# Should I run Kubernetes on bare metal or on VMs?

> Should you run Kubernetes on bare metal or on VMs? Bare metal for 3+ servers and local NVMe, VMs for one host, Talos or many clusters. No KYC, crypto.

- URL: https://lumenfer.com/kubernetes-bare-metal-vs-vms/
- Last updated: 2026-11-01
- Publisher: Lumenfer (https://lumenfer.com)

The same rented machines can carry Kubernetes directly or through Proxmox VE. What each design gives, what it costs, and when a single server is enough.

**Quick answer — Should I run Kubernetes on bare metal or on VMs?**

Should you run Kubernetes on bare metal or on VMs? Bare metal, when you have three or more servers and want every core and NVMe drive for pods. VMs, when you have one or two servers, need several clusters or want Talos. At **Lumenfer** both run on the same machines, with Proxmox VE 9 as an image. From $91.65 per 30 days, no KYC, paid in crypto.

## Kubernetes on bare metal or on VMs, on rented Lumenfer servers

|  | On bare metal | On VMs (Proxmox VE) |
| --- | --- | --- |
| Under the kubelet | The Linux host only | A hypervisor and a guest kernel |
| Servers for a highly available control plane | Three | Three hosts; one host gives quorum, not redundancy |
| Storage for pods | NVMe drives directly | Virtual disks on the host's storage |
| Talos Linux | No: there is no Talos image or console here | Yes: the ISO boots inside a VM |
| Snapshot before an upgrade | No: back up etcd and volumes yourself | Yes: snapshot the VMs |
| Console when networking breaks | None for customers | The VM console in the Proxmox web interface |
| Several clusters per machine | No | Yes |
| Public addresses | Each node holds its server's IPv4 | VMs sit behind NAT; the host forwards ports |
| 110-pod limit per node | One limit per machine | One limit per VM, so more pods per machine |

## When is bare metal the better place for Kubernetes?

**When you have at least three servers and your workloads want the hardware itself.** Three machines give three etcd members on three separate failure domains. Pods reach the NVMe drives with no virtual disk layer, which suits databases on local volumes. There is one kernel to patch and one network stack to debug.

The network stays simple too. Each node holds its server's own IPv4, so k3s ServiceLB or an ingress controller answers on real addresses. Our guide on [how to run Kubernetes on bare metal servers](https://lumenfer.com/kubernetes-bare-metal/ "How do I run Kubernetes on bare metal servers?") has the steps.

The weak spot is recovery. There is no IPMI or KVM-over-IP console for customers: you request reboots and reinstalls in the dashboard and the team performs them. A bad network change on a bare node can end in a reinstall. A replacement server is quick, live within the hour, but rebuilding a node still takes work. Keep enough capacity to lose one node for hours, not minutes.

## When are VMs the better choice?

**When you have one or two machines, several clusters to run, or Talos in mind.** A Proxmox host turns one server into a staging cluster, a production cluster and a test bed. Each VM can be snapshotted before an upgrade and rolled back after a failed one.

- **Talos Linux.** It boots only from its own ISO, PXE script or disk image. A Proxmox VM mounts the ISO, so Talos works here only in VMs.
- **A console of last resort.** The host has none, but every VM has one in the Proxmox web interface. A broken guest network is fixed from that console, not by a reinstall.
- **Pod density.** Kubernetes v1.37 is designed for no more than 110 pods per node. Splitting Lumen 32 (32 cores / 64 threads) into VMs gives each VM its own 110.

Sidero's Proxmox guide sizes Talos control-plane VMs at 2 or more cores and 4 GB. Workers get 4 or more cores and 8 GB. It sets the CPU type to `host` for AVX-512 and x86-64-v2, and notes that this blocks live migration.

## Is a three-VM cluster on one server highly available?

**No: it survives a crashed VM, not a dead server.** etcd keeps quorum while two of three members run, so one VM can fail. When the host fails, all three members go at once.

The Kubernetes high-availability guide names this risk for stacked control planes: losing a node loses an etcd member and a control-plane instance together. On one host, that node is the whole cluster. **Three VMs on one machine are a quorum, not redundancy.** Use it for staging, labs and Talos trials, and keep production on three machines.

## Can I combine both: Kubernetes VMs on a Proxmox cluster?

**Yes: three Proxmox hosts with one control-plane VM each is the most flexible design.** Each host is its own failure domain, and VMs keep their snapshots and consoles. The price is two clustering layers, corosync for Proxmox and etcd for Kubernetes, and both want low latency. Keep the three hosts in one city.

VMs on different hosts need a shared network. Proxmox SDN builds one with a VXLAN zone, but its documentation warns that VXLAN alone has no encryption. Carry it over WireGuard or another VPN between hosts. VXLAN also takes 50 bytes per packet, so guests use an MTU of 1450.

Our guide on [building a Proxmox cluster with rented dedicated servers](https://lumenfer.com/proxmox-cluster-dedicated-servers/ "How do I build a Proxmox cluster with rented dedicated servers?") covers quorum and corosync on public addresses.

## Does a hypervisor slow Kubernetes down?

**It adds a layer, and Lumenfer has measured nothing, so we give no figure.** A VM sees virtual CPUs, a virtual disk and a virtual network card. CPU features reach the guest only when the CPU type passes them through. Memory assigned to one VM is gone for the others and for the host.

Test your own deployment on a bare node and on a VM before you commit. Our guide to [finding honest benchmarks](https://lumenfer.com/dedicated-server-benchmarks/ "Where can I find honest benchmarks for dedicated server hosting?") explains how to measure a rented server yourself.

## Which Lumenfer plans fit each design?

**Bare metal needs three servers; VMs can start with one large one.** Prices are per server; 6- and 12-month terms cost less per month.

**13 of 18 no-KYC crypto hosts in Lumenfer's check of 30 October 2026 sold any dedicated server with a 2022-or-newer CPU and NVMe only.** Those servers made up 113 of 229 comparable plans. Every plan is on [what a Lumenfer server costs](https://lumenfer.com/pricing/ "Lumenfer bare-metal plans and prices per 30 days"), and the migration case is in [whether to move from AWS to bare metal](https://lumenfer.com/aws-to-bare-metal/ "Should I move from AWS to bare metal servers?").

### Three designs on Lumenfer plans

| Design | Servers | Example plan | Per server, 30 days | Per server, 12 months | Good for |
| --- | --- | --- | --- | --- | --- |
| Kubernetes on bare metal | 3 | Lumen 8: AMD Ryzen 7 9700X, 64 GB DDR5 | $109.37 | $918.70 | Production with local NVMe |
| Kubernetes VMs on one host | 1 | Lumen 24: AMD EPYC 9254, 256 GB DDR5 ECC | $356.25 | $2992.50 | Staging, labs, Talos trials |
| Kubernetes VMs on a Proxmox cluster | 3 | Lumen 16: AMD Ryzen 9 9950X, 128 GB DDR5 | $205.27 | $1724.26 | Several clusters with snapshots |

## Kubernetes on bare metal or VMs: FAQ

### Is Kubernetes on bare metal faster than on VMs?

It removes the hypervisor layer, but the gap depends on the workload. Lumenfer publishes no benchmark of it. Run the same deployment on a bare node and on a VM, then decide.

### Can I run Kubernetes on Proxmox VE?

Yes, inside VMs. Proxmox VE runs the VMs; kubeadm, k3s or Talos runs Kubernetes on top. The Proxmox VE 9 image is offered at checkout.

### How many servers do I need for highly available Kubernetes?

Three. Kubernetes asks for at least three stacked control-plane nodes, and k3s for an odd number of servers, three or more. On VMs, that means three hosts.

### Should I use k3s or Talos in VMs?

Either works. k3s installs over SSH on any of our Linux images. Talos is immutable, managed only through its API, and boots from an ISO that a Proxmox VM can mount.

### Do Kubernetes nodes need ECC memory?

Stateless pods rarely care; etcd and databases benefit. The EPYC plans list 32 GB DDR5 ECC, 256 GB DDR5 ECC and 384 GB DDR5 ECC. See [whether you need ECC RAM](https://lumenfer.com/ecc-ram-dedicated-server/ "Do I need ECC RAM on a rented dedicated server?").

**Pick bare metal or Proxmox VE at checkout** Same machines, either design: a person sets each one up once your crypto payment confirms. No KYC. [Order Lumen 16](https://lumenfer.com/app/checkout/?plan=lumen-16&lang=en) · [Compare plans](https://lumenfer.com/pricing/ "How much does a Lumenfer bare-metal server cost?")

## Sources

1. [Options for Highly Available Topology](https://kubernetes.io/docs/setup/production-environment/tools/kubeadm/ha-topology/) — Kubernetes documentation (read 2026-11-01)
2. [Considerations for large clusters (110 pods per node)](https://kubernetes.io/docs/setup/best-practices/cluster-large/) — Kubernetes documentation (read 2026-11-01)
3. [High Availability Embedded etcd](https://docs.k3s.io/datastore/ha-embedded) — K3s documentation (read 2026-11-01)
4. [Hardware recommendations](https://etcd.io/docs/v3.6/op-guide/hardware/) — etcd documentation (read 2026-11-01)
5. [Proxmox (Talos v1.14: VM sizing, CPU type)](https://docs.siderolabs.com/talos/v1.14/platform-specific-installations/virtualized-platforms/proxmox) — Sidero Labs documentation (read 2026-11-01)
6. [Boot Assets (Talos v1.14)](https://docs.siderolabs.com/talos/v1.14/platform-specific-installations/boot-assets) — Sidero Labs documentation (read 2026-11-01)
7. [What is Talos Linux? (API-managed, immutable)](https://docs.siderolabs.com/talos/v1.14/overview/what-is-talos) — Sidero Labs documentation (read 2026-11-01)
8. [Software-Defined Network (VXLAN zones, MTU, WireGuard)](https://pve.proxmox.com/pve-docs/chapter-pvesdn.html) — Proxmox VE documentation (read 2026-11-01)
9. [Cluster Manager (corosync network requirements)](https://pve.proxmox.com/pve-docs/chapter-pvecm.html) — Proxmox VE documentation (read 2026-11-01)

---

Lumenfer: Lumenfer rents current-generation bare-metal servers in Amsterdam, Reykjavik, Zurich and Bucharest, paid in crypto with no KYC: no ID, no selfie, no documents. Every machine has a CPU launched in 2022 or later, DDR5 memory and NVMe drives, and is rented whole by one customer with root access. It is run by DomaineGoat LLC, a Panama company that ignores DMCA notices, answers only to Panamanian law, allows Tor exit relays and keeps no logs. On the 30-day term, every plan is priced at least 25% under a named comparable current-generation server from 13 no-KYC crypto hosts, read on 30 October 2026.