Kubernetes
      Back to home
      1. Academy
      2. Tutorials
      3. Kubernetes

      Create a Kubernetes cluster with OpenStack Magnum on Fuga Cloud

      You will learn how to create a Kubernetes Cluster with the OpenStack CLI and Magnum.

       

      The terminology used in this document;

      OpenStack - Our Cloud infrastructure software used at Fuga
      Magnum - The engine in OpenStack that installs Kubernetes
      Heat - Orchestration service in OpenStack
      Stack - Object in Heat with a list of items
      Cinder - Storage backend in OpenStackRequirements

      Requirements

      To use the example scripts and files you need a few things;
      • OpenStack CLI setup and working
      • the jq binary to parse json in the scripts, check how to install on your OS
      • kubectl. make sure its the same as used in the cluster you are deploying
        At the time of writing 1.13.10 is deployed. More info on how to install kubectl
      • Clone of https://github.com/FugaCloud/magnum_additions

      Setting up the OpenStack CLI

      You need Python3 and its advised to use a virtual environment (virtualenv). To setup run;

      python3 ~/virtualenv/openstack
      . ~/virtualenv/openstack/bin/activate
      pip install openstackclient python-magnumclient

      Before you move on, you have to be able to use the OpenStack command and talk to OpenStack, for example, list your servers;

      openstack server list

      Verify it works as expected.

      Creating a template

      Before you can deploy a cluster, you need a template. This is the syntax;

      openstack coe cluster template create --image ${image} \
                                            --external-network public \
                                            --master-flavor ${master_flavor} \
                                            --flavor ${node_flavor} \
                                            --coe kubernetes \
                                            --volume-driver cinder \
                                            --network-driver flannel \
                                            --docker-volume-size ${volume_in_GB} \
                                            --labels kube_dashboard_enabled=0,kube_tag=${version_tag} \
                                            ${name}

       

      ${image} = "ac6c15cc-9073-4537-98d9-00f4ccfefa25"

      this is the fedora atomic image in Fuga's OpenStack infra at R2. Do not change this

       

      ${master_flavor} = "c2.small"

      this depends on your cluster sizing, for testing you can safely use c2.small

       

      ${node_flavor} = "c2.medium"

      also depends on personal usage, sizing, and preference

       

      ${volume_in_GB} = "choose size"

      All your nodes will have a volume attached. Set the sizing of these volumes in GB.

       

      ${version_tag} = "v1.13.10"

      this is appended to the tag pulled from docker.io. openstackmagnum on docker.io maintains these images and availability
      at this time, the highest verified working version is 1.13.10, so using tag "v1.13.10" is advised.

       

      ${name} = "your-cluster-template"

      Choose a cluster template name

      Submit your request and you will get your template information in return.

      Creating a cluster

      openstack coe cluster create --cluster-template ${template_name} \
                                   --master-count 1 \
                                   --node-count ${node_count} \
                                   --keypair ${keypair} \
                                   ${name}

       

      ${template_name}

      as created in the step 'Creating a template'

       

      master-count = "1"

      at this time, we do not support multi masters, so choose 1.

       

      ${node_count}

      how many worker nodes to deploy, they will have anti-affinity, OpenStack will try to divide them amongst the hypervisors

       

      ${keypair}

      the ssh keypair you wish to use deploying your servers (login user fedora)

       

      ${name}

      Choose a cluster name

       

      You can monitor the installation while you wait;

      openstack coe cluster list

       

      At some point, the status should switch from CREATE_IN_PROGRESS to CREATE_COMPLETE. When it's complete, we can retrieve our config ($name is your cluster name);

      openstack coe cluster config $name

       

      This will download a file called 'config' and place it in the working directory, and it will show you how to use it.

      export KUBECONFIG=/home/user/dir/config

       

      Before you can talk to the cluster, you need to open the Security Group rule (firewall). As a precaution, you cannot connect to 6443 yet.

      Go to the Security Groups tab in our dashboard and add a new rule to the Security Group of your master.

      Open Port: Port

      Port: 6443

      CIDR: <your_IP>/32

       

      Now kubectl should be able to use this, and you should be able to communicate with your cluster;

      kubectl get nodes -o wide

       

      Storage

      To configure Kubernetes for storage in OpenStack, apply the storageclass yaml;

      kubectl apply -f ../magnum_additions/storageclass.yaml

      Now Cinder is available and it's set as the default storageclass. Now the nodes need a label to be able to schedule nodes that need access to the volumes.
      You can use the supplied script

      ./node.sh

      Examples

      Nginx

      The cluster is now ready for deployment with storage. Try the Nginx example, which deploys Nginx with storage and a load balancer;

      kubectl apply -f nginx_example.yaml

       

      Once deployed, you should be able to list the services, and see Nginx;

      kubectl get svc

      nginx        LoadBalancer   10.254.253.43   <pending|ip>   80:30363/TCP

       

      You should see either 'pending' or an IP. When 'pending' it's still provisioning the load balancer. Once it's done, it will show you an IP.
      If your security group on the load balancer allows it, you can check Nginx on port 80, or use the script to open it for you;

      ./open_nginx_http.sh

       

      Now a curl on the IP should show you the default Nginx page. Once done, you can delete Nginx;

      kubectl delete -f nginx_example.yaml

      Conclusion

      You now learned how to create a Kubernetes Cluster through the Fuga Cloud CLI. If you have any questions please don't hesitate to start a chat. We are happy to help you.