Master Docker Swarm 1.12: Build, Scale, and Manage High‑Availability Clusters

Background

Docker has become familiar to many due to agile deployment, and Docker 1.12 introduced many features such as Swarm mode, health checks, node identity encryption, Service API, constraint filtering, built‑in routing, and multi‑platform support.

Previously Docker lacked a true cluster management solution; Docker 1.12 embeds Swarm mode for multi‑host, multi‑container orchestration.

Swarm Cluster New Features

1. Batch Service Creation

Docker 1.12 adds the docker service command, which can manage containers across multiple hosts. Example with one manager and five workers:

docker network create -d overlay mynet

docker service create --replicas 3 --name frontend --network mynet --publish 80:80/tcp frontend_image:latest

docker service create --name redis --network mynet redis:latest

First an overlay network is created to enable inter‑host communication, then three frontend replicas and one Redis replica are launched on the same overlay network.

2. High Availability

If two of the three frontend nodes fail, Swarm detects the desired replica count (3) and automatically starts the missing replicas on remaining nodes, maintaining service health.

3. Service Scalability

Scaling is simple: docker service scale frontend=6 creates additional replicas instantly. To run one replica on every node, use global mode:

docker service create --mode=global --name extend_frontend frontend_image:latest

4. Scheduling Constraints

Constraints allow placement on nodes matching criteria, e.g., only on SSD‑equipped hosts:

docker service create --replicas 3 --name frontend --network mynet --publish 80:80/tcp --constraint engine.labels.com.example.storage=ssd frontend_image:latest

5. Building a Swarm Cluster

Prepare three machines (Node1 192.168.133.129, Node2 192.168.133.137, Node3 192.168.133.139) and open ports 2377/tcp, 7946/udp, 4789/udp. Initialize the manager: docker swarm init Join workers: docker swarm join --token <TOKEN> 192.168.133.129:2377 Verify nodes with docker node ls.

6. Demo Service

A simple Python Flask app (env.py) returns the container’s hostname:

from flask import Flask

import os

app = Flask(__name__)

@app.route("/")

def env():

return os.environ["HOSTNAME"]

app.run(host="0.0.0.0")

Build the image and create a service:

docker service create --name test -p 5000:5000 demo python env.py

Use docker service ps to see task placement; Docker schedules tasks on the node with the lowest load.

7. Scaling the Demo

Scale to six replicas: docker service scale test=6 The three nodes each run two replicas, providing a natural HA cluster.

8. Simulating Node Failure

Stopping Docker on one node causes its tasks to be rescheduled on the remaining nodes automatically, preserving load balancing.

9. Load Balancing and Service Discovery

Docker’s built‑in discovery uses an overlay network and a distributed key‑value store. Services are reachable via internal DNS (e.g., test:5000) and IPVS provides layer‑4 load balancing.

Example Nginx reverse proxy configuration:

server {

listen 80;

location / {

proxy_pass http://test:5000;

}

}

10. External Service Discovery with Consul

Docker can use an external KV store such as Consul. After running Consul on a node, start registrator to register containers, then generate Nginx upstream configuration from Consul data.

11. Rolling Updates

Swarm supports rolling updates with controlled parallelism and delay:

docker service update --update-parallelism 2 --image demo:2.0 --update-delay 10s test

If an update fails, Docker rolls back to the previous task version.

Conclusion

Docker 1.12’s Swarm mode adds powerful cluster management, high availability, scaling, and service discovery features, making Docker suitable for production‑grade container orchestration.