Instructions for setting up a Kubernetes cluster.

Documentation for Kubernetes v1.7 is no longer actively maintained. The version you are currently viewing is a static snapshot. For up-to-date documentation, see the latest version.

Edit This Page

Using kubeadm to Create a Cluster

This quickstart shows you how to easily install a Kubernetes cluster on machines running Ubuntu 16.04+, CentOS 7 or HypriotOS v1.0.1+. The installation uses a tool called kubeadm which is part of Kubernetes. As of v1.6, kubeadm aims to create a secure cluster out of the box via mechanisms such as RBAC.

This process works with local VMs, physical servers and/or cloud servers. It is simple enough that you can easily integrate its use into your own automation (Terraform, Chef, Puppet, etc).

See the full kubeadm reference for information on all kubeadm command-line flags and for advice on automating kubeadm itself.

kubeadm assumes you have a set of machines (virtual or real) that are up and running. It is designed to be part of a large provisioning system - or just for easy manual provisioning. kubeadm is a great choice where you have your own infrastructure (e.g. bare metal), or where you have an existing orchestration system (e.g. Puppet) that you have to integrate with.

If you are not constrained, there are other higher-level tools built to give you complete clusters:

kubeadm Maturity

Aspect Maturity Level
Command line UX beta
Config file alpha
Selfhosting alpha
kubeadm alpha commands alpha
Implementation alpha

The experience for the command line is currently in beta and we are trying hard not to change command line flags and break that flow. Other parts of the experience are still under active development. Specifically, kubeadm relies on some features (bootstrap tokens, cluster signing), that are still considered alpha. The implementation may change as the tool evolves to support even easier upgrades and high availability (HA). Any commands under kubeadm alpha (not documented here) are, of course, alpha.

Be sure to read the limitations. Specifically, configuring cloud providers is difficult.

Before you begin

  1. One or more machines running Ubuntu 16.04+, CentOS 7 or HypriotOS v1.0.1+
  2. 1GB or more of RAM per machine (any less will leave little room for your apps)
  3. Full network connectivity between all machines in the cluster (public or private network is fine)



(1/4) Installing kubeadm on your hosts

See Installing kubeadm.

Note: If you already have kubeadm installed, you should do a apt-get update && apt-get upgrade or yum update to get the latest version of kubeadm.

The kubelet is now restarting every few seconds, as it waits in a crashloop for kubeadm to tell it what to do.

(2/4) Initializing your master

The master is the machine where the control plane components run, including etcd (the cluster database) and the API server (which the kubectl CLI communicates with).

To initialize the master, pick one of the machines you previously installed kubeadm on, and run:

kubeadm init


Please refer to the kubeadm reference doc if you want to read more about the flags kubeadm init provides.

kubeadm init will first run a series of prechecks to ensure that the machine is ready to run Kubernetes. It will expose warnings and exit on errors. It will then download and install the cluster database and control plane components. This may take several minutes.

You can’t run kubeadm init twice without tearing down the cluster in between (unless you’re upgrading from v1.6 to v1.7), see Tear Down.

The output should look like:

[kubeadm] WARNING: kubeadm is in beta, please do not use it for production clusters.
[init] Using Kubernetes version: v1.7.0
[init] Using Authorization modes: [Node RBAC]
[preflight] Running pre-flight checks
[preflight] Starting the kubelet service
[certificates] Generated CA certificate and key.
[certificates] Generated API server certificate and key.
[certificates] API Server serving cert is signed for DNS names [kubeadm-master kubernetes kubernetes.default kubernetes.default.svc kubernetes.default.svc.cluster.local] and IPs []
[certificates] Generated API server kubelet client certificate and key.
[certificates] Generated service account token signing key and public key.
[certificates] Generated front-proxy CA certificate and key.
[certificates] Generated front-proxy client certificate and key.
[certificates] Valid certificates and keys now exist in "/etc/kubernetes/pki"
[kubeconfig] Wrote KubeConfig file to disk: "/etc/kubernetes/admin.conf"
[kubeconfig] Wrote KubeConfig file to disk: "/etc/kubernetes/kubelet.conf"
[kubeconfig] Wrote KubeConfig file to disk: "/etc/kubernetes/controller-manager.conf"
[kubeconfig] Wrote KubeConfig file to disk: "/etc/kubernetes/scheduler.conf"
[apiclient] Created API client, waiting for the control plane to become ready
[apiclient] All control plane components are healthy after 16.502136 seconds
[token] Using token: <token>
[apiconfig] Created RBAC rules
[addons] Applied essential addon: kube-proxy
[addons] Applied essential addon: kube-dns

Your Kubernetes master has initialized successfully!

To start using your cluster, you need to run (as a regular user):

  mkdir -p $HOME/.kube
  sudo cp -i /etc/kubernetes/admin.conf $HOME/.kube/config
  sudo chown $(id -u):$(id -g) $HOME/.kube/config

You should now deploy a pod network to the cluster.
Run "kubectl apply -f [podnetwork].yaml" with one of the options listed at:

You can now join any number of machines by running the following on each node
as root:

  kubeadm join --token <token> <master-ip>:<master-port>

Make a record of the kubeadm join command that kubeadm init outputs. You will need this in a moment.

The token is used for mutual authentication between the master and the joining nodes. The token included here is secret, keep it safe — anyone with this token can add authenticated nodes to your cluster. These tokens can be listed, created and deleted with the kubeadm token command. See the reference guide.

(3/4) Installing a pod network

You must install a pod network add-on so that your pods can communicate with each other.

The network must be deployed before any applications. Also, kube-dns, a helper service, will not start up before a network is installed. kubeadm only supports Container Network Interface (CNI) based networks (and does not support kubenet).

Several projects provide Kubernetes pod networks using CNI, some of which also support Network Policy. See the add-ons page for a complete list of available network add-ons.

New for Kubernetes 1.6: kubeadm 1.6 sets up a more secure cluster by default. As such it uses RBAC to grant limited privileges to workloads running on the cluster. This includes networking integrations. As such, ensure that you are using a network system that has been updated to run with 1.6 and RBAC.

You can install a pod network add-on with the following command:

kubectl apply -f <add-on.yaml>

NOTE: You can install only one pod network per cluster.

Please select one of the tabs to see installation instructions for the respective third-party Pod Network Provider.

The official Calico guide is here.


  • In order for Network Policy to work correctly, you need to pass --pod-network-cidr= to kubeadm init.
  • Calico works on amd64 only.
kubectl apply -f

The official Canal set-up guide is here.


  • For Canal to work correctly, --pod-network-cidr= has to be passed to kubeadm init.
  • Canal works on amd64 only.
kubectl apply -f
kubectl apply -f


  • For flannel to work correctly, --pod-network-cidr= has to be passed to kubeadm init.
  • flannel works on amd64, arm, arm64 and ppc64le, but for it to work on an other platform than amd64 you have to manually download the manifest and replace amd64 occurences with your chosen platform.
kubectl apply -f

Kube-router relies on kube-controll-manager to allocate pod CIDR for the nodes. Therefore, use kubeadm init with the --pod-network-cidr flag.

Kube-router provides pod networking, network policy, and high-performing IP Virtual Server(IPVS)/Linux Virtual Server(LVS) based service proxy.

For information on setting up Kubernetes cluster with Kube-router using kubeadm please see official setup guide.

The official Romana set-up guide is here.

Note: Romana works on amd64 only.

kubectl apply -f

The official Weave Net set-up guide is here.

Note: Weave Net works on amd64, arm and arm64 without any extra action required. Weave Net sets hairpin mode by default. This allows Pods to access themselves via their Service IP address if they don’t know their PodIP.

export kubever=$(kubectl version | base64 | tr -d '\n')
kubectl apply -f "$kubever"

Once a pod network has been installed, you can confirm that it is working by checking that the kube-dns pod is Running in the output of kubectl get pods --all-namespaces. And once the kube-dns pod is up and running, you can continue by joining your nodes.

If your network is not working or kube-dns is not in the Running state, check out the troubleshooting section below.

Master Isolation

By default, your cluster will not schedule pods on the master for security reasons. If you want to be able to schedule pods on the master, e.g. for a single-machine Kubernetes cluster for development, run:

kubectl taint nodes --all

With output looking something like:

node "test-01" untainted
taint key="dedicated" and effect="" not found.
taint key="dedicated" and effect="" not found.

This will remove the taint from any nodes that have it, including the master node, meaning that the scheduler will then be able to schedule pods everywhere.

(4/4) Joining your nodes

The nodes are where your workloads (containers and pods, etc) run. To add new nodes to your cluster do the following for each machine:

The output should look something like:

[kubeadm] WARNING: kubeadm is in beta, please do not use it for production clusters.
[preflight] Running pre-flight checks
[discovery] Trying to connect to API Server ""
[discovery] Created cluster-info discovery client, requesting info from ""
[discovery] Cluster info signature and contents are valid, will use API Server ""
[discovery] Successfully established connection with API Server ""
[bootstrap] Detected server version: v1.7.0
[bootstrap] The server supports the Certificates API (
[csr] Created API client to obtain unique certificate for this node, generating keys and certificate signing request
[csr] Received signed certificate from the API server, generating KubeConfig...
[kubeconfig] Wrote KubeConfig file to disk: "/etc/kubernetes/kubelet.conf"

Node join complete:
* Certificate signing request sent to master and response
* Kubelet informed of new secure connection details.

Run 'kubectl get nodes' on the master to see this machine join.

A few seconds later, you should notice this node in the output from kubectl get nodes when run on the master.

(Optional) Controlling your cluster from machines other than the master

In order to get a kubectl on some other computer (e.g. laptop) to talk to your cluster, you need to copy the administrator kubeconfig file from your master to your workstation like this:

scp root@<master ip>:/etc/kubernetes/admin.conf .
kubectl --kubeconfig ./admin.conf get nodes

Note: If you are using GCE, instances disable ssh access for root by default. If that’s the case you can log in to the machine, copy the file someplace that can be accessed and then use gcloud compute copy-files.

(Optional) Proxying API Server to localhost

If you want to connect to the API Server from outside the cluster you can use kubectl proxy:

scp root@<master ip>:/etc/kubernetes/admin.conf .
kubectl --kubeconfig ./admin.conf proxy

You can now access the API Server locally at http://localhost:8001/api/v1

(Optional) Installing a sample application

Now it is time to take your new cluster for a test drive. Sock Shop is a sample microservices application that shows how to run and connect a set of services on Kubernetes. To learn more about the sample microservices app, see the GitHub README.

Note that the Sock Shop demo only works on amd64.

kubectl create namespace sock-shop
kubectl apply -n sock-shop -f ""

You can then find out the port that the NodePort feature of services allocated for the front-end service by running:

kubectl -n sock-shop get svc front-end

Sample output:

front-end   <nodes>       80:30001/TCP   59s

It takes several minutes to download and start all the containers, watch the output of kubectl get pods -n sock-shop to see when they’re all up and running.

Then go to the IP address of your cluster’s master node in your browser, and specify the given port. So for example, http://<master_ip>:<port>. In the example above, this was 30001, but it may be a different port for you.

If there is a firewall, make sure it exposes this port to the internet before you try to access it.

To uninstall the socks shop, run kubectl delete namespace sock-shop on the master.

Tear down

To undo what kubeadm did, you should first drain the node and make sure that the node is empty before shutting it down.

Talking to the master with the appropriate credentials, run:

kubectl drain <node name> --delete-local-data --force --ignore-daemonsets
kubectl delete node <node name>

Then, on the node being removed, reset all kubeadm installed state:

kubeadm reset

If you wish to start over simply run kubeadm init or kubeadm join with the appropriate arguments.


Instructions for upgrading kubeadm clusters can be found here.

Explore other add-ons

See the list of add-ons to explore other add-ons, including tools for logging, monitoring, network policy, visualization & control of your Kubernetes cluster.

What’s next


Version skew policy

The kubeadm CLI tool of version vX.Y may deploy clusters with a control plane of version vX.Y or vX.(Y-1). kubeadm CLI vX.Y can also upgrade an existing kubeadm-created cluster of version vX.(Y-1).

Due to that we can’t see into the future, kubeadm CLI vX.Y may or may not be able to deploy vX.(Y+1) clusters.

Example: kubeadm v1.7 can deploy both v1.6 and v1.7 clusters and upgrade v1.6 kubeadm-created clusters to v1.7.

kubeadm is multi-platform

kubeadm deb/rpm packages and binaries are built for amd64, arm (32-bit), arm64, ppc64le, and s390x following the multi-platform proposal.

Only some of the network providers offer solutions for all platforms. Please consult the list of network providers above or the documentation from each provider to figure out whether the provider supports your chosen platform.


Please note: kubeadm is a work in progress and these limitations will be addressed in due course.

  1. The cluster created here has a single master, with a single etcd database running on it. This means that if the master fails, your cluster loses its configuration data and will need to be recreated from scratch. Adding HA support (multiple etcd servers, multiple API servers, etc) to kubeadm is still a work-in-progress.

    Workaround: regularly back up etcd. The etcd data directory configured by kubeadm is at /var/lib/etcd on the master.


You may have trouble in the configuration if you see Pod statuses like RunContainerError, CrashLoopBackOff or Error.

  1. There are Pods in the RunContainerError, CrashLoopBackOff or Error state. Right after kubeadm init there should not be any such Pods. If there are Pods in such a state right after kubeadm init, please open an issue in the kubeadm repo. kube-dns should be in the Pending state until you have deployed the network solution. However, if you see Pods in the RunContainerError, CrashLoopBackOff or Error state after deploying the network solution and nothing happens to kube-dns, it’s very likely that the Pod Network solution that you installed is somehow broken. You might have to grant it more RBAC privileges or use a newer version. Please file an issue in the Pod Network providers’ issue tracker and get the issue triaged there.

  2. The kube-dns Pod is stuck in the Pending state forever. This is expected and part of the design. kubeadm is network provider-agnostic, so the admin should install the pod network solution of choice. You have to install a Pod Network before kube-dns may deployed fully. Hence the Pending state before the network is set up.

  3. I tried to set HostPort on one workload, but it didn’t have any effect. The HostPort and HostIP functionality is available depending on your Pod Network provider. Please contact the author of the Pod Network solution to find out whether HostPort and HostIP functionality are available.

    If not, you may still use the NodePort feature of services or use HostNetwork=true.

  4. Pods cannot access themselves via their Service IP. Many network add-ons do not yet enable hairpin mode which allows pods to access themselves via their Service IP if they don’t know about their podIP. This is an issue related to CNI. Please contact the providers of the network add-on providers to get timely information about whether they support hairpin mode.

  5. If you are using VirtualBox (directly or via Vagrant), you will need to ensure that hostname -i returns a routable IP address (i.e. one on the second network interface, not the first one). By default, it doesn’t do this and kubelet ends-up using first non-loopback network interface, which is usually NATed. Workaround: Modify /etc/hosts, take a look at this Vagrantfileubuntu-vagrantfile for how this can be achieved.

  6. The following error indicates a possible certificate mismatch.

# kubectl get po                           
Unable to connect to the server: x509: certificate signed by unknown authority (possibly because of "crypto/rsa: verification error" while trying to verify candidate authority certificate "kubernetes")       

Verify that the $HOME/.kube/config file contains a valid certificate, and regenerate a certificate if necessary. Another workaround is to overwrite the default kubeconfig for the “admin” user:

mv  $HOME/.kube $HOME/.kube.bak
mkdir -p $HOME/.kube
sudo cp -i /etc/kubernetes/admin.conf $HOME/.kube/config
sudo chown $(id -u):$(id -g) $HOME/.kube/config
  1. If you are using CentOS and encounter difficulty while setting up the master node, verify that your Docker cgroup driver matches the kubelet config:
docker info |grep -i cgroup
cat /etc/systemd/system/kubelet.service.d/10-kubeadm.conf

If the Docker cgroup driver and the kubelet config don’t match, change the kubelet config to match the Docker cgroup driver.





Then restart kubelet:

systemctl daemon-reload
systemctl restart kubelet

The kubectl describe pod or kubectl logs commands can help you diagnose errors. For example:

kubectl -n ${NAMESPACE} describe pod ${POD_NAME}

kubectl -n ${NAMESPACE} logs ${POD_NAME} -c ${CONTAINER_NAME}


Create an Issue Edit this Page