LUCKVM / CLOUD INFRASTRUCTURE

Explore LuckVM
Home Global Acceleration Domains Support News Company
LUCKVM / ARCHITECTURE

Design a multi-tenant Proxmox hosting platform

Plan tenant isolation, IP allocation and recovery before choosing a virtualization host. This is an architecture checklist, not a benchmark or a promise that a specific IP block is included.

Define the failure and trust boundaries

Separate tenant traffic, hypervisor management, storage and cluster communication. Restrict the management plane to a VPN or trusted administration network; never expose the Proxmox API or web console to tenant VLANs.

Size compute and storage for sustained load

Budget host overhead and peak tenant memory before applying CPU overcommit. Measure random I/O, write latency and backup traffic with representative workloads. Reserve headroom for a host failure; adding nodes alone does not create usable failover capacity.

Build an address plan before ordering IPv4

A /24 contains 256 addresses; usable tenant addresses depend on routing, gateways and provider reservations. An offer with 125 IPv4 addresses is not automatically a /24 allocation. Confirm prefix ownership, routing mode, reverse DNS, IPv6 and IP reassignment rules before checkout.

Choose routing and isolation deliberately

Decide whether guests use routed public addresses, VLAN isolation or NAT. Validate permitted MAC addresses and anti-spoofing rules with the provider. CN2 is a transit-route characteristic, not a tenant-isolation mechanism. Confirm whether BGP sessions and customer prefixes are actually supported before designing around them.

Design quorum and recovery separately

A two-node cluster needs an explicit quorum strategy such as an external QDevice. Shared or replicated storage, fencing and spare compute capacity must be designed together. Test restoration from independent backups: high availability is not backup.

Define a repeatable acceptance test

Create isolated test tenants and verify VLAN boundaries, firewall defaults, CPU and memory limits, disk quotas, IPv4/IPv6 reachability and reverse DNS. Record peak-hour route tests, backup duration and restore time before accepting production tenants.