[Part I] IoT with docker

To use the correct tools to implement a task is so important. When playing with IoT applications, you will inevitably touch:

  • MQTT, for handling easily the messaging,
  • InfluxDB database, to store the data in a timestamped fashion (time-series arrays), and
  • Grafana, to visualise the data

This is good, but do you mean that I need to install everything from scratch? Ohh, that’s a lot.

Fortunately, everything is so easy. Just use docker containers and set up that chain a few minutes.

Do you want to start with me? Okay. Let’s start.

We need to install Docker by downloading and executing the convenience script in our ubuntu machine. Type in the terminal:

curl -sSL https://get.docker.com | sh

Then, we add our user (non-root user) to the docker group in order to run containers without using the sudo command.

sudo usermod -aG docker <username>

So replace <username> with the name of the user that you use. Finally, test that everything works as expected:

docker run hello-world

Great. Now it’s time for our MQTT broker (server) to be installed and become live. One popular option is to use mosquito by the eclipse foundation. Easy:

docker pull eclipse-mosquitto
docker run -it -p 1883:1883 -p 9001:9001 eclipse-mosquitto

In the second command, we run the container using the following switches:

  • -it: run it in interactive mode by allocating a tty for the container process
  • -p: expose the container’s internal port to the host port

Now, MQTT broker is up and running; you could use. Do you want to test it directly? Let me guide you.

First, we need to install mosquitto clients and use it directly from the terminal commands to publish and listen to topics.

sudo apt-get update
sudo apt-get install mosquitto-clients

So now, if we decide that we want to publish in a topic called “test” the temperature of the kitchen, we could type in terminal:

mosquitto_pub -m "10" -t "test"

To listen for messages in this specific topic, we need to run:

mosquitto_sub -t "test"

The switches in the commands are self-explanatory. If you are a GUI type of person, you could use MQTT.fx, download the deb package and install it, easy as that:

sudo dpkg -i ./mqttfx-1.7.1-64bit.deb

Now, we need to set the InfluxDB. We need to pull it from the hub and run it.

docker pull influxdb
docker run -d -p 8086:8086 -v influxdb:/var/lib/influxdb --name influxdb influxdb

In the run command, we execute the container in the detached mode (-d) and create and connect a docker volume to store data associated with the InfluxDB container (-v). Finally, we assign the container the influxdb name (–name).

Before we continue, we need to configure a database. How to do that in our case that influxDB runs in a detached mode? Easy. First, let’s get in the container with exec command

docker exec -it influxdb influx

Then, we create a database called “sensors”, we create a user “telegraf” and we give him full permissions in the newly created database.

create database sensors
use sensors
create user telegraf with password 'password'
grant all on sensors to telegraf

The database is there and we need to connect with the MQTT listening to the data in a specific topic. No, you do not need to write code for that. You just need to use telegraf. Let’s install it and configure it to listen in a specific MQTT topic and store the time series data directly to the database.

docker pull telegraf
docker run --rm telegraf telegraf config > telegraf.conf

Yeap, as you guessed, we need to edit the telegraf.conf file.

First, to remove the # for comment and enable the input

  servers = ["tcp://"]
  topics = ["sensors"]
  data_format = "influx"

and then similarly the output

  urls = [""]
  database = "sensors"
  skip_database_creation = true
  username = "telegraf"
  password = "password"

Yes, as you correctly guess, the IP address of the machine that I execute both MQTT broker and InfluxDB server is the same “” machine. The last step is to start the telegraf module like:

docker run -v /home/<username>/:/etc/telegraf:ro telegraf

To test the flow of data, we can publish a message and check the database to find out if the point stored there, as expected.

The command to publish a measurement into the topic is:

mosquitto_pub -h -m "temp,site=kitchen value=20" -t "test"

The measurement called “temp”, it is equal to 20 and it has a tag with the room location (i.e. kitchen)

Now, if we login again in the database using the docker exec command, we can list the measurements writing something like:

use sensors
select * from temp

Last but not least, we use again a docker container to set up Grafana and we run as usual

docker pull grafana/grafana
docker run -d -p 3000:3000 grafana/grafana