Table of Contents

• ROS Noetic Docker 1
• Docker Swarm
  • What is a docker Swarm?
  • How Docker Swarm Works?
  • How Run this Didactic Project?
    • Dependencies
    • Let's Start!
      • Starting the Environment
      • Cluster Overview
      • Deploying apps in the Cluster
    • What Else?

ROS Noetic Docker 1

That is a simple Docker that install the ros-noetic-ros-base environment and launches a roscore for demostrating that it works.

Docker Swarm

With this example, I pretend to simulate the performance of a cluster formed by different nodes. Before to start, I am going to explain... what is a docker swarm?

What is a docker Swarm?

You can find more info here but, briefly, Docker Swarm is a docker based technology that allow us to run a docker container in a distributed system formed by N nodes connected to each other. In this way, Docker Swarm provides a load balancing for the optimal deployment of the different tasks between all nodes.

How Docker Swarm Works?

Docker Swarm follows a master-slave architecture. So, if we have 4 nodes, one of then will be the master and the others will be the slaves. In the Docker terminology, master is named manager and slave is named worker.

In this way, we would have something like this:

---

First, I will tell how this project works and how reproduce its result.

How Run this Didactic Project?

Dependencies

• Virtual Box: You need the 5.2 version or older of Virtual Box. You can install it from here

• Vagrant: It is a virtual machines manager. You can install it with sudo apt install vagrant.

NOTE: If you don't have activated the virtualization in BIOS, that is the moment. Otherwise, you won't can start the virtual machine.

Let's Start!

Starting the Environment

• You need to download from Releases the following files: docker_swarm.box.aa, docker_swarm.box.ab and docker_swarm.box.ac. These files are only one file that was splitted because it was very heavy.

• We are going to get the .box file from the three files with the next command:

  $ cat docker_swarm.box.a* > docker_swarm.box
  

Now, we have the packaged virtual machine. We need to import it.

• Import the docker_swarm.box file with the next command:

  $ vagrant box add docker_swarm docker_swarm.box
  

With this command, we will have imported an image named docker_swarm.

• The next step is to create the virtual box from this image previously imported. For doing this, we need type the next commands:

  $ mkdir docker_swarm && cd docker_swarm
  $ vagrant init docker_swarm
  

  NOTE: Probably, we have to wait a little time

• Once the virtual box has been created, we are ready to run it:

  vagrant up --provider virtualbox
  

  NOTE: We only need to specify the flag '--provider virtualbox' the first time that we run it.

At this moment, we have the virtual machine running successfully. But... we want to enter on it to start working!

• The last step is to enter to the virtual machine typing vagrant ssh.

Now, we are inside the virtual box!

Cluster Overview

Before all, maybe you need to open more than one shell with vagrant ssh.

If we type the command lxc list we well see three lxd containers:

• master
• esclavo1
• esclavo2

Each of these containers is simulating a node of the cluster where master is the manager node and esclavo1 and esclavo2 are the workers nodes.

The first thing we have to do is run all nodes (lxd containers). For doing this, we need to type the following command:

$ lxc start --all

After waiting a few seconds, if we type lxc list again, we will see that all containers are now running. So, we should have something like that:

We can see that IP addresses on eth0 interfaces are:

• master: 10.222.123.10
• esclavo1: 10.222.123.11
• esclavo2: 10.222.123.12

It is very important that nodes are connected each other, so this is the next thing we are going to check:

NOTE: we can enter to the lxd containers typing master, esc1 or esc2 in each case.

• Enter to the master lxd container (master node) typing master. Then we are going to check the connection with the two workers:

  $ ping esclavo1
  

  Passed a few seconds, press CTRL+C. The output should be something like that:

  

  A we can see, in this example, three packets were transmitted and three packets were received. Any package lost. That means that the connection works successfully.

  In the same way, check the connection with the second worker:

  $ ping esclavo2
  

  Now, exit from master pressing CTRL+D and check the connection with master and esclavo2 from esclavo1 and the connection with master and esclavo1 from esclavo2. Remember! The commands are:

  $ ping master
  $ ping esclavo1
  $ ping esclavo2
  

---

Arrived to this point, I recommend open three terminals, and enter in one node in each one.

Also, we are going to be working with Docker so... we need root privileges. Then, we need to type sudo su inside each node.

But... **Where is the Cluster?!**

Go to the master node! Now, type the next command:

$ docker node list

We should are seeing something like that:

docker node list informs us that there is three nodes (all are activated), their host names (esclavo1, esclavo2 and master), and it also indicates us which is the manager node (Leader). As we can see, master is the manager.

So, Cluster is already deployed. Now, we are going how it works running a Telegram Bot. If you dont have a Telegram bot, you can create it so fast following this instructions.

NOTE: It is very important to save the token

Deploying apps in the Cluster

• I have prepared a docker that runs a Telegram bot but, to make it work, you must paste your Telegram bot Token in the file /home/ubuntu/telegram_bot/context/token.txt.

  NOTE: You have to do this in the three nodes.

• Now, let's to create the Docker image in the three nodes. It is very important that the three nodes have the image because, otherwise, the manager will not be able to distribute the task between them. To create the image, go to /home/ubuntu/telegram_bot folder. Then, run:

  $ ./build.sh
  

  NOTE: You have to do this in the three nodes.

Wait the Docker image build finish at three nodes.

• Now, we can launch the Telegram bot. For doing this, go to the master node and type:

$ docker service create --name telegram_bot fgonzalezr1998/telegram_bot

We have to wait for a few seconds.

• If we type on master docker service list we will see the telegram_bot process running. If we type docker service ps telegram_bot We can see which node is running it. Here there is an example:

As we can see, in this example, we have created only one replica, but, we can scale the service:

$ docker service scale telegram_bot=3

Run this command in the master node will create three replicas of the service.

You can try to run docker ps in each node and see what is shown. You also can try to run again the previous commands in the master.

Also we have reduce the number of replicas:

$ docker service scale telegram_bot=2

master node decide which worker execute each task.

But... What is exactly to scale a docker service? Here you have more information.

• To conclude... try to send messages to your telegram Bot and see how it answer!

NOTE: To remove the service, type docker service rm telegram_bot

What Else?

Now, you can try to create your own Docker image to run other tasks. Also you can add more lxd containers as workers (esclavo3, esclavo4, etc...).