Setting up Docker Community Edition Swarm on Azure


I wanted to test as a Control Pane for Docker Swarm, but in order to do this I first need some infrastructure running in the cloud, and for this article I will focus on getting Docker Community Edition  (CE) to run on Azure.


  • Setup Docker CE as a Docker Swarm on Azure

I will be following this guide to setup Docker CE on Azure:

and this guide specifies that I need some prerequisites:

Access to an Azure account with admin privileges

Yep, got that. If not then create for free =>

SSH key that you want to use when accessing your completed Docker install on Azure

To avoid bloating my own developer laptop, I create my own workspace in a Container and work from there and keep the files on Bitbucket. That way I can move my workspace around on different computers without caring about snowflakes. The workspace setup used is shared here:

So if you do not have “ssh-keygen” installed then it is easy to start a linux Container and execute something like:

ssh-keygen -t rsa -b 4096 -C -f /app/ubuntu/geircode_19f93204_rsa

Or click on the “” and then “docker-compose.up.bat”. This will start a container and open up an interactive terminal. A shared volume is set up between Docker Host and the Container so that the files you see inside the Container on /app are the same files you are seeing on the Docker Host. Running the above command will create the certificate inside the /ubuntu folder which is also accessible on the Docker Host because of the shared volume.

Creating a “Service principal”

Execute something like:

docker run -ti docker4x/create-sp-azure docker_ce_19f93204 docker-ce-19f93204_rg westeurope

If the script fails for some reason, then navigate to Azure Portal to delete the Service Principal that gets created before running the script again.

root@40198a525e8f:/app# docker run -ti docker4x/create-sp-azure docker_ce_19f93204 docker-ce-19f93204_rg westeurope

info: Executing command login

/info: To sign in, use a web browser to open the page and enter the code XXXXXXXXX to authenticate.

info: Added subscription Visual Studio Premium med MSDN

info: Setting subscription "Visual Studio Premium" as default


info: login command OK

Using subscription xxxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx

info: Executing command account set

info: Setting subscription to "Visual Studio Premium" with id "xxx".

info: Changes saved

info: account set command OK

Creating AD application docker_ce_19f93204

Created AD application, APP_ID= xxxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx

Creating AD App ServicePrincipal

Created ServicePrincipal ID= xxxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx

Create new Azure Resource Group docker-ce-19f93204_rg in westeurope

info: Executing command group create

+ Getting resource group docker-ce-19f93204_rg

+ Creating resource group docker-ce-19f93204_rg

info: Created resource group docker-ce-19f93204_rg

data: Id: /subscriptions/ xxxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx /resourceGroups/docker-ce-19f93204_rg

data: Name: docker-ce-19f93204_rg

data: Location: westeurope

data: Provisioning State: Succeeded

data: Tags: null


info: group create command OK

Resource Group docker-ce-19f93204_rg created

Waiting for account updates to complete before proceeding ...

Creating role assignment for xxxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx scoped to docker-ce-19f93204_rg

info: Executing command role assignment create

+ Finding role with specified name

/data: RoleAssignmentId : /subscriptions/ xxxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx /resourcegroups/docker-ce-19f93204_rg/providers/Microsoft.Authorization/roleAssignments/ xxxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx

data: RoleDefinitionName : Contributor

data: RoleDefinitionId : xxxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx

data: Scope : /subscriptions/ xxxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx/resourcegroups/docker-ce-19f93204_rg

data: Display Name : docker_ce_19f93204

data: SignInName : undefined

data: ObjectId : xxxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx

data: ObjectType : ServicePrincipal



info: role assignment create command OK

Successfully created role assignment for xxxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx

Test login...

Waiting for roles to take effect ...

info: Executing command login

-info: Added subscription Visual Studio Enterprise


info: login command OK

Your access credentials ==================================================

AD ServicePrincipal App ID: xxxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx

AD ServicePrincipal App Secret: xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

AD ServicePrincipal Tenant ID: xxxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx

Resource Group Name: docker-ce-19f93204_rg

Resource Group Location: westeurope


Great success.. What now?

Click on “Deploy Docker Community Edition (CE) for Azure (stable)” and this will load a custom deployment on Azure.

Let’s fill in the details:

Click on “Purchase” and wait 3-4 minutes.

Listing the resources created:

The new VM’s are hiding inside the Virtual machine scale sets “swarm-xxx-vmss”. In order to save VM costs during testing setups, you can Deallocate them from there.

Deploy your app on Docker for Azure

Navigate to “Outputs” on the deployment:

Copy the URL of “SSH TARGETS” to a browser to get this:

Time to login into the Swarm manager. I will use my Container to do this, but first I need to copy the Certificate from the volume shared directory “/app” to somewhere else outside this directory. This is because when a volume is shared to Docker Host, all files inside will get new file modes, and SSH demands a very specific file mode for the certificate.

Trying to connect to the Swarm manager:

root@40198a525e8f:/ubuntu# ssh -i /ubuntu/geircode_19f93204_rsa docker@

ssh: connect to host port 22: Connection refused

Ok, so that did not work. Aha, each swarm manager has different port i.e. port 50000.

root@40198a525e8f:/ubuntu# ssh -i /ubuntu/geircode_19f93204_rsa -p 50000 docker@

The authenticity of host '[]:50000 ([]:50000)' can't be established.

RSA key fingerprint is ca:2b:5a:34:39:46:32:b5:f5:31:81:d9:68:ec:03:13.

Are you sure you want to continue connecting (yes/no)? yes

Warning: Permanently added '[]:50000' (RSA) to the list of known hosts.

Enter passphrase for key '/ubuntu/geircode_19f93204_rsa':

Welcome to Docker!


Cool. I have now logged into my first Docker Swarm in Azure. Or so I thought.

Run “docker info”

Why is “Swarm: inactive”?

Ok, apparently I need to add the Swarm workers manually via the Swarm manager through SSH. Why didn’t the Azure template just do this automatically? Oh well.

Connecting to your Linux worker nodes using SSH

I tried to configure SSH agent forwarding and got this error:

root@40198a525e8f:~$ ssh-add

Could not open a connection to your authentication agent.

According to, I need to start the ssh-agent first!

root@40198a525e8f:/ubuntu# eval `ssh-agent -s`

Agent pid 96

root@40198a525e8f:/ubuntu# ssh-add -L

The agent has no identities.

root@40198a525e8f:/ubuntu# ssh-add /ubuntu/geircode_19f93204_rsa

Enter passphrase for /ubuntu/geircode_19f93204_rsa:

Identity added: /ubuntu/geircode_19f93204_rsa (/ubuntu/geircode_19f93204_rsa)


root@40198a525e8f:/ubuntu# ssh -p 50000 -A docker@

Welcome to Docker!


Yay. My SSH agent is working.

So where do I find my Swarm workers IP?

Go to the resource list and find the vnet setup: i.e. “dockerswarm-vnet”.

swarm-manager000000:~$ ssh -A docker@

The authenticity of host ' (' can't be established.

RSA key fingerprint is SHA256:c260W6he0ppCfmik+oa7TN42K4/xfPigAK2VysCSe6U.

Are you sure you want to continue connecting (yes/no)? yes

Warning: Permanently added '' (RSA) to the list of known hosts.

Welcome to Docker!


Join the worker to the Docker Swarm:

swarm-worker000000:~$ docker swarm join --token SWMTKN-1-591r0uu76vnar02h8f16n1e2p5p0dbt2pzpm4g89o0zrymhn39-a9q5n6bv4svr3amjscmvp2x55

This node joined a swarm as a worker.

swarm-worker000000:~$ exit

Connection to closed.

swarm-manager000000:~$ docker node ls


py4y0q2jmk8jgi806nc0jofgc * swarm-manager000000 Ready Active Leader 18.03.0-ce

1qv3m227fz9n9mz3fg0trwhxt swarm-worker000000 Ready Active 18.03.0-ce



To test my swarm I will use

Login Swarm Manager node and run:

git clone

cd example-voting-app/

docker stack deploy --compose-file docker-stack.yml vote

But first I need to scale up my workers by increasing here:

And joining the Worker to the Manager like before. Now I have this:

NB: At this point I ran in several quirks:

  • First, at the moment the Manager also works as a Node. So if you try to create a Docker service, it will start up on the Manager Node and not be available on the LoadBalancer URL that is only connected to the Worker Nodes.
  • All network traffic to the Worker Nodes are blocked by default. You have to create a Rule for every Port that is going to be open for the internet.

Add a rule for each port that is going to be reachable from the internet.

Deploy the app:

swarm-manager000000:~/example-voting-app$ docker stack deploy --compose-file docker-stack.yml vote

Creating network vote_backend

Creating network vote_frontend

Creating network vote_default

Creating service vote_result

Creating service vote_worker

Creating service vote_visualizer

Creating service vote_redis

Creating service vote_db

Creating service vote_vote


Find the URL for the public Loadbalancer for the Worker nodes. In my case it was “dockerswarm-externalLoadBalancer-public-ip”:

Open the URL in a browser with the correct Port:


Conclusion: Setting up Docker Swarm in Azure this way had way too many unnecessary steps than I first thought. This should have rather been done in a declarative way such as i.e.: “I want 1 Manager, 2 Workers. Give it to me please, and connect the swarm to my local Docker CLI”.

Article is also available here =>

Install “Docker for Windows” in Azure Nested Virtualization and Debug in VS2017

July this year, Azure got some new interesting VM types that where it’s possible to run Virtual Machines inside each other. This is called nested virtualization which previously has only been possible on bare-metal machines.

Before we get started there is one prerequisite:

  • Azure account

Goals of this post (if the title was not enough):

  • Start a new VM with the nested virtualization
  • Install Visual Studio 2017
  • Install Docker for Windows
  • Debug a .NET CORE 2 service

Open Azure Portal to create a VM, and it looks like Azure already had an image ready:

Apparently nested virtualization is not yet available everywhere in the world:

Select your region to see if your region is supported. Since my region is Europa then only “West Europa” have support for nested virtualization.

When choosing a VM size, look for all VMs starting with “D” or “E” and ending with “_V3”.

My choice was the “D4_V3”:

By default, the Linux image from “Docker for Windows” use 2GB memory so 16GB memory on the Host should be plenty. Tips: Avoid the use of “Premium disk support” if you are just testing stuff because this will cost you a lot even though the VM is shut down and deallocated. The new “Auto-Shutdown” option is also nice.

Click “Create” and wait in intense suspense for deployment.

After a few minutes, the VM has started and it’s time for the fun stuff like “will it actually work out of the box??”.

So far so good. First, I want to find out what exactly how much comes “out of the box”. It turns out that Hyper-V is disabled by looking at the “Turn Windows Features on or off” list.

So, what happens if we enable this? Ok, click through the “Features” and tick the “allow destination computer to restart”. Install Hyper-V.

It will restart automatically after installing Hyper-V.

After rebooting the Task Manager is now showing that Virtualization is Enabled. Great.

If you do not find the newly installed Hyper-V Manager by searching for it, it’s because Windows have not yet indexed it. Just a bit annoying. Anyways, it will be somewhere here: C:\ProgramData\Microsoft\Windows\Start Menu\Programs\Administrative Tools.

Installing “Docker for Windows”.


Err, was it too good to be true? Hmm. Troubleshooting time!

Clicking “Reset to factory defaults” button, but got new error:

Restarting the VM to see if that will fix this, and behold!

“Docker for Windows” running on an Azure VM with nested virtualization.

Debug a .NET CORE 2 service with Visual Studio 2017

Visual Studio 2017 Community was included in the Azure VM image, but it probably did not include .NET CORE and Docker tooling.

Open “Visual Studio Installer” from the start menu and click on “modify”. To my surprise, Microsoft have installed everything😊

But the Azure image did not container .NET CORE 2, so we need to install that and restart Visual Studio:

Creating a new service:

Remember to tick “Enable Docker Support” and choose “Linux”.

Make sure the “docker-compose” is set to be “StartUp Project”.

Click on “ ” to start debugging.

Eventually you might get this popup. Click on “Share it” to enable access to your source files directly in the Container.

So nested virtualization in Azure together with “Docker for Windows” is definitely working!

Debugging is also working!

A few small bumps to get there, but they were trivial to fix. Great success.


Exploring Docker Cloud Swarm

This feature has been in beta in Docker Cloud for some time now, and I wanted to find out what all the fuss was about.


  • Create and Connect a Docker Swarm running in Azure
  • Run a simple web app in the Swarm


  • Azure account (Global Admin)
  • Docker Cloud account

The Docker guide to setup the Docker Cloud to Azure is pretty straightforward. Navigate to Docker Cloud and enable “Swarm mode”. Open “Cloud settings”.

The integration process is almost completely automatic. Just type in the Subscription Id, hit click, login Azure and give access. Tada!

The only ‘challenge’ that might arise is that the Azure account must be Global Admin. If you only have partial access, you will need to get in touch with the supreme owner of the Azure account to activate this integration between Docker Cloud and Azure.

Creating the swarm!

For some reason, I always think of beeeeeeesssss when I hear this.

Navigate to “Swarms” tab, hit “Create” and follow this guide:

The guide offered me no resistance.

Starting up!

When it is finished, click the swarm and get this:

Copy and paste into a shell. I recommend trying out Ubuntu for Windows; it is great for doing Linux stuff without an actual Linux instance. Running “docker ps” will now show you this:

The command/shell is now running against the new Swarm and running “docker ps” will now show you this:

It is also possible to get here without the copy and pasting of the docker run command. In Docker for Windows, right click the white whale at the bottom of the screen and expand the “Swarms” option. Click the swarm (in this case the “swarm-poc”) and it will open a CMD where the Docker CLI is connected to the Swarm.

Since Docker CLI is now running against the selected Swarm, any command now runs directly against the swarm and not the Docker Host running locally in Hyper-V.

To see the swarm running, run this command:

>> docker node ls

>> docker run -it –rm hello-world

This however will create a Hello World container running on the Manager, and that is not correct at all. To make run Hello World on one or more of the Workers we need to create a Docker Service.

Running Hello World in the Swarm

The commands to be used with Docker Swarm is listed here:

>> docker service create -p 13337:80 tutum/hello-world

The “Hello world” webapplication is now running in one of the workers and is reachable from the internet. You can get the public IP from either Docker Cloud or the Azure Portal:

In Azure portal, find the Resource Group and “externalLoadBalancer”:

Deallocate and save money

Deallocate the VMSS instances in Azure in order to save money. They do not show up in the Virtual Machine tab, but click on the Resource Group for the new swarm will show them. Remember to start them up again before using the swarm.

My three nodes (1 Manager, 2 Workers) does not cost much though. Yet. So far they only have a running fare at approx. 0.5 euros a day or 10€ (11$ or 101 NOK) for a month. Some people will call that inexpensive, low-priced, low-cost, economical, competitive, affordable, reasonable or free. At least Google Dictionary do. Except the last one.

In conclusion, we have now a Docker Swarm up and running about that can run Containers with workloads remotely instead of running them in your local Docker Host instance.