1 minute read

Scope and Assumptions

This article builds directly on Part 1. Nothing introduced here replaces earlier guidance.

Assumptions:

  • Three fresh Proxmox nodes
  • No existing VMs
  • No existing cluster
  • Managed switch + stateful firewall already in place

If your cluster already exists or hosts VMs, do not follow this guide. That scenario is covered in Part 3.


VLAN Architecture (Expanded)

VLAN Name Purpose
90 Management Web UI, SSH
99 Cluster Corosync (L2 preferred)
200+ VM Guest workloads

Key Rules

  • Corosync must not traverse user or VM networks
  • Management traffic must remain routable and firewalled
  • VM traffic must never touch VLAN 90 or 99

Logical Topology

Cluster Topology


Host Addressing

Host Mgmt (VLAN 90) Cluster (VLAN 99)
hv1 192.168.90.11 192.168.99.11
hv2 192.168.90.12 192.168.99.12
hv3 192.168.90.13 192.168.99.13

Switch Configuration

Each Proxmox port:

  • Tagged VLANs: 90, 99, 200+
  • Native VLAN: None

This ensures deterministic traffic separation.


Firewall Responsibilities

Critical Requirement

If your firewall cannot enforce inter-VLAN policy, this architecture collapses.

A “VLAN-capable” router without firewall rules provides zero isolation.


Proxmox Network Configuration (All Nodes)

Management Bridge

auto vmbr0
iface vmbr0 inet static
    address 192.168.90.11/24
    gateway 192.168.90.1
    bridge-ports eno1.90

Cluster Interface

auto vmbr99
iface vmbr99 inet static
    address 192.168.99.11/24
    bridge-ports eno1.99

VLAN 99 should ideally be non-routed.


Cluster Creation

From hv1:

pvecm create secure-cluster --bindnet0 192.168.99.0

Join hv2 and hv3 using their VLAN 99 addresses only.


Threat–Mitigation Matrix

Threat Mitigation
User device compromise Mgmt VLAN isolation
VM breakout Separate VM VLANs
Corosync instability Dedicated cluster VLAN
Lateral movement Firewall policy enforcement
Mis-tagged traffic No native VLAN

Common Pitfalls

  • Using management IPs for Corosync
  • Allowing VLAN 99 to route
  • Forgetting to firewall east–west traffic

Diagram


What’s Next

Part 3 covers the hardest scenario:

  • Existing cluster
  • Live VMs
  • Native VLAN dependency
  • Zero-downtime migration strategy

End of Part 2