Control plane on confidential VMs
Three masters with embedded etcd, each on its own confidential VM. The processor encrypts their memory, and the disk unlocks only after attestation.
Managed Kubernetes (RKE2) with three masters on confidential VMs. The kubeconfig is sealed to a key made in your browser, and the masters refuse any node you didn't sign.
Memory encrypted by the processor (AMD SEV-SNP). The disk unlocks only after attestation.
Three masters with embedded etcd, each on its own confidential VM. The processor encrypts their memory, and the disk unlocks only after attestation.
The masters seal the kubeconfig to your cluster key, and it is decrypted in your browser. After launch, Hippius holds no readable kubeconfig, no certificate authority and no shell on the nodes.
Adding a node is signed in your browser with your key. The masters refuse any node you didn't sign, so Hippius can't add a machine of its own.
In the console: Cloud, Confidential Compute, Kubernetes, then New cluster. One form, price in the summary.
Pick the country the three masters run in, Compact (workloads on the masters) or Standard (workloads on worker pools), and the master size.
List the IPv4 ranges allowed to reach the API on port 6443, fixed once the cluster is created, then download your cluster key file and confirm you stored it.
hippius-cluster-prod-eu.keyLoad your key file to download the kubeconfig, then add worker pools of up to 50 identical nodes each, signed with your key.
kubectl get nodesYour cluster key is made in your browser, and only its public half ever reaches us. So a few things are simply not possible.
The masters seal it to your key. It is decrypted in your browser, and neither your key nor the kubeconfig is sent anywhere.
Every node is signed with your key. The masters check the signature and refuse any node you didn't sign.
Nobody, Hippius included, can reset it. Keep it backed up, for example in a password manager.
Services and jobs that handle data the servers they run on must not be able to read.
Workers talk over your private network and have no public IP. The API answers only the ranges you allow.
Daily Velero backups, on unless you turn them off at creation, are encrypted inside the cluster before they reach your own Object Storage, and restore into any Kubernetes cluster, on Hippius or not.
Billed per second from your balance, charged hourly: the VMs it runs on, plus $15 per cluster per month, whatever its size or number of workers.
Smallest control plane, 3 Small masters
$0.2986/h
About $218 a month: $0.263/h for the 3 masters, $0.015/h for their addresses and $0.0205/h for the cluster fee, before workers.
Monthly figures are estimates over 730 hours. Prices can change: the console's live price list and every create form show the current ones. See pricing
An RKE2 cluster (v1.34.3) with three masters running embedded etcd, each on its own confidential VM with a public IPv4 for the API. Nodes talk over your private network with the Canal network plugin. All three masters run in the country you choose.