaching a limit, and violation. A dense wall of gauges may look advanced while hiding the one condition that requires action.
Show both the average and the spatial variation
A container farm is compact but not necessarily uniform. Airflow, racks, lamps, cooling distance, and crop density can produce different conditions at different locations. A single average may appear acceptable while the upper rack or one end of the container is overheating.
Group the 27 nodes by left, middle, and right zones and by bottom, middle, and top layers. Display the mean, minimum, maximum, spread, and missing-node count. This turns the dashboard from a room thermometer into a spatial diagnostic tool.
Put actuator states beside climate trends
Temperature without AC status tells only half the story. Overlay or align actuator timelines with temperature, humidity, and CO₂ trends so users can see when equipment switched, how quickly the environment responded, and whether excessive cycling or overshoot occurred.
Record commands and acknowledgements separately when possible. A command marked ON is not proof that a relay, compressor, pump, or solenoid actually operated.
Treat sensor health as a first-class panel
Missing data must never be silently converted to zero. Show the last-seen time, message gaps, implausible jumps, values that remain frozen too long, and battery voltage when available. Mark unavailable observations as no data.
This distinction prevents a communication failure from being mistaken for an extreme environmental event and makes research records easier to audit.
Alert on conditions that require action
Useful alerts include prolonged temperature or humidity violations, a sensor that stops reporting, excessive differences between zones, and an AC that is active without a corresponding temperature response. Each alert needs a priority, timestamp, affected device or zone, and a first diagnostic action.
Avoid alarm fatigue. A rule that fires repeatedly without requiring action should be redesigned, delayed, grouped, or removed.
Use Grafana for visibility, not as the scientific result
Grafana supports exploration, daily operation, and early fault detection. It can also display controller metrics such as setpoint error, compliance time, actuator duty, energy use, overshoot, and settling time.
Publication results should still come from a reproducible analysis pipeline with documented queries, cleaning rules, time windows, and statistical methods. A compelling screenshot is evidence of a dashboard—not proof of a scientific conclusion.
Secure the path from the internet to the dashboard
Do not expose Grafana, the message broker, or the database directly to the public internet. Cloudflare Tunnel provides an outbound-only path from the facility, but the application still needs authenticated access, strong passwords, updates, audit logs, and least-privilege roles.
Monitoring accounts should be read-only by default. Remote control, if introduced, belongs behind stronger authorization, safe command limits, local interlocks, and a complete audit trail.
Recommended dashboard panels
Current temperature, humidity, CO₂, and last update
Minimum, maximum, mean, and spatial spread
Left–middle–right and bottom–middle–top comparison
AC, humidifier, CO₂ solenoid, and lighting status
Offline nodes, stale values, and message gaps
Setpoint compliance and violation duration
Daily actuator runtime and electrical energy
Prioritized alerts with a first diagnostic action