After seven years running Proxmox in my homelab, I'm tackling my most complex project yet—prototyping a complete SMB infrastructure with 11+ VMs, 8 network segments, and comprehensive automation.
Why Prototype Virtually?
Testing infrastructure in VMs before production deployment catches problems early:
Network misconfigurations discovered in VLANs before buying physical switches
Resource constraints identified before ordering hardware
Backup failures found during testing instead of disasters
Automation issues debugged in isolated environments
Disaster recovery practiced without actual disasters
The Infrastructure
Network Design (8 VLANs)
10.0.100.0/24 – Management and monitoring
10.0.110.0/24 – Base infrastructure servers
10.0.120.0/24 – Application servers
10.0.130.0/24 – General workstations
10.0.131.0/24 – Manager workstations
10.0.132.0/24 – IT workstations
10.0.140.0/24 – Guest Wi-Fi and IoT
10.0.150.0/24 – Public-facing services
All inter-VLAN routing handled by an OPNSense VM. This lets me test firewall rules, routing policies, and network segmentation before deploying to physical infrastructure.
User Management
Proxmox supports PAM and Proxmox VE users. I use both:
PAM admin: SSH access + web UI (root disabled)
PVE users: Limited permissions, no shell access
Ansible user: API-only access for automation
Pro tip: Shut down pveproxy when not using the web UI:
systemctl stop pveproxy # Stop when not needed
systemctl start pveproxy # Start when needed
Resource Pools (Used Correctly)
I previously misused resource pools as tags. They're actually for delegation and access control:
smb-servers: Core infrastructure
smb-workstations: Desktop/laptop VMs
smb-project-admin@pve: Full access across pools
smb-admin@pve: Server pool only
Backup Strategy
Dual-layer backups for redundancy:
Local (10TB): 7 daily, 4 weekly, 2 monthly
External (4TB): 1 daily, 2 weekly, 1 monthly
And I actually test restore procedures. Backups are worthless if you've never validated them.
Automation
Ansible user with Proxmox API access enables infrastructure as code:
- VM provisioning from templates
- Network configuration (VLAN assignments)
- Resource management
- Backup scheduling
Configurations stored in GitHub—destroy everything and rebuild from source.
Is This Overkill?
For a homelab? Yes. For learning? Absolutely not.
Overengineering in the lab teaches enterprise concepts (VLANs, RBAC, disaster recovery, automation) without production risk. When something breaks, fixing it builds troubleshooting skills.
Hardware
Dual-socket Lenovo D20 (24 cores) with:
- CPU host passthrough for VMs
- Memory ballooning across VMs
- Initial allocation: 4GB servers, 8GB workstations
What's Next
I'm documenting the full build over 3-6 months:
- SMB infrastructure planning
- Ansible automation setup
- Samba Active Directory deployment
- File and print services
- Linux workstation configuration
- SELinux hardening
- Monitoring and backup automation
Read the Full Article
Complete details on my Proxmox prototyping methodology: Prototyping a Larger Project with Proxmox