Verifying that the Existing Kubernetes Cluster is Ready for Bootstrapping

If you will be deploying the Compute Spaces (CSpaces) on an existing Kubernetes cluster, do the following to ensure that your Kubernetes cluster is configured and ready for bootstrapping:
  • Verify that kubectl is set up. Use the kubectl version command to verify that kubectl is installed in your client machine as a CLI access to your Kubernetes cluster.
  • Verify that the Kubernetes cluster running on bare-metal is set up and in context. To verify, get the context of your cluster before bootstrapping by running the kubectl config current-context command, and ensure that current-context is set to the IP address of the master. If the current-context is not set to the IP address of the master, use the kubectl config set-context command to rest the context.
  • List the nodes in your cluster to verify that cluster IP is displayed by running the kubectl get nodes command, and check the following:
    • Check if the number of nodes is as expected (a minimum 5-node cluster is required).
    • Check if one of the nodes is a master.
      Note: The master node is not available on GKE.
    • Check if the Status of nodes is Ready.
  • Use the kubectl get ns command to verify that the default Kubernetes namespaces are active on your cluster. The command should show the following active namespaces:
    • default
    • kube-public
    • Kube-system
  • Ensure that kubectl is configured with super-user access to your Kubernetes cluster. Your ID should have cluster-admin-level privileges to bootstrap. Use the following commands to provide the permissions:
    kubectl create clusterrolebinding user-cluster-admin-binding
      --clusterrole=cluster-admin
      --user=<USER>
    For more information, see Service Account Permissions.
  • On OpenShift clusters, run the following command to allow pods to run with root-user access. Without root-user access, Spark drivers can fail to start on OpenShift (described in the MapR Kubernetes Ecosystem release note known issue K8S-1082):
    oc adm policy add-scc-to-group anyuid system:authenticated