Docker Swarm: A Complete Guide for Beginners

Docker is a go-to choice for application deployment for ALL the distributed systems. But, managing the application after deployment is as essential as deploying it. Docker Swarm is one such management tool developed by Docker Inc. In this blog, I will not only explain what is Docker Swarm, but I will also walk you through the topics mentioned below.

Let me give a quick introduction about Docker to get everybody on the same page.

What are Docker and Docker Container?

Docker is an open-source tool used to automate application deployment in the form of lightweight containers. And Docker containers are not only lightweight but also make the application platform-independent. So, the application that runs on your computer will run the same way on your friend’s computer!

The Docker Containers overcome a lot of problems faced by the virtual machines. Hence, the Docker Container wins the game of Docker Container vs. Virtual Machines any day! To enumerate, Docker Containers are faster, portable, provide isolation, use less memory, etc.

But just the Docker containers cannot do everything independently, and that is where an orchestrator comes in. So, let’s check out how Docker Swarm helps better management of Docker containers.

What is Docker Swarm?

Docker Swarm is native to the Docker platform. The Docker Swarm is integrated into Docker itself, and the Docker Swarm is installed when you install Docker.

Docker Swarm is a small-scale management tool that provides a good orchestration system for Dockerized applications. Docker Swarm turns a pool of Docker hosts into a single virtual server allowing clustering with the built-in Swarm orchestration. In essence, we can utilize the Docker Swarm model to efficiently deploy, manage, and scale a cluster of nodes on Docker.

Features of Docker Swarm

Have a look at some of the prominent features of the Docker Swam mode, and why do we need it:

• Simple and Fast: The Swarm mode offers simplicity and speed for development and deployment.
• Decentralized access: Swarm makes it pretty easy for organizations to enter and manage the environment
• High security: The communication between the manager and a worker nodes is highly secure within the swarm
• Autoload balancing: We also have an autoload balancing option within your environment, and you can write your own script for this.
• High scalability: With load balancing comes high scalability.
• Roll-back: It also allows you to roll back environments to previous safe environments

Docker Swarm Mode Concepts

So, whenever I refer to ‘Docker Swarm’, I was actually referring to Docker Swarm Mode but not the deprecated Docker Swarm. Let’s look at some of the key concepts of swarm mode:

Service and Tasks

We use services to launch our Docker containers. And they can be deployed in either global or replicated ways.

• Global services are used to monitor containers that want to run on a swarm node.
• Whereas a Replicated service is to specify the number of similar tasks that we require on a host machine.
• Scaling becomes effortless with Docker Swarm mode.
• Services can be accessed by any of the nodes in the same cluster.
• A service is a description of a task, whereas a task performs the work.
• A developer creates a service with Docker, which can start tasks. However, when a task is allotted to a node, the same task can’t be assigned to another node.

Node

• A node in  Swarm mode is an instance of the Docker engine.
• It is possible to run multiple nodes on a single server. But in production, nodes are spread across various devices.

The architecture and Working of Docker Swarm Mode

The idea of how you make use of Docker Swarm is simple: you declare the applications as stacks of services and let Docker handle the rest. The key components of a Docker Swarm are Docker Nodes, Docker Services, and Docker Tasks.

There are two kinds of Docker Nodes, the Manager Node, and the Worker Node. And a Swarm consists of at least one node, be it physical or virtual machines running Docker version 1.12 or later.

Manager Node

The manager node, as the name suggests, is responsible for the cluster management tasks:

• To maintain the cluster state
• Scheduling of Services
• And serving Swarm mode HTTP API endpoints.

But, for persistence, we use something called Raft implementation. The managers maintain a consistent internal state of services and the whole Swarm cluster itself. The best practice in production/non-testing purposes we should implement more than one manager node.

Worker Node

The Worker nodes are nothing but the instances of Docker Engine which are created to run Docker Containers. They are not involved in managing tasks. A manager node can exist without a worker node; a worker cannot exist without a manager. A manager node is also a worker node by default.

The worker nodes talk to the manager node via API over HTTP. As said earlier, services can be deployed and accessed by any node of the same cluster. So, while creating a service, you’ll have to specify the container image. You can set up commands and services to either global or replicated: a global service will run on every Swarm node, and on a replicated service, the manager node assigns tasks to worker nodes.

Now comes the most interesting section. I will show you how to create a cluster, upgrade nodes, start a service and, much more!

Demo on Docker Swarm for Beginners

Now comes the most interesting part, hands-on with Docker Swarm! I will show you how to set up a Swarm cluster with 1 manager node and 2 worker nodes. Check out the prerequisite to get started:

1. Three Ubuntu (or any other Linux distros) virtual machines or
2. Three Amazon AWS ec2 instances (I will be using ec2)
   Know how to create a free AWS account.
3. Interest to try out new things 😇

Step 1:  Install Docker

Install Docker on all three instances or VMs with the following command

curl -fsSL https://get.docker.com -o get-docker.sh
sh get-docker.sh

You can check if it is installed by the below command:

docker --version

Step 2: Create a Swarm Manager

We will have to create a manager to handle our worker nodes. And, you should also know that the manager is also a worker node with some special privileges.

docker swarm init --advertise-addr [private or public ip address]

Use private IP addresses if all the instances/VMs are in the same network, if not you should be using the pubic IP instead.

Note: Make sure you copy the above-selected command which will be required in the next step.

Step 3: Add Worker Nodes

The next step is to join our two worker nodes to the Swarm cluster by using the token which was generated earlier.

docker swarm join --token <token-id> <ip:port>

Worker 1:

Worker 2:

Step 5: Check the Nodes

Now, let us check the status of our Swarm cluster. Perform the command below.

docker node ls

Hurray! Mission accomplished. We have successfully created a Swarm cluster with 1 Manager node and 2 Worker Nodes. Wait no more and start deploying your applications on the swarm cluster.