Kubernetes, with the control plane run for you.

We run the API server, etcd and the upgrades, and a cluster is ready about 4 minutes after the command. Each node is a cloud server at its usual price.

From €3.49 a month

Features

A cluster that fits the rest of the project.

  • Versions 1.35 to 1.37

    CNCF-conformant clusters on the three minor versions upstream supports. Patch releases install in a maintenance window you choose.

  • Node pools and autoscaling

    Group nodes in pools by plan, shared, dedicated or GPU. The cluster autoscaler adds and removes nodes within the limits you set.

  • Services become load balancers

    A Service of type LoadBalancer creates a load balancer and keeps its targets in step with the nodes.

  • Volumes and file storage

    The CSI driver provisions volumes for ReadWriteOnce claims and file storage for ReadWriteMany, and snapshots both.

  • Cilium on a private network

    Nodes talk over a private network with Cilium as the CNI, which brings network policies and eBPF load balancing.

  • Upgrades a node at a time

    A minor upgrade replaces each node in turn. It drains the pods, starts a fresh node and only then removes the old one.

Pricing

A cluster costs what its nodes cost.

The standard control plane is not billed. Each node costs what a cloud server of its plan costs.

Shared vCPU

PlanvCPUMemoryNVMe diskAn hourA month, at most
s-112 GB40 GB€0.0056€3.49
s-224 GB80 GB€0.0100€6.49
s-448 GB160 GB€0.0190€12.49
s-8816 GB320 GB€0.0370€24.49
s-161632 GB640 GB€0.0730€48.49

Dedicated vCPU

PlanvCPUMemoryNVMe diskAn hourA month, at most
d-228 GB80 GB€0.0230€14.90
d-4416 GB160 GB€0.0450€29.90
d-8832 GB240 GB€0.0890€59.90
d-161664 GB360 GB€0.1770€119.90
d-3232128 GB600 GB€0.3530€239.90
Kubernetes, high-availability control planeper cluster
€29.90 a month, at most€0.0450 an hour

A high-availability control plane runs three API servers and three etcd members on separate hosts. GPU node pools cost what GPU servers cost.

Code

A cluster, then kubectl.

lc k8s creates the cluster and writes its kubeconfig. From there, every tool that speaks to Kubernetes works unchanged.

  • Node pools by plan, with a minimum and a maximum size.
  • A kubeconfig with a token scoped to the one cluster.
  • The same cluster as a Terraform resource, when you prefer code.
shell
$ lc k8s create shop --version 1.37 --region pt1 --ha \
    --pool web --plan s-4 --min 3 --max 6
Creating shop in pt1 ... ready
$ lc k8s kubeconfig shop > ~/.kube/config
$ kubectl get nodes
NAME          STATUS   ROLES    AGE   VERSION
shop-web-7k2  Ready    <none>   2m    v1.37.1
shop-web-9fq  Ready    <none>   2m    v1.37.1
shop-web-c4m  Ready    <none>   2m    v1.37.1

Questions

Questions about Kubernetes.

FAQ
What does the standard control plane leave out?
It runs as one replica, which restarts on another host if its own host fails. The API is out of reach for up to 5 minutes, while your pods keep running. The high-availability control plane runs three replicas on separate hosts and keeps serving through the loss of one.
How long is each version supported?
We support the three minor versions upstream maintains, 1.35 to 1.37. When a version leaves upstream support, its clusters upgrade in their maintenance window after a month’s notice.
Can one cluster’s nodes run in several regions?
No. A cluster’s nodes all run in the cluster’s own region. Reach databases and servers in other regions over a private network, which spans all of them.

Start with one server and one command.

Choose a region and a plan, and add an SSH key. The server answers SSH 38 seconds after the command, at the price on its spec sheet.

logi-ciel is a design concept

This site is a portfolio piece by Soheyb. The brand, the pages and the price list are real design work; the cloud behind them is not. There’s no console to sign in to, and no server to rent.

Back to the site