Run Proxmox on Hetzner
Single node, cluster, HA & support

Yes. Proxmox VE can run perfectly fine as a single node and is a typical starting point before potentially clustering later. What matters is proper operations: backups, monitoring, updates, and a solid restore plan.

Single Node means: Everything runs on one host – there's no automatic failover if hardware fails. An HA cluster consists of multiple nodes, so VMs/containers can start on another node in case of failure (if design/storage allows).

For stable quorum and proper HA, 3+ nodes is the standard recommendation. Two nodes are theoretically possible but require an additional quorum concept (e.g., QDevice) and are more error-prone in operation.

Proxmox itself is open source; costs arise mainly from hardware (Dedicated/Cloud) and your desired operational scope (setup, monitoring, backups, support). We calculate transparently based on RAM/CPU/Storage and your requirements (RPO/RTO, SLA, growth).

Our standard is: Cluster/storage traffic doesn't belong on the public internet. We plan a private network (e.g., separate VLAN/private segment) plus firewalling and minimize open management ports.

Yes – at Hetzner this works via vSwitch, which can "connect" Dedicated Servers to Hetzner Cloud Private Networks. This allows Cloud and Dedicated resources to communicate over private links.

At minimum, we separate Management (GUI/SSH/API) from Cluster/Corosync. If Ceph or other storage is added, you ideally need a third, separate storage network – this significantly improves performance and stability.

For Single Node, ZFS is often the sweet spot: fast, stable, simple. For HA clusters, Ceph is interesting because it provides distributed storage – but requires more planning (network, disks, sizing) and "operational discipline".

Rule of thumb: plan for 2-3x your current VM data needs plus backup buffer and growth. With Ceph, replication adds up (more raw capacity needed), for backups factor in retention and change rate.

In practice: Dedicated for workloads/storage + optionally Cloud for bastion/monitoring/edge services. This leverages Hetzner's strengths and keeps the design clean.

We start with IP/hostname planning (important for clustering), install Proxmox VE, harden access and set up network + storage + backup according to your target architecture.

Management access runs via VPN/bastion, not "wildly" open on the internet. Additionally: least-privilege accounts, 2FA where possible, logging and clear emergency access.

Single Node: good Dedicated server + ZFS + solid backup + monitoring. HA: 3+ nodes + separated networks + clear storage concept (Ceph or shared storage).

We define RPO/RTO, set up automated backups, regularly test restores and build in offsite/immutable as needed. The goal is: Restore is a practiced process, not hope.

Single Node: planned in maintenance window with rollback plan. Cluster: Rolling updates node-by-node, so workloads (where possible) keep running.

We do this as a planned cutover: Assessment → Test migration → Performance/restore tests → final move. For VMware exits, details depend heavily on storage/network/guest OS.

The most common: poor network design (everything in one network), missing restore tests and unclear update process. That's exactly why we deliver runbooks + monitoring + clear separation.

Yes – typically as hour packages (for "when it burns" and further development) or as retainer/managed operations (predictable, SLA-based).