Kubernetes Bare-Metal Worker Nodes

Kubernetes bare-metal deployments give you cloud-native orchestration without cloud markup. We stock dense-compute servers for K8s worker nodes: Dell PowerEdge R740/R750 and HPE ProLiant DL380 Gen10/Gen11 with high core counts (32-48 cores per node) and plenty of memory (256-512 GB).

Common K8s distributions on bare metal: Rancher (RKE2 / K3s), Red Hat OpenShift, Mirantis Kubernetes Engine, SUSE Rancher, EKS Anywhere, Azure Arc-enabled K8s. We have hardware tested against each — request the validation report for your chosen distro.

Networking: Calico, Cilium, and Flannel CNIs all work well on dual 10/25 GbE NICs. For high-performance workloads, use SR-IOV with Mellanox ConnectX-5/6 25-100 GbE cards (we stock these). Storage: Longhorn, Rook+Ceph, or Portworx work well with local NVMe attached to each node.

Sizing rule of thumb: 1 worker node = 16-48 vCPUs, 64-256 GB RAM. A 3-node cluster (the minimum HA configuration) gives you 48-144 vCPUs and 192 GB-768 GB RAM. Scale horizontally by adding nodes; Kubernetes handles workload distribution automatically.

Frequently Asked Questions

How many nodes do I need for a Kubernetes cluster?

Minimum 3 control-plane nodes (etcd requires odd-number quorum) + at least 2 worker nodes for HA. For production: 3 control-plane + 4+ worker nodes. Small dev/test clusters can run 3 mixed control-plane+worker nodes (single-node K8s like K3s also works for edge/lab).

What is the difference between Kubernetes on bare metal vs cloud?

Bare metal gives you better per-cycle cost (no cloud markup), full hardware control (GPU, FPGA, custom NICs), no egress fees, but requires more operational overhead (managing nodes, upgrades, hardware failures). Cloud K8s (EKS, GKE, AKS) trades cost for managed simplicity.

Do I need SSDs for Kubernetes worker nodes?

Yes — SSDs are recommended for the OS disk and any local persistent volumes. Most workloads benefit from NVMe SSDs for low latency. Don't use spinning disks for K8s worker nodes — etcd will throw timeouts and pods will OOM-kill at the wrong moments.

Can I run OpenShift on refurbished hardware?

Yes — Red Hat OpenShift Container Platform runs on any Red Hat Enterprise Linux-certified hardware, including refurbished Dell PowerEdge and HPE ProLiant. The Red Hat Hardware Catalog (access.redhat.com/ecosystem) lists certified platforms — most of our refurbished SKUs are on it.

Other Use-Case Hardware

Part of

Server Virtualization & HCI Hub

View all 54 pages →

VMware vSphere, Hyper-V, Proxmox, KVM, Nutanix AHV — hardware sizing, licensing math, deployment.

Related Pages