OVHcloud Managed Kubernetes

What OVHcloud Managed Kubernetes actually is

OVHcloud Managed Kubernetes Service (MKS) is a managed Kubernetes control plane running on OVHcloud Public Cloud: OVHcloud installs, upgrades, and operates the API servers and etcd, and you supply the worker nodes, which are ordinary Public Cloud instances you pay for by the hour or month. It is available in OVHcloud's Public Cloud regions — that means France (Gravelines, Strasbourg), Germany, the UK, and Poland in Europe, plus Canada, the US, and Asia-Pacific locations such as Singapore, Mumbai, and Sydney — not just the four EU countries our database listed.

The product has two tiers, and the tiering is the single most important thing to understand. The Free plan gives you a managed control plane at zero cost, permanently: a 99.5% monthly availability commitment, up to 100 nodes per cluster, and a 400 MiB etcd quota on a shared, single-AZ control plane. The Standard plan costs $0.099 per cluster per hour, moves you to a dedicated control plane with cross-availability-zone resilience, up to 500 nodes per cluster, an 8 GiB etcd quota, and rolling Kubernetes upgrades, and targets a 99.99% SLA in 3-AZ regions. OVHcloud's own wording is careful: the 99.99% figure is "offered after the beta phase" for Standard, and the legal SLA is explicit that because the Managed Kubernetes Service is provided free of charge, no service credits apply to its 99.50% commitment — the instances behind your nodes carry their own SLAs (99.99% for Public Cloud instances).

In our view this is one of the quieter strong options in /category/kubernetes-managed, aimed at teams that want Kubernetes without operating a control plane and without paying EKS/AKS/GKE-style control-plane fees.

Pricing: what you really pay

The control plane is free on the Free plan and $0.099/hour on Standard — roughly $72 a month per cluster if you leave it running. Everything else is Public Cloud consumption. A worked example from a 2026 test of the Free tier: three D2-8 worker nodes (4 vCPU, 8 GB RAM, 50 GB NVMe each) at €25.92/month per node (€0.0355/hour), plus an Octavia load balancer at €9.99/month, for about €87.75/month total for a three-node HA cluster with a free control plane. Smaller flavors cost less; bigger and GPU flavors cost a lot more. Block Storage volumes and public IP addresses are billed separately, and private networking over vRack is included.

There is also a real trial path: new Public Cloud projects get $200 in credit, enough to run that three-node cluster for roughly a month before paying anything. That is a better onboarding story than most managed Kubernetes services offer, and it is what we point readers toward when they ask whether MKS is worth evaluating.

Two honest caveats on the money. First, the headline "free control plane" understates total cost: the bill is dominated by nodes, storage, and load balancers, exactly like every other managed Kubernetes service. Second, the Standard plan's per-cluster fee is billed while the cluster exists, so staging clusters you forget about add up quietly. Turn idle clusters down; the API and Terraform provider make that scriptable.

Performance & infrastructure

Underneath, MKS runs on the same Public Cloud fleet as everything else OVHcloud sells: KVM-virtualized instances with NVMe-backed storage, 100 Tbps of network capacity behind them, and the always-on Anti-DDoS infrastructure that covers every OVHcloud product at no extra cost. Kubernetes versions track upstream with a documented support window (recent guides show 1.30 through 1.34 supported, chosen at cluster creation). The feature set is genuinely modern: cluster and node-pool autoscaling, rolling updates, OIDC-based access control, security update policies, persistent volumes backed by Block and Object Storage classes (LUKS-encrypted), Public Cloud Load Balancer integration, and private networking on vRack with separate node and load-balancer subnets.

The automation story is better than the marketing. There is an official Terraform provider, Pulumi and CDK for Terraform support, plus the OVHcloud API and CLI; a cluster and its node pools define cleanly in infrastructure-as-code, which is how we would run it. The control plane being managed means upgrades and etcd operations are OVHcloud's problem; node OS upgrades remain yours, via managed node-pool upgrade flows.

We have not benchmarked MKS in our own April 2026 test round, so we will attach no numbers to node performance — the worker nodes are Public Cloud instances, and their throughput is a function of the flavor you pick. Compare instance-level figures in /benchmarks before sizing a production pool.

Regional design is the practical differentiator versus the hyperscalers: you can keep clusters, control planes, and data inside EU regions with a European provider, which matters under GDPR and for Schrems-II-era data-residency arguments. The Standard plan's 3-AZ option covers the high-availability case in regions that have multiple availability zones; the Free plan is single-AZ and should be treated accordingly.

Worth comparing head-to-head with the hyperscalers: AWS, Google, and Azure all charge for the control plane or wrap it into per-cluster fees that compound across environments, and their instance prices for comparable vCPU and memory are generally higher in EU regions. MKS does not win on features against those platforms — it wins on the arithmetic of running several clusters in Europe without paying rent on each control plane. If your cluster count is one and your budget is generous, the difference shrinks and the hyperscaler ecosystem starts to matter more.

Where it falls short

The control panel is the most common complaint we read, and it matches our own impression of OVHcloud's console generally: functional and complete, but denser and less forgiving than DigitalOcean or GKE. Kubernetes-specific UX (cluster dashboards, node-pool management screens) is workable rather than pleasant; most serious users end up living in kubectl and Terraform instead.

Support is OVHcloud's standard ticket-based model unless you buy a premium tier. For a platform service where an incident at 3 a.m. is an outage for your customers, that is a real consideration: compare it with the enterprise support packages available from AWS, Google, or Azure before betting production on it. The Standard plan's SLA wording also deserves scrutiny — 99.99% is promised "after the beta phase", and the free tier's 99.5% commitment comes with no credits at all.

The ecosystem is thinner than the hyperscalers: no first-party marketplace of add-ons, fewer deep integrations (identity, observability, service mesh) than EKS or GKE, and multi-cluster management is not offered in-product. And the etcd limits bite earlier than you expect: 400 MiB on the Free plan is fine for modest CRD usage, but teams running operators with fat custom resources should model that before committing.

Finally, node lifecycle remains your job: OS patching cadence, drain strategies, and right-sizing are on you. That is normal for managed Kubernetes, but readers moving from a fully hosted PaaS sometimes expect otherwise.

Day to day, expect kubectl and Terraform to be the interface rather than the console. That suits the audience MKS is built for — platform engineers who already treat clusters as code — and it is why the official provider and the clean node-pool resources matter more to us than the dashboard's looks. Teams that need a guided, click-ops Kubernetes experience will be happier elsewhere.

Who should pick this

We pick MKS for EU-resident workloads that want managed Kubernetes economics without control-plane rent: cost-sensitive startups, agencies running client workloads on vRack-private networks, and platform teams already standardized on Terraform. The Free plan is, as far as we know, the most generous permanent free control plane among European providers — for staging, side projects, and learning clusters it is the obvious choice, and the $200 trial credit covers the first month of real nodes.

We would not pick it for teams that need deep managed-service integration (identity, observability, marketplace add-ons) or an enterprise support contract on the standard plan. Multi-cluster fleets at scale will likely outgrow the console and want a hyperscaler's tooling. And if your requirement is a contractual 99.99% today, read the fine print on the Standard plan's beta wording before signing.

Key takeaways

• Managed control plane on OVHcloud Public Cloud; you pay for worker nodes, storage, and load balancers.
• Free plan: $0 control plane forever, 99.5% availability commitment, up to 100 nodes, single AZ.
• Standard plan: $0.099/hour per cluster, dedicated multi-AZ control plane, up to 500 nodes, 99.99% target after beta.
• Real cost example: three D2-8 nodes plus a load balancer land around €87.75/month with a free control plane.
• Strong automation: official Terraform provider, Pulumi, API/CLI, autoscaling, vRack private networking.
• Our pick for EU-resident, Terraform-managed clusters on a budget; not for teams wanting hyperscaler-grade managed ecosystems.

Sources

• https://www.ovhcloud.com/en/public-cloud/kubernetes/
• https://us.ovhcloud.com/legal/sla/public-cloud/
• https://itechguides.com/best/container-orchestration-software/ovhcloud-managed-kubernetes
• https://eucloudcost.com/blog/ovhcloud-cluster
• https://docs.ovhcloud.com/en/guides/public-cloud/containers-orchestration/managed-kubernetes/mks-plans.md
• https://us.ovhcloud.com/security/anti-ddos