How to Create a Windows Containers Cluster

In this blog post, I’ll show you how I create a Windows Containers Cluster with two Container Hosts on Windows Server 2016 Core.

When productize Containers we should think about creating a cluster for HA reasons in case the host goes offline or need to be patched.

Creating a Windows Containers Cluster is not as hard as creating other Clusters like Exchange, SQL, etc, however, you will need to follow the steps below to get it working.

My recommendation is that any host participating in the Cluster should have a static IP address and must run Docker version 1.13.0 or later.

In this lab, I’m using two Windows Server Core 2016 Insider build running as virtual machines and the latest docker build.

Swarm Mode

In the Docker and Windows Containers, A cluster is called Swarm Mode and it combines features like Clustering, container orchestration capabilities and scheduling of container workloads.

When we group few Container Hosts together we form a “Swarm” cluster that allows us to use the three feature above.


A windows Containers cluster consist of two types of nodes:

Manager Nodes – This is the main cluster node that all management commands are run from, We can have more than one Manager Node in a Swarm cluster

Worker Nodes – Executed tasks that are assigned to the Manager nodes

Firewall Requirements

For Swarm to work the ports below need to be opened and or you can disable the Firewall using the PowerShell cmdlet below.

TCP port 2377 for cluster management communications

TCP and UDP port 7946 for communication among nodes

UDP port 4789 for overlay network traffic

I used the cmdlet below, to disable the Windows Firewall on my two Container Hosts.

Set-NetFirewallProfile -Profile Domain,Public,Private -Enabled false
Create a Swarm Cluster

To create a Swarm Cluster I run the cmdlet below from the host I would like to be the Manager Node.

docker swarm init --advertise-addr= --listen-addr

As you can see above once the cluster has been created it will return the token key which I will use to add Worker Nodes to the Swarm Cluster.

Add Worker Nodes

Using the token I have received in the previous command, I’ll add a Worker Node using the command below:

docker swarm join --token SWMTKN-1-4xls514s7de3d2jwwqj6nog0mmcs7j7m53zbi1lq44mkipo2zo-5sh77de2iqbb3olbornca2c9u

To add another Manager Node I’ll use the command below:

docker swarm join-token manager

To retrieve the token from the Manager Node run the command below:

docker swarm join-token worker -q

To view, all members of the Swarm Cluster and their role run the cmdlet below

Docker node ls

Overlay Network

Once my Swarm Cluster is done, I’ll create the network to support it:

docker network create --driver=overlay swarmnet

Create a Service

In Swarm mode, we create containers using a Service.

The Service can include a pre-configured image of a web server for example and the network we created before.

Note: Once we have created the image we should copy it to each of container host.

Below, I’ll create a Swarm Service with a Windows Nano Server 2016 Image (In my case I’ve not used pre-configured the image) and use the Overlay network I created before.

Docker service create -name webapp -endpoint-mode=dnsrr --network=swarmnet microsoft/nanoserver-insider

To view the service run the command below:

Docker service ls

By default, when creating a service, swarm will only deploy one container.

If we want to have a few containers running the service we can scale it using the line below.

Docker service scale webapp=4

To view the deployed Containers run the cmdlet below

Docker service ps webapp

To create more than one container with my app I’ll run the cmdlet below

Limitation and conclusion

Docker Swarm can be scaled to 1000+ nodes easily and I don’t think there is a point even finding out the maximum number because of the massive scale of the technology.

Swarm is an amazing cluster technology that makes the entire HA process easy to scale up and achieve high availability.

I strongly recommend using Swarm In any production environment as it will allow a more consistent behavior.

It will also give better performance of applications and services hosted in docker.

Categorized as Swarm

1 comment

  1. Did this ever worked for anyone?
    I was trying like 10 different variation for master and worker nodes(bare metal, VMware VMs, HyperV VMs), all of them running on Win Server 1709(tried pre 1709 and Win10 either) using windowsservercore as a base container. I never managed to get services up and running, they just keep on closing once up for a few seconds, faced dozens on different issues, fixed em and still faced something else. Sad.

Leave a comment

Your email address will not be published.

This site uses Akismet to reduce spam. Learn how your comment data is processed.