Instructions for setting up a Kubernetes cluster.

Edit This Page

Setting up Kubernetes with Juju

Ubuntu 16.04 introduced the Canonical Distribution of Kubernetes, a pure upstream distribution of Kubernetes designed for production usage. This page shows you how to deploy a cluster.

Before you begin

Deployment overview

Out of the box the deployment comes with the following components on 9 machines:

The Juju Kubernetes work is curated by the Big Software team at Canonical Ltd, let us know how we are doing. If you find any problems please open an issue on our tracker so we can find them.

Support Level

IaaS Provider Config. Mgmt OS Networking Docs Conforms Support Level
Amazon Web Services (AWS) Juju Ubuntu flannel, calico* docs   Commercial, Community
OpenStack Juju Ubuntu flannel, calico docs   Commercial, Community
Microsoft Azure Juju Ubuntu flannel docs   Commercial, Community
Google Compute Engine (GCE) Juju Ubuntu flannel, calico docs   Commercial, Community
Joyent Juju Ubuntu flannel docs   Commercial, Community
Rackspace Juju Ubuntu flannel docs   Commercial, Community
VMWare vSphere Juju Ubuntu flannel, calico docs   Commercial, Community
Bare Metal (MAAS) Juju Ubuntu flannel, calico docs   Commercial, Community

For support level information on all solutions, see the Table of solutions chart.

Configure Juju to use your cloud provider

Deployment of the cluster is supported on a wide variety of public clouds, private OpenStack clouds, or raw bare metal clusters. Bare metal deployments are supported via MAAS.

After deciding which cloud to deploy to, follow the cloud setup page to configure deploying to that cloud.

Load your cloud credentials for each cloud provider you would like to use.

In this example

juju add-credential aws
credential name: my_credentials
select auth-type [userpass, oauth, etc]: userpass
enter username: jorge
enter password: *******

You can also just auto load credentials for popular clouds with the juju autoload-credentials command, which will auto import your credentials from the default files and environment variables for each cloud.

Next we need to bootstrap a controller to manage the cluster. You need to define the cloud you want to bootstrap on, the region, and then any name for your controller node:

juju update-clouds # This command ensures all the latest regions are up to date on your client
juju bootstrap aws/us-east-2

or, another example, this time on Azure:

juju bootstrap azure/centralus

You will need a controller node for each cloud or region you are deploying to. See the controller documentation for more information.

Note that each controller can host multiple Kubernetes clusters in a given cloud or region.

Launch a Kubernetes cluster

The following command will deploy the initial 9-node starter cluster. The speed of execution is very dependent of the performance of the cloud you’re deploying to:

juju deploy canonical-kubernetes

After this command executes the cloud will then launch instances and begin the deployment process.

Monitor deployment

The juju status command provides information about each unit in the cluster. Use the watch -c juju status --color command to get a real-time view of the cluster as it deploys. When all the states are green and “Idle”, the cluster is ready to be used:

juju status


Model    Controller     Cloud/Region   Version
default  aws-us-east-2  aws/us-east-2  2.0.1

App                    Version  Status       Scale  Charm                  Store       Rev  OS      Notes
easyrsa                3.0.1    active           1  easyrsa                jujucharms    3  ubuntu  
etcd                   3.1.2    active           3  etcd                   jujucharms   14  ubuntu  
flannel                0.6.1    maintenance      4  flannel                jujucharms    5  ubuntu  
kubeapi-load-balancer  1.10.0   active           1  kubeapi-load-balancer  jujucharms    3  ubuntu  exposed
kubernetes-master      1.6.1    active           1  kubernetes-master      jujucharms    6  ubuntu  
kubernetes-worker      1.6.1    active           3  kubernetes-worker      jujucharms    8  ubuntu  exposed
topbeat                         active           3  topbeat                jujucharms    5  ubuntu  

Unit                      Workload     Agent  Machine  Public address  Ports            Message
easyrsa/0*                active       idle   0                      Certificate Authority connected.
etcd/0                    active       idle   3    2379/tcp         Healthy with 3 known peers.
etcd/1*                   active       idle   4    2379/tcp         Healthy with 3 known peers. (leader)
etcd/2                    active       idle   5    2379/tcp         Healthy with 3 known peers.
kubeapi-load-balancer/0*  active       idle   7    443/tcp          Loadbalancer ready.
kubernetes-master/0*      active       idle   8   6443/tcp         Kubernetes master services ready.
  flannel/3               active       idle                      Flannel subnet
kubernetes-worker/0*      active       idle   9                    Kubernetes worker running.
  flannel/2               active       idle                      Flannel subnet
kubernetes-worker/1       active       idle   10                     Kubernetes worker running.
  flannel/0*              active       idle                       Flannel subnet
kubernetes-worker/2       active       idle   11                    Kubernetes worker running.
  flannel/1               active       idle                      Flannel subnet

Machine  State    DNS            Inst id              Series  AZ
0        started    i-06e66414008eca61c  xenial  us-east-2c
3        started   i-0038186d2c5103739  xenial  us-east-2b
4        started   i-0ac66c86a8ec93b18  xenial  us-east-2a
5        started   i-078cfe79313d598c9  xenial  us-east-2c
7        started   i-00fd70321a51b658b  xenial  us-east-2c
8        started  i-0109a5fc942c53ed7  xenial  us-east-2b
9        started  i-0ab63e34959cace8d  xenial  us-east-2b
10       started   i-0108a8cc0978954b5  xenial  us-east-2a
11       started  i-0f5562571c649f0f2  xenial  us-east-2c

Interacting with the cluster

After the cluster is deployed you may assume control over the cluster from any kubernetes-master, or kubernetes-worker node.

First you need to download the credentials and client application to your local workstation:

Create the kubectl config directory.

mkdir -p ~/.kube

Copy the kubeconfig file to the default location.

juju scp kubernetes-master/0:config ~/.kube/config

Fetch a binary for the architecture you have deployed. If your client is a different architecture you will need to get the appropriate kubectl binary through other means. In this example we copy kubectl to ~/bin for convenience, by default this should be in your $PATH.

mkdir -p ~/bin
juju scp kubernetes-master/0:kubectl ~/bin/kubectl

Query the cluster:

kubectl cluster-info


Kubernetes master is running at
Heapster is running at
KubeDNS is running at
Grafana is running at
InfluxDB is running at

Congratulations, you’ve now set up a Kubernetes cluster!

Scale up cluster

Want larger Kubernetes nodes? It is easy to request different sizes of cloud resources from Juju by using constraints. You can increase the amount of CPU or memory (RAM) in any of the systems requested by Juju. This allows you to fine tune the Kubernetes cluster to fit your workload. Use flags on the bootstrap command or as a separate juju constraints command. Look to the Juju documentation for machine details.

Scale out cluster

Need more workers? We just add more units:

juju add-unit kubernetes-worker

Or multiple units at one time:

juju add-unit -n3 kubernetes-worker

You can also ask for specific instance types or other machine-specific constraints. See the constraints documentation for more information. Here are some examples, note that generic constraints such as cores and mem are more portable between clouds. In this case we’ll ask for a specific instance type from AWS:

juju set-constraints kubernetes-worker instance-type=c4.large
juju add-unit kubernetes-worker

You can also scale the etcd charm for more fault tolerant key/value storage:

juju add-unit -n3 etcd

It is strongly recommended to run an odd number of units for quorum.

Tear down cluster

If you want stop the servers you can destroy the Juju model or the controller. Use the juju switch command to get the current controller name:

juju switch
juju destroy-controller $controllername --destroy-all-models

This will shutdown and terminate all running instances on that cloud.

More Info

The Ubuntu Kubernetes deployment uses open-source operations, or operations as code, known as charms. These charms are assembled from layers which keeps the code smaller and more focused on the operations of just Kubernetes and its components.

The Kubernetes layer and bundles can be found in the kubernetes project on

Feature requests, bug reports, pull requests and feedback are appreciated.


Create an Issue Edit this Page