Temperature monitoring with Grafana and InfluxDB

Exploring DIY temperature monitoring

Date
2024-07-18
Tags
pi, influxdb, grafana
Length
3 min

I’ve been eager to work on a project involving time series databases. With summer heating up the northern hemisphere, I recently had the idea to create a small project at home to monitor the temperature of the two Pis I have.

Since the configurations of the two devices I have are different (Pi4 - Pi Zero), I planned to write the script that measures temperature data in shell script to find a common ground. This shell script would run every minute and send the results to an InfluxDB on a remote machine.

As I had never written a shell script before, I turned to our dear friend GPT for help. After a few attempts, I managed to obtain the exact script I wanted.

#!/bin/bash

# Determine the node name based on hostname
NODE=$(hostname)

# InfluxDB settings
INFLUX_URL="your_influx_db_host"
INFLUX_TOKEN="your_influx_token"

# Function to get CPU temperature
get_cpu_temp() {
  vcgencmd measure_temp | grep -oP '\d+\.\d+'
}

# Function to get GPU temperature (if available)
get_gpu_temp() {
  vcgencmd measure_temp | grep -oP '\d+\.\d+' 2>/dev/null || echo "N/A"
}

# Get CPU temperature
CPU_TEMP=$(get_cpu_temp)

# Get GPU temperature
GPU_TEMP=$(get_gpu_temp)

# Epoch timestamp in nanoseconds
EPOCH_NANO=$(date +%s%N)

# Print date and time
echo "Current date and time:"
echo "----------------------"
echo "$(date)"

# Construct data to send to InfluxDB
DATA="temp,node=${NODE} cpu_temperature=${CPU_TEMP} ${EPOCH_NANO}"
if [ "$GPU_TEMP" != "N/A" ]; then
  DATA+=$'\n'"temp,node=${NODE} gpu_temperature=${GPU_TEMP} ${EPOCH_NANO}"
fi

# Print the data to be sent
echo -e "\nData to be sent:"
echo "----------------"
echo -e "$DATA"

# Send data to InfluxDB
echo -e "\nSending data to InfluxDB..."
curl --request POST \
  --url $INFLUX_URL \
  --header "Authorization: Token ${INFLUX_TOKEN}" \
  --header "Content-Type: text/plain; charset=utf-8" \
  --header "Accept: application/json" \
  --data-binary "$DATA"

echo "Data successfully sent to InfluxDB."
Current date and time:
----------------------
Thu Jul 18 13:49:09 +03 2024

Data to be sent:
----------------
temp,node=pi4 cpu_temperature=53.5 1721299749389821197
temp,node=pi4 gpu_temperature=53.5 1721299749389821197

Sending data to InfluxDB...
Data successfully sent to InfluxDB.

The script is running smoothly 😄

Let’s add the script to cron jobs to run once every minute.

crontab -e

Let’s use the nano editor to append the following line to the end of the opened file.

* * * * * /bin/bash /home/kaan/scripts/temp_info.sh >/dev/null 2>&1

When I check the cron logs, I can see that our file is running even though there are no error or success messages.

journalctl _COMM=cron | grep temp_info
Jul 17 12:25:01 pi4 CRON[2098647]: (kaan) CMD (/bin/bash /home/kaan/scripts/temp_info.sh >/dev/null 2>&1)
Jul 17 12:26:01 pi4 CRON[2098707]: (kaan) CMD (/bin/bash /home/kaan/scripts/temp_info.sh >/dev/null 2>&1)
Jul 17 12:27:01 pi4 CRON[2098785]: (kaan) CMD (/bin/bash /home/kaan/scripts/temp_info.sh >/dev/null 2>&1)

Let’s check the data flow in InfluxDB

InfluxDB Data Explorer showing temperature readings arriving from the Pis
yayyy..

Everything looks good. Let’s move on to the visualization part.

Data visualization with Grafana

Let’s add Influx, our data source, to Grafana and write a few queries using Flux.

Grafana panel editor with a Flux query for the average CPU temperature
Average temperature data
from(bucket: "home_lab_bucket")
  |> range(start: v.timeRangeStart , stop: v.timeRangeStop)
  |> filter(fn: (r) => r._measurement == "temp")
  |> filter(fn: (r) => r.node == "pi4")
  |> filter(fn: (r) => r._field == "cpu_temperature")
  |> mean()

The part that gave me a bit of trouble here was adding the selected time range information from Grafana to the query. The variables v.timeRangeStart and v.timeRangeStop seamlessly incorporated the selected date-time information from Grafana into the query.

Let’s add a few useful panels to our Grafana dashboard.

Grafana time-series panel of temperature readings
Temperature data recorded over time axis
from(bucket:"home_lab_bucket")
    |> range(start: v.timeRangeStart , stop: v.timeRangeStop)

With a few hours of effort, I successfully created a simple dashboard just as I wanted.

The finished Grafana dashboard: average gauges for both Pis, a combined temperature graph and a CPU graph for each

Until the next post, see you soon!