Managed Proxmox · Ceph · AS47263

High-availability infrastructure for critical systems, operated in Germany

From three nodes. Three copies of every block. One network of our own.

We design, run and monitor high-availability Proxmox clusters for companies with business-critical applications: automatic failover, Ceph storage with three replicas, 24/7 monitoring with on-call duty, a dedicated contact and a published fixed price.

  • AS47263our own network
  • 3×replication per data block
  • 24/7monitoring & on-call duty
  • from €449.99net per month
Diagram 01 Failover in a three-node cluster
  • Active traffic
  • Replication
  • Node active
  • Node failed
If node B fails, the Proxmox HA Manager automatically restarts the affected virtual machines on node A or C. Services running on several nodes behind the load balancer remain available.
Diagram as text
  1. The load balancer distributes requests across nodes A, B and C.
  2. Ceph keeps every data block in three replicas on different nodes.
  3. Node B fails; the load balancer removes it from rotation.
  4. The Proxmox HA Manager restarts the virtual machines from node B on node A or C.

Facts you can verify

In brief

High availability hosting in Germany, explained in brief

Short answer

Our high availability hosting in Germany runs on managed HA clusters: at least three servers (nodes) that work as one system. If a node fails, its virtual machines restart automatically on another node. We handle setup, availability, updates, monitoring and support for a fixed monthly fee; you remain responsible for your application.

managed-cluster.net is a brand of Cube-Hosting.net (owner: Jesse Röhsler, Gerolzhofen, Germany). We operate managed Proxmox and high-availability clusters in German data centers (Maincubes FRA01, NBG01 Nuremberg) on our own network, AS47263.

At least three nodes ensure that a majority, the quorum, can still make decisions when one node goes down. You do not need your own hardware, data center or on-call rotation: platform, network and operations all come from one provider under one contract.

Our company page shows who is behind managed-cluster.net.

Who it is for

Hosting for business-critical applications: four typical starting points

managed-cluster.net is built for companies whose applications cannot simply go offline. The starting points differ, but the operations behind them are the same: 24/7 monitoring, on-call duty and written documentation.

Takeover

Hand over an existing Proxmox cluster

After an audit, we connect your cluster to our monitoring and take over operations, updates and on-call duty, moving it step by step toward the target architecture if you wish.

Have your existing Proxmox cluster managed

VMware migration

From VMware to Proxmox VE

We migrate to a Proxmox cluster built in parallel and convert ESXi VMs to KVM; the switchover per VM typically takes a matter of minutes. After the analysis, you receive a fixed-price proposal for the migration and ongoing managed operations.

Proxmox vs VMware compared

More use cases: Shopware & Magento · Managed servers

Services

Four service areas, one operations contract

Operations at managed-cluster.net are organized into four service areas across the lifecycle of your cluster. All four belong to one operations contract at a fixed monthly price with no à-la-carte add-ons, and you have a dedicated contact throughout.

Design & setup

Requirements analysis, sizing with headroom, and a redundancy design with N+1 capacity and quorum planning. Before handover we test failover and document the emergency plan.

High availability cluster design

Operations & monitoring

Monitoring from hardware up to the application and 24/7 alerting with an escalation chain. Every incident ends with a post-incident report, every month with an operations report.

24/7 infrastructure monitoring

Updates & security

Security patches with priority for critical vulnerabilities, rolling updates node by node and system hardening. Each update runs on one node first and is checked there before the remaining nodes follow.

Managed patch management

Failover testing & runbooks

Failover tests before go-live and regularly during operations, each with a written record. Emergency plans cover scenarios beyond automatic failover.

Failover testing with written records

All services

Architecture

Reference architecture: the high-availability three-node cluster

The managed-cluster.net reference architecture has three layers: redundant firewalls and load balancers with health checks, three Proxmox VE nodes in one cluster, and Ceph storage with three replicas of every data block. Every layer is redundant. We discuss the hardware of the cluster nodes (CPU, RAM, storage, network) individually and define it in the design proposal based on your load profile.

Diagram 02 Access, nodes, storage
  • active traffic
  • replication
  • external
  • node active
Three-node Proxmox cluster with a redundant firewall and load balancer layer and Ceph storage across all nodes.
Diagram as text
  1. Requests from the internet reach the floating IP of the redundant firewall and load balancer layer.
  2. The load balancers distribute requests across nodes 1, 2 and 3 and continuously check their health.
  3. The nodes form a cluster with quorum via Corosync; the Proxmox HA Manager monitors the virtual machines.
  4. Ceph spreads storage across all nodes and keeps every data block in three replicas.

Together, quorum and fencing, the safe isolation of a failed node, prevent a split brain, the state in which two parts of a cluster keep working independently of each other. Our guide to split brain, quorum and fencing explains how the two work together.

The redundant load balancers share a floating IP, continuously check the health of the nodes and take failed systems out of rotation. Management, storage and public traffic run on separate networks (VLANs).

The Essentials and Business packages use Ceph; we use ZFS replication in individually designed, more compact setups. How platform, storage and network fit together is described on our managed Proxmox cluster page.

What happens when a node fails

  1. A NODE FAILS, for example due to a hardware fault or a power supply failure.
  2. Corosync detects the failure; the remaining nodes keep quorum.
  3. The Proxmox HA Manager restarts the affected VMs on another node.
  4. The on-call alert goes out, the hardware is replaced, and you receive a post-incident report.

The load balancer removes a failed node from rotation within seconds. How long the virtual machines take to restart depends on fencing and on the guest system's boot time.

Step-by-step failover

If a disk or an entire node fails, no data is lost: Ceph keeps every block in three copies on different nodes and restores the target redundancy on its own as soon as enough disks and nodes are available for it. Recovery increases I/O load; with several simultaneous failures, some data may be temporarily unavailable until recovery completes. Our page on Ceph storage with three replicas covers the technical details.

Infrastructure

Managed Proxmox hosting in Germany: data centers and our own network

Your systems run exclusively in German data centers: the main site is Maincubes FRA01 in the Rhine-Main region, the second site NBG01 in Nuremberg. They are connected through AS47263, Cube-Hosting's autonomous system with its own routing via the Border Gateway Protocol (BGP).

Data centers FRA01 and NBG01

According to its operator, Maincubes, the FRA01 data center is certified to ISO/IEC 27001 and ISO 9001; the certifications belong to the operator, not to managed-cluster.net. Also according to the operator, FRA01 has a redundant, uninterruptible power supply as well as multi-level access control and a security service around the clock.

Source: FRA01 site page on maincubes.com, as of October 2026.

  • At FRA01, defective components are replaced via remote hands, which the operator states are available around the clock.
  • NBG01 in Nuremberg holds backup copies outside the main site and is the second site for multi-site deployments.
  • On request, we name the site including its address in your contract.

Our data centers FRA01 and NBG01

Network AS47263

AS47263 is Cube-Hosting's network, with fixed IP address space (several IPv4 prefixes and native IPv6) and redundant core routers.

  • Peering at ERA-IX Frankfurt, NL-ix and frankonIX, among others
  • Network-level DDoS protection
  • Prefixes and peerings publicly documented, for example on bgp.tools

Our own network AS47263

Security & evidence

Security in operations: what we cover

managed-cluster.net supports your security measures in four areas: incident handling, backup and recovery, updates and vulnerabilities, and network security. The excerpt shows our contribution and the evidence you receive in each area.

Scope of this overview

This overview describes how our operations support your security measures. It is neither a certification nor legal advice. Your company's legal and regulatory obligations, and your management's accountability for them, remain with you.

Responsibilities in operations (excerpt)
AreaOur contributionEvidence
Incident handling24/7 monitoring, alerting with escalation paths and on-call duty; analysis, containment and recoveryPost-incident report on cause and follow-up measures for every incident
Backup and recoveryBackup infrastructure separate from the cluster with defined retention periods, regular restore tests, documented emergency plansWritten records of failover tests, written emergency plans
Updates and vulnerabilitiesSecurity patches with priority for critical vulnerabilities, rolling updates, interim measures such as firewall rules where possible, system hardeningVersion-controlled, traceable configuration
Network securitySeparate networks via VLAN, firewalling on every layer, network-level DDoS protection; the Enterprise package adds VPN, IP allowlisting and role-based accessNetwork architecture in the documentation, publicly documented prefixes and peerings of AS47263

What stays with you is set out in the responsibility overview on our security and compliance page. How separate backups and regular restore tests work is covered on our managed backup and disaster recovery page.

Contract

What the operations contract and SLA define

At managed-cluster.net, binding commitments belong in the operations contract and its service level agreement (SLA). We agree your availability target, response times and recovery targets with you, which is why this website shows no blanket uptime percentages.

  • Availability targets matched to your application and its level of redundancy
  • Response times by priority level: critical alerts reach on-call duty immediately and are handled right away, around the clock.
  • Service hours and escalation paths for emergencies
  • Recovery targets: RTO (recovery time objective) for the maximum outage, RPO (recovery point objective) for the maximum tolerable data loss
  • Monthly operations report covering availability, incidents and utilization; the Business package adds SLA reports
  • Data processing agreement under Art. 28 GDPR for processing personal data on your behalf
  • German contract law for the operations contract, SLA and DPA
  • Data location: we process and store your cluster data exclusively in Germany.
  • Documentation belongs to you: architecture, access details and emergency plans are provided in writing.

Our availability calculator shows what an availability target means in downtime per year.

Pricing

Managed HA cluster pricing: packages and fixed prices

A managed HA cluster from managed-cluster.net comes at a fixed monthly price covering setup, operations, monitoring, updates and support. The published prices apply to the Essentials and Business packages. We price Enterprise environments by data volume, RPO target and switchover time and state the fixed price in the proposal.

Monthly, net Managed Proxmox cluster packages at a glance
PackageScopePrice, net per month
Essentials · Ceph HA3 nodes with quorum, Ceph HA storage with triple replication and self-healing, 24/7 monitoring and alerting, fully managed including updates, DDoS protection€449.99
Business · 2 sites6 nodes across FRA01 and NBG01, SLA reports covering availability and incidents, extended monitoring with escalation, backups with restore tests at the second site€1,099.99
EnterpriseIndividual RPO/RTO and SLA targets, separate networks, VPN, IP allowlisting, role-based access, multi-site architecture designed to your requirementson request

All prices are net, plus VAT. As of October 2026.

Not ready for a cluster yet? Dedicated managed servers start at €99.99 net per month. They run in the same data centers and on the same network as our clusters, with an upgrade path to a cluster.

Compare all packages and prices

FAQ

Frequently asked questions about high availability hosting

What is a managed HA cluster?

A group of at least three servers that works as one system and that we design, build, monitor and maintain. If a node fails, the Proxmox HA Manager restarts its virtual machines on another node. You look after your application and we take care of everything underneath it, at a fixed monthly price.

Which applications justify a high availability cluster?

Applications whose outage hits your business directly, such as line-of-business software, databases, portals or online stores. The cluster absorbs the failure of individual servers, disks and network paths; it replicates operator errors and malware as well, so a separate backup is still required. As long as your business can absorb a brief interruption after a hardware fault, a managed server from €99.99 net per month is enough.

What do you take care of, and what stays with us?

We take care of everything below your application: hardware, network, firewall, Proxmox VE and the cluster, storage, backup infrastructure, monitoring with on-call duty and platform updates. The data center operator provides the building, power and cooling; at FRA01 that is Maincubes. You are responsible for your application; availability and recovery targets are agreed jointly.

How quickly will I get an assessment and a proposal?

You receive an initial assessment within one business day; design and proposal usually follow within a few business days. Our contact page explains what happens after your inquiry.

Who is behind managed-cluster.net?

Cube-Hosting.net, owned by Jesse Röhsler and based in Gerolzhofen, Germany; managed-cluster.net is a brand of the company. Company profile, address and infrastructure are on our page about managed-cluster.net.

What does a managed HA cluster cost?

From €449.99 net per month for the Essentials package with three nodes and Ceph; Business with six nodes costs €1,099.99 net, and Enterprise is priced individually. Our pricing page compares what each package includes.

What happens if a server fails?

The affected virtual machines restart automatically on another node, and on-call duty is alerted. The full sequence is described on our high availability cluster page.

Does an HA cluster replace backups?

No. Replication also copies mistakes such as an accidental deletion to every replica at once. Backups therefore run on infrastructure separate from the cluster; see backup and disaster recovery.

Contact

Request an architecture call

Briefly describe your application and your requirements for availability and load. Getting from first contact to a cluster in operation takes three steps.

  1. Inquiry: You describe your application, load and availability requirements in a few sentences and receive an initial assessment within one business day, free and without obligation.
  2. Design and proposal: We clarify load profile, availability and recovery targets. Usually within a few business days, you receive a written design with an architecture diagram, a component list and a fixed price for setup and operations.
  3. Build, failover test, operations: Build including monitoring and backups and the failover test before handover typically take one to three weeks; you set the date for the migration. Day-to-day operations follow.

* Required field

What should run on the cluster? What are your requirements for availability, load or recovery?

By submitting, you agree to the processing of your details for the purpose of handling your inquiry. See the privacy policy for details. Alternatively, you can email us directly.