How the automatic scaling works in docker swarm

Implementing Docker Swarm involves setting up a cluster of Docker nodes that can manage multiple services and scale horizontally to handle increased workloads. Below are the steps to implement Docker Swarm involving two services, and an explanation of how to specify multiple nodes, handle scalability, and manage node crashes:

Steps to Implement Docker Swarm

1. Initialize Docker Swarm

First, you need to initialize Docker Swarm on one of your nodes. This node will act as the Swarm manager.

docker swarm init --advertise-addr <MANAGER-IP>

• Replace <MANAGER-IP> with the IP address of the node that will serve as the Swarm manager.
• This command initializes a new Swarm and generates a join token that other nodes can use to join the Swarm.

2. Join Worker Nodes

To add worker nodes to the Swarm, use the join token generated during initialization. Run the following command on each worker node:

docker swarm join --token <TOKEN> <MANAGER-IP>:2377

• Replace <TOKEN> with the join token obtained from the docker swarm init command output.
• Replace <MANAGER-IP> with the IP address of the Swarm manager node.

3. Deploy Services

Once your Swarm cluster is set up, you can deploy services. A service in Docker Swarm represents a scalable application, such as a microservice or a web server.

docker service create --name service1 --replicas 3 -p 8080:8080 my-service1-image
docker service create --name service2 --replicas 2 -p 9090:9090 my-service2-image

• --name: Specifies the name of the service.
• --replicas: Specifies the number of replicas (instances) of the service to run across the Swarm. Docker Swarm handles distributing these replicas across available nodes.
• -p: Maps a container port to a host port for accessing the service externally.
• my-service1-image, my-service2-image: Docker images used to create containers for service1 and service2.

4. Scaling Services

You can scale services up or down dynamically to meet changing demand:

docker service scale service1=5

This command scales service1 to 5 replicas. Docker Swarm automatically distributes these replicas across the available worker nodes.

5. Handling Node Failures

Docker Swarm handles node failures and ensures high availability of services:

• Automatic Recovery: When a node fails, Docker Swarm automatically reschedules tasks (containers) from the failed node to other healthy nodes in the Swarm.

• Fault Tolerance: If a container or node fails, Docker Swarm ensures that the desired number of replicas (specified during service creation) are always running, maintaining service availability.

Example Scenario: Scalability and Node Crash

Let’s illustrate how Docker Swarm handles scalability and node crashes in a practical scenario:

• Initial Setup: You have a Docker Swarm with one manager node (Node A) and two worker nodes (Node B and Node C).

• Deploy Services:

  docker service create --name webapp --replicas 3 -p 80:80 my-webapp-image
  

  This command deploys a web application (webapp) with 3 replicas across the Swarm.

• Scalability: As traffic increases, you decide to scale the webapp service to 5 replicas:

  docker service scale webapp=5
  

  Docker Swarm automatically distributes the additional replicas (4th and 5th replicas) across Node B and Node C, balancing the workload.

• Node Crash: Suppose Node B experiences a hardware failure and goes offline:

  • Docker Swarm detects the node failure and marks Node B as unreachable.
  • Tasks (containers) running on Node B are rescheduled to other available nodes (Node A and Node C) to maintain the desired replica count (5 instances).

• Recovery: Once Node B is back online or replaced with a new node, Docker Swarm automatically resynchronizes and reschedules tasks to restore the desired state.