website

8.4 KiB

Raw Blame History

Stop. This guide has been superseded by Minikube which is the recommended method of running Kubernetes on your local machine.

The following instructions show you how to set up a simple, single node Kubernetes cluster using Docker.

Here's a diagram of what the final result will look like:

TOC {:toc}

Prerequisites

Note: These steps have not been tested with the Docker For Mac or Docker For Windows beta programs.

You need to have docker installed on one machine.
Decide what Kubernetes version to use. Set the ${K8S_VERSION} variable to a released version of Kubernetes >= "v1.2.0". If you'd like to use the current stable version of Kubernetes, run the following:

export K8S_VERSION=$(curl -sS https://storage.googleapis.com/kubernetes-release/release/stable.txt)

and for the latest available version (including unstable releases):

export K8S_VERSION=$(curl -sS https://storage.googleapis.com/kubernetes-release/release/latest.txt)

Run it

export ARCH=amd64
docker run -d \
    --volume=/:/rootfs:ro \
    --volume=/sys:/sys:rw \
    --volume=/var/lib/docker/:/var/lib/docker:rw \
    --volume=/var/lib/kubelet/:/var/lib/kubelet:rw \
    --volume=/var/run:/var/run:rw \
    --net=host \
    --pid=host \
    --privileged \
    gcr.io/google_containers/hyperkube-${ARCH}:${K8S_VERSION} \
    /hyperkube kubelet \
        --containerized \
        --hostname-override=127.0.0.1 \
        --api-servers=http://localhost:8080 \
        --config=/etc/kubernetes/manifests \
        --cluster-dns=10.0.0.10 \
        --cluster-domain=cluster.local \
        --allow-privileged --v=2

Note that --cluster-dns and --cluster-domain is used to deploy dns, feel free to discard them if dns is not needed.

If you would like to mount an external device as a volume, add --volume=/dev:/dev to the command above. It may however, cause some problems described in #18230

Architectures other than amd64 are experimental and sometimes unstable, but feel free to try them out! Valid values: arm, arm64 and ppc64le. ARM is available with Kubernetes version v1.3.0-alpha.2 and higher. ARM 64-bit and PowerPC 64 little-endian are available with v1.3.0-alpha.3 and higher. Track progress on multi-arch support here

If you are behind a proxy, you need to pass the proxy setup to curl in the containers to pull the certificates. Create a .curlrc under /root folder (because the containers are running as root) with the following line:

proxy = <your_proxy_server>:<port>

This actually runs the kubelet, which in turn runs a pod that contains the other master components.

** SECURITY WARNING ** services exposed via Kubernetes using Hyperkube are available on the host node's public network interface / IP address. Because of this, this guide is not suitable for any host node/server that is directly internet accessible. Refer to #21735 for addtional info.

Download `kubectl`

At this point you should have a running Kubernetes cluster. You can test it out by downloading the kubectl binary for ${K8S_VERSION} (in this example: {{page.version}}.0).

Downloads:

The generic download path is:

http://storage.googleapis.com/kubernetes-release/release/${K8S_VERSION}/bin/${GOOS}/${GOARCH}/${K8S_BINARY}

An example install with linux/amd64:

curl -sSL "https://storage.googleapis.com/kubernetes-release/release/{{page.version}}.0/bin/linux/amd64/kubectl" > /usr/bin/kubectl
chmod +x /usr/bin/kubectl

On OS X, to make the API server accessible locally, setup a ssh tunnel.

docker-machine ssh `docker-machine active` -N -L 8080:localhost:8080

Setting up a ssh tunnel is applicable to remote docker hosts as well.

(Optional) Create kubernetes cluster configuration:

kubectl config set-cluster test-doc --server=http://localhost:8080
kubectl config set-context test-doc --cluster=test-doc
kubectl config use-context test-doc

Test it out

List the nodes in your cluster by running:

kubectl get nodes

This should print:

NAME        STATUS    AGE
127.0.0.1   Ready     1h

Run an application

kubectl run nginx --image=nginx --port=80

Now run docker ps you should see nginx running. You may need to wait a few minutes for the image to get pulled.

Expose it as a service

kubectl expose deployment nginx --port=80

Run the following command to obtain the cluster local IP of this service we just created:

ip=$(kubectl get svc nginx --template={{.spec.clusterIP}})
echo $ip
{% endraw %}```

On Linux the IP is directly accessible via a web browser.

On OS X, since docker is running inside a VM, run the following command instead:

```shell
docker-machine ssh `docker-machine active` curl $ip

Deploy a DNS

See here for instructions.

Turning down your cluster

Delete all the containers including the kubelet:

Many of these containers run under the management of the kubelet binary, which attempts to keep containers running, even if they fail. So, in order to turn down the cluster, you need to first kill the kubelet container, and then any other containers.

You may use docker rm -f $(docker ps -aq), note this removes all containers running under Docker, so use with caution.

Cleanup the filesystem:

On OS X, first ssh into the docker VM:

docker-machine ssh `docker-machine active`

grep /var/lib/kubelet /proc/mounts | awk '{print $2}' | sudo xargs -n1 umount
sudo rm -rf /var/lib/kubelet

Troubleshooting

Node is in `NotReady` state

If you see your node as NotReady it's possible that your OS does not have memcg enabled.

Your kernel should support memory accounting. Ensure that the following configs are turned on in your linux kernel:

CONFIG_RESOURCE_COUNTERS=y
CONFIG_MEMCG=y

Enable the memory accounting in the kernel, at boot, as command line parameters as follows:

GRUB_CMDLINE_LINUX="cgroup_enable=memory=1"

NOTE: The above is specifically for GRUB2. You can check the command line parameters passed to your kernel by looking at the output of /proc/cmdline:

$ cat /proc/cmdline
BOOT_IMAGE=/boot/vmlinuz-3.18.4-aufs root=/dev/sda5 ro cgroup_enable=memory=1

Support Level

IaaS Provider	Config. Mgmt	OS	Networking	Docs	Conforms	Support Level
Docker Single Node	custom	N/A	local	docs		Project (@brendandburns)

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

8.4 KiB Raw Blame History