Before you build clusters

• Ensure that all cluster nodes are up and running.

• Verify that the inventory file is updated as mentioned in the inventory sample file.

  • For Kubernetes, all the applicable groups are kube_control_plane, kube_node, and etcd.

  • The centralized authentication server inventory group, that is auth_server, is common for both Slurm and Kubernetes.

  • For secure login node functionality, ensure to add the login group in the provided inventory file.

• Verify that all nodes are assigned a group. The inventory file is case-sensitive. Follow the format provided in the sample file link.

Note

• The inventory file accepts both IPs and FQDNs as long as they can be resolved by DNS.

• In a multi-node setup, an IP cannot be listed as a control node and a compute node simultaneously. That is, don’t include the kube_control_plane IP address in the compute node group. In a single node setup, the compute node and the kube_control_plane must be the same.

• Users should also ensure that all repositories are available on the cluster nodes.

• If the cluster requires more than 10 kubernetes nodes, use a docker enterprise account to avoid docker pull limits.

If you have any feedback about Omnia documentation, please reach out at omnia.readme@dell.com.