Managed Kubernetes

Vanilla.
Upstream.
Yours.

No fork. No proprietary CRDs. No "AKS-flavored" anything. We run the same Kubernetes you'd install yourself — patched, scaled, and HA — and we get out of your way.

Spin up a cluster Reference architecture

1.32 · 1.31 · 1.30·CNCF certified·Cilium · Talos

Cluster · prod-euhealthy

v1.32.1 · 3-AZ control plane

kube-system ✓ 9 / 9

cilium ✓ 18 / 18

cert-manager ✓ 3 / 3

ingress-nginx ✓ 6 / 6

Nodes9 / 12 ready

Pods218 running

API server p997 ms

etcd commit p994 ms

Built for production

Pretend the control plane doesn't exist.

HA control plane

Three etcd members across three AZs. We handle quorum, snapshots, etcd-backup-rotation. You don't think about it.

Cilium, default

eBPF networking, kube-proxy replaced, ClusterMesh between regions. Hubble UI free in the console.

Talos node OS

Minimal, immutable, API-driven. No sshd, no apt, no surprise drift. Patched on a schedule you control.

Karpenter built in

Bin-packed nodes, mixed-instance fleets, spot interruption handled. Your bill matches your actual workload, not your worst hour.

Pod-level isolation

Kata-runtime available cluster-wide for untrusted workloads. Hardware virtualization at pod boundary, no perf cliff.

Honest pricing

€72/month per cluster for the control plane. Worker nodes billed as Compute. No hidden API-call fees.

In code

From `arv k8s create` to `kubectl apply` in 90 seconds.

$ arv k8s create prod-eu \
    --version 1.32 \
    --region  eu-hel-1 \
    --pool    "default:a8:6:6:18" \
    --cni     cilium

# control plane up: 84s · workers ready: 92s
$ arv k8s kubeconfig prod-eu > ~/.kube/prod-eu

$ kubectl --kubeconfig ~/.kube/prod-eu get nodes
NAME                       STATUS   ROLES    AGE   VERSION
prod-eu-default-1a-001     Ready    worker   34s   v1.32.1
prod-eu-default-1b-001     Ready    worker   34s   v1.32.1
prod-eu-default-1c-001     Ready    worker   33s   v1.32.1
…

resource "arveon_kubernetes_cluster" "prod" {
  name    = "prod-eu"
  version = "1.32"
  region  = "eu-hel-1"
  cni     = "cilium"

  node_pool {
    name     = "default"
    plan     = "a8"
    min_size = 6
    max_size = 18
    spread   = "az"
  }
}

Customer story · Cardinal Data

"We left a managed K8s service that kept charging us for API calls. Arveon's per-cluster price means we stopped doing capacity math and started shipping."

Nadia Kowalski

Principal SRE · Cardinal Data

Production clusters

3.4k

Pods average

Surprise upgrades

7ms

API server p99

Vanilla.Upstream.Yours.