Service

On-premise Kubernetes setup and management on your own hardware

Bare-metal and on-premise cluster design, storage, load balancing and day-2 operations — for workloads that cannot sit in a public cloud.

On a managed cloud cluster the provider quietly supplies the control plane, the load balancer and the storage layer. On your own hardware, all three become your problem, and that is where most on-premise Kubernetes projects stall. We build clusters on bare metal or on-premise virtualisation for the cases that genuinely call for it — data residency, an existing hardware investment, plant-floor latency, or an air-gapped network — and we will tell you when the case is not there.

What the engagement covers

Cluster design on your hardware

Control-plane topology and etcd sizing, node roles, and a distribution choice — kubeadm, RKE2, k3s or Talos — matched to who has to operate it after we leave.

Load balancing without a cloud LB

MetalLB or a BGP-based setup, plus ingress, so a Service of type LoadBalancer means something on a network that has no cloud provider behind it.

Storage

Persistent volumes backed by Ceph, Longhorn or an existing SAN/NFS estate, with the failure and backup behaviour tested rather than assumed.

Air-gapped and restricted networks

A private registry, mirrored charts and an offline upgrade path, for sites where pulling from the internet at deploy time is not an option.

Hybrid connectivity

Linking on-premise clusters to cloud services where that split is deliberate, with consistent identity and observability across both.

Day-2 operations

Node replacement, certificate rotation, etcd backup and restore, and a rehearsed upgrade path — the operations a cloud provider would otherwise be doing for you.

Tools we work with

Not an exhaustive list, and not a claim of certification — just what we reach for most often on this kind of work.

  • Kubernetes
  • kubeadm
  • RKE2
  • k3s
  • Talos
  • MetalLB
  • Ceph
  • Longhorn
  • Harbor
  • Argo CD
  • Prometheus

Common questions

Should we run Kubernetes on-premise at all?
Only for a reason you can name — data that legally cannot leave the building, hardware you have already bought, latency to equipment on the floor, or a genuinely air-gapped network. If none of those apply, a managed cluster is less work to operate and we will say so.
Can you run one cluster on-premise and one in the cloud?
Yes, and it is a common shape: production close to the equipment, non-production in the cloud. We keep delivery, identity and observability the same across both so they do not become two different systems to learn.
What about hardware — do you supply it?
No. We specify what the workload needs and review what you already have, but we do not resell hardware, which keeps the sizing recommendation free of any interest in selling you more of it.

Talk to us about on-premise kubernetes

Send over what you are dealing with. We will reply with a straight assessment of whether we are the right team for it.