How to get started, and achieve tasks, using Kubernetes

Edit This Page

Kubernetes on Mesos

About Kubernetes on Mesos

Mesos allows dynamic sharing of cluster resources between Kubernetes and other first-class Mesos frameworks such as Hadoop, Spark, and Chronos. Mesos also ensures applications from different frameworks running on your cluster are isolated and that resources are allocated fairly among them.

Mesos clusters can be deployed on nearly every IaaS cloud provider infrastructure or in your own physical datacenter. Kubernetes on Mesos runs on-top of that and therefore allows you to easily move Kubernetes workloads from one of these environments to the other.

This tutorial will walk you through setting up Kubernetes on a Mesos cluster. It provides a step by step walk through of adding Kubernetes to a Mesos cluster and starting your first pod with an nginx webserver.

NOTE: There are known issues with the current implementation and support for centralized logging and monitoring is not yet available. Please file an issue against the kubernetes-mesos project if you have problems completing the steps below.

Further information is available in the Kubernetes on Mesos contrib directory.


Note: You can, but you don’t have to deploy Kubernetes-Mesos on the same machine the Mesos master is running on.

Deploy Kubernetes-Mesos

Log into the future Kubernetes master node over SSH, replacing the placeholder below with the correct IP address.

ssh jclouds@${ip_address_of_master_node}

Build Kubernetes-Mesos.

git clone
cd kube-mesos-framework

Set some environment variables. The internal IP address of the master may be obtained via hostname -i.

export KUBERNETES_MASTER_IP=$(hostname -i)

Note that KUBERNETES_MASTER is used as the api endpoint. If you have existing ~/.kube/config and point to another endpoint, you need to add option --server=${KUBERNETES_MASTER} to kubectl in later steps.

Deploy etcd

Start etcd and verify that it is running:

sudo docker run -d --hostname $(uname -n) --name etcd \
  -p 4001:4001 -p 7001:7001 \
  --listen-client-urls \
  --advertise-client-urls http://${KUBERNETES_MASTER_IP}:4001
$ sudo docker ps
CONTAINER ID   IMAGE                        COMMAND   CREATED   STATUS   PORTS                NAMES
fd7bac9e2301   "/etcd"   5s ago    Up 3s    2379/tcp, 2380/...   etcd

It’s also a good idea to ensure your etcd instance is reachable by testing it

curl -L http://${KUBERNETES_MASTER_IP}:4001/v2/keys/

If connectivity is OK, you will see an output of the available keys in etcd (if any).

Start Kubernetes-Mesos Services

Update your PATH to more easily run the Kubernetes-Mesos binaries:

export PATH="$(pwd)/_output/local/go/bin:$PATH"

Identify your Mesos master: depending on your Mesos installation this is either a host:port like mesos-master:5050 or a ZooKeeper URL like zk://zookeeper:2181/mesos. In order to let Kubernetes survive Mesos master changes, the ZooKeeper URL is recommended for production environments.

export MESOS_MASTER=<host:port or zk:// url>

Create a cloud config file mesos-cloud.conf in the current directory with the following contents:

$ cat <<EOF >mesos-cloud.conf
        mesos-master        = ${MESOS_MASTER}

Now start the kubernetes-mesos API server, controller manager, and scheduler on the master node:

$ km apiserver \
  --address=${KUBERNETES_MASTER_IP} \
  --etcd-servers=http://${KUBERNETES_MASTER_IP}:4001 \
  --service-cluster-ip-range= \
  --port=8888 \
  --cloud-provider=mesos \
  --cloud-config=mesos-cloud.conf \
  --secure-port=0 \
  --v=1 >apiserver.log 2>&1 &

$ km controller-manager \
  --master=${KUBERNETES_MASTER_IP}:8888 \
  --cloud-provider=mesos \
  --cloud-config=./mesos-cloud.conf  \
  --v=1 >controller.log 2>&1 &

$ km scheduler \
  --address=${KUBERNETES_MASTER_IP} \
  --mesos-master=${MESOS_MASTER} \
  --etcd-servers=http://${KUBERNETES_MASTER_IP}:4001 \
  --mesos-user=root \
  --api-servers=${KUBERNETES_MASTER_IP}:8888 \
  --cluster-dns= \
  --cluster-domain=cluster.local \
  --v=2 >scheduler.log 2>&1 &

Disown your background jobs so that they’ll stay running if you log out.

disown -a

Validate KM Services

Interact with the kubernetes-mesos framework via kubectl:

$ kubectl get pods
# NOTE: your service IPs will likely differ
$ kubectl get services
NAME             LABELS                                    SELECTOR   IP(S)          PORT(S)
k8sm-scheduler   component=scheduler,provider=k8sm         <none>   10251/TCP
kubernetes       component=apiserver,provider=kubernetes   <none>     443/TCP

Lastly, look for Kubernetes in the Mesos web GUI by pointing your browser to http://<mesos-master-ip:port>. Make sure you have an active VPN connection. Go to the Frameworks tab, and look for an active framework named “Kubernetes”.

Spin up a pod

Write a JSON pod description to a local file:

$ cat <<EOPOD >nginx.yaml
apiVersion: v1
kind: Pod
  name: nginx
  - name: nginx
    image: nginx
    - containerPort: 80

Send the pod description to Kubernetes using the kubectl CLI:

$ kubectl create -f ./nginx.yaml

Wait a minute or two while dockerd downloads the image layers from the internet. We can use the kubectl interface to monitor the status of our pod:

$ kubectl get pods
nginx     1/1       Running   0          14s

Verify that the pod task is running in the Mesos web GUI. Click on the Kubernetes framework. The next screen should show the running Mesos task that started the Kubernetes pod.

Launching kube-dns

Kube-dns is an addon for Kubernetes which adds DNS-based service discovery to the cluster. For a detailed explanation see DNS in Kubernetes.

The kube-dns addon runs as a pod inside the cluster. The pod consists of three co-located containers:

The skydns container offers DNS service via port 53 to the cluster. The etcd communication works via local communication

We assume that kube-dns will use

Note that we have passed these two values already as parameter to the apiserver above.

A template for an replication controller spinning up the pod with the 3 containers can be found at cluster/addons/dns/ in the repository. The following steps are necessary in order to get a valid replication controller yaml file:

In addition the service template at cluster/addons/dns/ needs the following replacement:

To do this automatically:

sed -e "s/{{ pillar\['dns_replicas'\] }}/1/g;"\
"s,\(command = \"/kube2sky\"\),\\1\\"$'\n'"        - --kube_master_url=${KUBERNETES_MASTER},;"\
"s/{{ pillar\['dns_domain'\] }}/cluster.local/g" \
  cluster/addons/dns/ > skydns-rc.yaml
sed -e "s/{{ pillar\['dns_server'\] }}/" \
  cluster/addons/dns/ > skydns-svc.yaml

Now the kube-dns pod and service are ready to be launched:

kubectl create -f ./skydns-rc.yaml
kubectl create -f ./skydns-svc.yaml

Check with kubectl get pods --namespace=kube-system that 3/3 containers of the pods are eventually up and running. Note that the kube-dns pods run in the kube-system namespace, not in default.

To check that the new DNS service in the cluster works, we start a busybox pod and use that to do a DNS lookup. First create the busybox.yaml pod spec:

cat <<EOF >busybox.yaml
apiVersion: v1
kind: Pod
  name: busybox
  namespace: default
  - image: busybox
      - sleep
      - "3600"
    imagePullPolicy: IfNotPresent
    name: busybox
  restartPolicy: Always

Then start the pod:

kubectl create -f ./busybox.yaml

When the pod is up and running, start a lookup for the Kubernetes master service, made available on by default:

kubectl  exec busybox -- nslookup kubernetes

If everything works fine, you will get this output:

Address 1:

Name:      kubernetes
Address 1:

## Support Level

IaaS Provider Config. Mgmt OS Networking Docs Conforms Support Level
Mesos/GCE       docs   Community (Kubernetes-Mesos Authors)

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

What next?

Try out some of the standard Kubernetes examples.

Read about Kubernetes on Mesos’ architecture in the contrib directory.

NOTE: Some examples require Kubernetes DNS to be installed on the cluster. Future work will add instructions to this guide to enable support for Kubernetes DNS.

NOTE: Please be aware that there are known issues with the current Kubernetes-Mesos implementation.


Create Issue Edit This Page