Skip to main content
In this step-by-step guide, you will learn how to bootstrap a resilient Docker Swarm cluster on CentOS 7.6, manage node roles, control availability, and gracefully remove nodes. By the end of this tutorial, you’ll have a three-node swarm (one manager, two workers) and know how to:
  1. Initialize the swarm
  2. Join worker nodes
  3. Promote and demote managers
  4. Drain and reactivate nodes
  5. Remove nodes from the cluster
This tutorial assumes all nodes can resolve each other by hostname, have open swarm ports, and run Docker Engine v19.03.8 or later.
  • Three CentOS 7.6 servers (managerone, workerone, workertwo)
  • 2 CPU cores and 4 GB RAM each
  • Hostname resolution (e.g., via /etc/hosts or DNS)
  • Open ports:
    • TCP 2377 for cluster management
    • TCP/UDP 7946 for node discovery
    • UDP 4789 for overlay networking
  • Docker Engine installed and running

​
System Overview

NodeOS ReleaseCPUsMemory (MiB)Docker Version
manageroneCentOS Linux release 7.6.1810 (Core)2378719.03.8
workeroneCentOS Linux release 7.6.1810 (Core)2378719.03.8
workertwoCentOS Linux release 7.6.1810 (Core)2378719.03.8

​
1. Initialize the Swarm on managerone

  1. Confirm Swarm is inactive: 
    docker system info | grep -i swarm
# Swarm: inactive
  2. Initialize with the manager’s advertise address: 
    docker swarm init --advertise-addr 172.31.42.232
    After initialization, note the worker join command output: 
    docker swarm join --token <SWARM_WORKER_TOKEN> 172.31.42.232:2377

​
2. Add the First Worker (workerone)

On workerone, run the join command displayed by the manager: 
docker swarm join \
  --token SWMTKN-1-3fdj9fgrjcrrj5t0pekb42n45tj96zgwxodtwd4ujv4qnhl-cop40spyhgc1tmzythfss49xn \
  172.31.42.232:2377
# This node joined a swarm as a worker.
Back on managerone, verify: 
docker node ls
# ID       HOSTNAME    STATUS  AVAILABILITY  MANAGER STATUS  ENGINE VERSION
# ...      managerone  Ready   Active        Leader          19.03.8
# ...      workerone   Ready   Active                        19.03.8

​
3. Add the Second Worker (workertwo)

​
3.1. Prepare workertwo

Ensure OS, CPU, memory, Docker version, and connectivity: 
cat /etc/centos-release
nproc
free -m
docker --version
systemctl status docker
ping -c3 managerone
ping -c3 workerone

​
3.2. Retrieve and Run the Join Token

On managerone, print the worker join command: 
docker swarm join-token worker
Copy the docker swarm join --token … line and execute it on workertwo. Then confirm on managerone: 
docker node ls
# ... workertwo   Ready   Active   19.03.8

​
4. Promote a Worker to Manager

  1. List current nodes: 
    docker node ls
  2. Promote workerone: 
    docker node promote workerone
# Node workerone promoted to a manager in the swarm.
  3. Verify the new manager status: 
    docker node ls
# workerone now shows MANAGER STATUS: Reachable
  4. (Optional) Inspect workerone in detail: 
    docker node inspect workerone --pretty

​
5. Demote a Manager back to Worker

If you need to revert workerone to a purely worker role: 
docker node demote workerone
docker node ls
# workerone returns to no MANAGER STATUS

​
6. Drain and Reactivate a Node

To prevent new tasks from scheduling on workerone: 
docker node update --availability drain workerone
docker node ls
# AVAILABILITY: Drain
When you’re ready to allow tasks again: 
docker node update --availability active workerone
docker node ls
# AVAILABILITY: Active

​
7. Remove a Node from the Swarm

  1. Drain workerone: 
    docker node update --availability drain workerone
  2. Attempt removal (will fail if the node is still up): 
    docker node rm workerone
# Error: node is not down and can't be removed
  3. On workerone, leave the swarm: 
    docker swarm leave
# Node left the swarm.
  4. Back on managerone, remove the node: 
    docker node rm workerone
docker node ls
# Only managerone and workertwo remain.
Congratulations! You’ve successfully created, scaled, and managed a Docker Swarm cluster on CentOS 7.6.

​
References and Further Reading

Watch Video