Skip to main content
Version: 0.7 (Next)

Private Clusters

Alpha

Civo Bare Metal is alpha functionality. Bare-metal clusters managed by Konstruct are priced per the pricing page — no upcharge while in alpha.

Private clusters are Kubernetes clusters running on your own hardware. Once provisioned, they are managed by Konstruct like any other cluster: deploy from the GitOps catalog, attach environments, apply Helm and pipeline templates, and govern access with the same organizations and RBAC.

Cluster types​

Colony provisions three cluster flavors onto discovered hardware:

TypeBest forHighlights
Talos LinuxProduction clustersImmutable, API-driven Linux purpose-built for Kubernetes; single-node to HA (3 control plane + N workers)
K3sEdge, dev/test, constrained hardwareLightweight; ships with Flannel CNI, Traefik ingress, local-path storage, ServiceLB out of the box
CivoStackFull cloud experience on your metalTalos + the Civo Stack Engine: Cilium, distributed block storage (Mayastor), Ceph object/block storage, ingress, monitoring, Autopilot — the platform that runs the Civo public cloud

Creation guides in the Colony docs: Talos · K3s · CivoStack

Declarative from end to end​

A cluster is declared as a single TenantCluster resource — control plane replicas, worker pools, hardware references, Kubernetes version, and network CIDRs. The Colony operator generates the provisioning plans, PXE-boots the machines, bootstraps the cluster, and publishes the kubeconfig as a Secret when the cluster reaches Ready:

Planning → ProvisioningControlPlane → Bootstrapping → ProvisioningWorkers → WaitingForNodes → Ready

Day-2 operations follow the same model — add nodes by extending a worker pool, upgrade by bumping the version in the spec: Add nodes · Access clusters.

Storage and data-heavy workloads​

For storage-intensive platforms, CivoStack clusters bring the storage stack Civo runs in production:

  • Mayastor — high-performance distributed block storage across worker nodes
  • Ceph (rook-ceph) — object storage (S3-compatible RGW) and block storage (RBD)
  • External S3 tiering — S3-compatible object storage integration (e.g. Wasabi)

Multi-site and failover posture​

Because every layer — hardware inventory, provisioning intent, cluster definitions, and the applications Konstruct delivers on top — is declared in Git, a cluster is rebuildable from its declarations. Combined with per-site isolation and multi-site management from one control plane, this is the foundation for cross-data-center failover designs: the desired state travels with the Git repository, not the building.