Skip to content

Kubernetes on Breqwatr Cloud

Breqwatr Cloud runs managed Kubernetes clusters as a project-level resource. Each cluster is a set of OpenStack instances (one or more masters, one or more workers) that Magnum provisions, joins, and exposes to you as a normal Kubernetes API endpoint with a downloadable kubeconfig.

How it works

Strata (the Breqwatr Cloud Portal) sends cluster lifecycle requests to OpenStack Magnum, which uses the magnum_cluster_api driver to provision the underlying resources via Cluster API semantics. The actual nodes, networking, load balancers, and storage are OpenStack resources sitting alongside everything else in your project.

You don't run any management cluster of your own — it's all operator-side plumbing. The cluster you create is just the workload cluster.

What the Portal exposes

For each cluster you can:

  • Create — pick a cluster template (operator-supplied; defines the Kubernetes version, OS image, and driver), set node counts and flavors, optionally toggle a highly-available control plane.
  • View status and node detail — provisioning state (CREATE_IN_PROGRESS → CREATE_COMPLETE), the master/worker layout, and the cluster's IP and API endpoint.
  • Download kubeconfig — single click once the cluster is ready.
  • Resize worker nodes — scale up or down. Master count is immutable.
  • Delete — tears the whole cluster down asynchronously.

Cluster templates control the heavy decisions (K8s version, OS image, networking driver). If you need one that isn't on the list, contact support.

Pages in this section

  • Create a Kubernetes cluster — the Portal flow end to end, plus the OpenStack CLI alternative.
  • Create a Kubernetes cluster with GPUs — GPU-flavor selection and the workload-side enablement (drivers, device plugins).
  • The Kubernetes AI sub-section covers cluster-side enablement for accelerator workloads — autoscaling, gang scheduling, KubeRay, Kubeflow, vulnerability-aware admission, and others. Those tutorials are mostly cloud-agnostic Kubernetes content (kubectl / helm / manifests) that runs the same way on any conformant cluster.

Prerequisites for this section

  • A project you can launch resources in (owner or admin role).
  • The kubectl CLI on your local machine for any post-cluster work.
  • An SSH key on your account — at least one key is required at cluster creation so you can reach the underlying nodes for debugging.

Next steps