MINISFORUM MS-A2 Migration Guide

Comprehensive guide for migrating from Intel NUC6i7KYK cluster to MINISFORUM MS-A2 systems with minimal downtime.

Table of Contents

Overview

This document provides a detailed migration plan from the current Intel NUC-based vSphere cluster to a new MINISFORUM MS-A2 AMD-based cluster, focusing on maintaining service continuity and optimizing performance.

Migration Objectives

  • Zero data loss: Comprehensive backup and validation procedures
  • Minimal downtime: Phased migration with rolling updates
  • Performance improvement: 4x compute capacity increase
  • Future readiness: Support for TKG/TAS workloads
  • Network enhancement: Leverage native 10G capabilities

Hardware Comparison

Current Platform: Intel NUC6i7KYK

Component Specification Total (3 units)
CPU Intel Core i7-6770HQ (4C/8T @ 2.6-3.5GHz) 12C/24T
Memory 64GB DDR4 each 192GB DDR4
Storage 250GB NVMe each 750GB total
Network 1x built-in GbE + 1x USB GbE 6x 1G ports
Form Factor 4.6” x 4.4” x 1.4” Mini desktop
Power ~35-50W each ~150W total

Target Platform: MINISFORUM MS-A2

Component Specification Total (3 units)
CPU AMD Ryzen 9 7945HX (16C/32T @ 2.5-5.4GHz) 48C/96T
Memory 32GB DDR5 each (expandable to 96GB) 96GB DDR5
Storage 1TB NVMe each + 2x M.2 slots 3TB+ potential
Network 2x SFP+ 10G + 2x 2.5G ethernet Native 10G
Form Factor 8.3” x 8.3” x 2.1” Rackmountable
Power ~60-80W typical, 129W peak ~240W typical

Performance Improvement Summary

  • CPU Performance: 4x cores, 2x per-core performance = ~8x total
  • Memory Bandwidth: DDR4-2400 → DDR5-4800 = 2x improvement
  • Storage: 4x capacity, similar NVMe performance
  • Network: 1G → 10G = 10x bandwidth
  • Efficiency: Better performance per watt

Migration Strategy

Phased Approach

Phase 1: Infrastructure Preparation (Week 1)

  • Install rack infrastructure
  • Deploy 10G networking
  • Setup first MS-A2 unit
  • Validate hardware configuration

Phase 2: vSphere Integration (Week 2)

  • Add MS-A2 to existing cluster
  • Configure mixed Intel/AMD cluster
  • Test EVC compatibility
  • Validate vMotion between platforms

Phase 3: Workload Migration (Weeks 3-4)

  • Begin VM migration to MS-A2
  • Validate application performance
  • Monitor for compatibility issues
  • Migrate storage-intensive workloads

Phase 4: Cluster Expansion (Weeks 5-6)

  • Deploy second MS-A2 unit
  • Continue workload migration
  • Begin decommissioning first NUC

Phase 5: Completion (Weeks 7-8)

  • Deploy third MS-A2 unit
  • Complete workload migration
  • Remove all NUCs from cluster
  • Optimize new configuration

Pre-Migration Preparation

Documentation and Assessment

Current Environment Inventory 

# vSphere cluster assessment
Get-Cluster | Get-VM | Select Name, NumCpu, MemoryGB, PowerState
Get-Cluster | Get-VMHost | Select Name, Version, Model

# Network configuration
Get-VMHost | Get-VirtualSwitch | ft
Get-VMHost | Get-VMHostNetworkAdapter | ft

# Storage configuration
Get-Datastore | Select Name, CapacityGB, FreeSpaceGB
Get-VM | Get-HardDisk | ft

Compatibility Validation

  1. EVC Mode: Determine compatible CPU feature set
  2. OS Compatibility: Validate guest OS support on AMD
  3. Application testing: Test critical applications on AMD
  4. Driver validation: Ensure hardware driver compatibility

Backup Procedures

Complete Environment Backup

  1. vCenter configuration: Full vCenter backup
  2. VM backups: Individual VM exports or snapshots
  3. Host configurations: ESXi configuration backups
  4. Network settings: Switch and router configurations
  5. Storage: NAS snapshot before migration

Recovery Testing

  • Test restore procedures on isolated environment
  • Validate backup integrity
  • Document recovery time objectives

Hardware Setup

MS-A2 Physical Installation

Rack Mounting

  1. Install rack shelf for MS-A2 unit
  2. Secure unit with appropriate mounting hardware
  3. Verify clearance for airflow and cable access
  4. Label unit for identification

Power Configuration 

# Power requirements per unit:
- Idle: 23-26W
- Typical: 60-80W
- Peak: 129W
- Power connector: Standard IEC C13

# Total cluster power:
- 3x MS-A2: ~240W typical, ~400W peak
- Plus network gear: ~40W
- Total: ~280W typical, ~440W peak

Network Configuration

10G Connectivity Setup 

# MS-A2 network interfaces per unit:
- 2x SFP+ 10G ports (primary networking)
- 2x 2.5G ethernet ports (management/backup)

# Connection plan:
Port 1 (SFP+): Primary 10G to USW-Aggregation
Port 2 (SFP+): Spare/future use
Port 3 (2.5G): Management network
Port 4 (2.5G): Backup/OOB management

VLAN Configuration 

# VLAN assignments for MS-A2 hosts:
VLAN 10: Management (2.5G interface)
VLAN 20: vMotion (10G interface)
VLAN 30: Storage (10G interface)
VLAN 100: VM Workload (10G interface)
VLAN 110: TKG Frontend (10G interface)

Storage Configuration

Local Storage Setup 

# MS-A2 storage configuration:
Slot 1: 1TB NVMe (included) - ESXi boot + local datastore
Slot 2: Available for expansion
Slot 3: Available for vSAN or additional storage

# ESXi partitioning:
- ESXi boot: 64GB
- Scratch partition: 4GB
- Local datastore: ~900GB

Shared Storage Integration

  • NAS connectivity: 10G connection to Synology
  • iSCSI configuration: Maintain existing LUN structure
  • NFS shares: Leverage improved 10G performance

vSphere Configuration

ESXi Installation and Setup

ESXi 8.0 Installation 

# Boot from USB installer
# Custom installation for MS-A2 hardware
# Configure management network on 2.5G interface
# Enable SSH and DCUI access

Host Configuration 

# Configure hostname and DNS
esxcli system hostname set --host="esxi-ms-a2-01"
esxcli network ip dns search add --domain="markalston.net"

# Configure NTP
esxcli system ntp set --server="pool.ntp.org"
esxcli system ntp set --enabled=true

# Configure advanced settings for AMD optimization
esxcli system settings advanced set -o /Power/CpuPolicy -s "Balanced"
esxcli system settings advanced set -o /Numa/CoreCapRatioThreshold -i 50

Cluster Integration

EVC Mode Configuration 

# Determine lowest common denominator between Intel and AMD
# Configure EVC mode in vCenter
# Test VM compatibility between platforms

# AMD-specific EVC considerations:
- AMD "Zen 2" or "Zen 3" family features
- Intel/AMD feature compatibility
- Workload-specific testing required

Network Configuration 

# Create distributed virtual switches
New-VDSwitch -Name "dvSwitch-10G" -Location (Get-Datacenter)

# Configure port groups for VLANs
New-VDPortgroup -VDSwitch "dvSwitch-10G" -Name "Management" -VlanId 10
New-VDPortgroup -VDSwitch "dvSwitch-10G" -Name "vMotion" -VlanId 20
New-VDPortgroup -VDSwitch "dvSwitch-10G" -Name "Storage" -VlanId 30
New-VDPortgroup -VDSwitch "dvSwitch-10G" -Name "Workload" -VlanId 100

# Add MS-A2 hosts to distributed switch
$VMHost = Get-VMHost "esxi-ms-a2-01"
Add-VDSwitchVMHost -VDSwitch "dvSwitch-10G" -VMHost $VMHost

Workload Migration

Migration Planning

Workload Prioritization

  1. Non-critical VMs: Test migration first
  2. Development environments: Early migration candidates
  3. Production services: Migrate during maintenance windows
  4. Storage-intensive workloads: Benefit most from 10G

Migration Methods 

# vMotion (preferred for running VMs)
Move-VM -VM "test-vm" -Destination "esxi-ms-a2-01"

# Cold migration (for powered-off VMs)
Move-VM -VM "test-vm" -Destination "esxi-ms-a2-01" -RunAsync

# Storage vMotion (change datastore)
Move-VM -VM "test-vm" -Datastore "ms-a2-local-01"

Performance Validation

Pre-Migration Baseline 

# CPU performance testing
# Memory bandwidth testing
# Storage IOPS measurement
# Network throughput validation

Post-Migration Validation 

# Application performance comparison
# Resource utilization monitoring
# User experience validation
# Automated testing where possible

Rollback Procedures

  • VM snapshots before migration
  • vMotion back to original hosts if issues
  • Documented rollback steps for each workload

Network Configuration

10G Network Optimization

ESXi Network Tuning 

# Enable jumbo frames for storage traffic
esxcli network vswitch standard set -v vSwitch0 -m 9000

# Optimize network buffer sizes
esxcli system settings advanced set -o /Net/TcpipHeapMax -i 1576960
esxcli system settings advanced set -o /Net/TcpipHeapSize -i 32

# Configure SR-IOV if supported
esxcli system module parameters set -m igb -p "max_vfs=8"

VLAN Configuration Validation 

# Test connectivity between VLANs
vmkping -I vmk1 192.168.20.10  # vMotion network
vmkping -I vmk2 192.168.30.10  # Storage network

# Validate routing between networks
esxcli network ip route ipv4 list

NSX-T Preparation

Network Requirements for NSX-T

  • Management network: Existing VLAN 10
  • Overlay transport: VLAN 20 (vMotion network)
  • Edge uplink: VLAN 100 (workload network)
  • TEP pool: Dedicated IP range for tunnel endpoints

Edge VM Placement Planning

  • Edge VMs: Deploy on MS-A2 hosts for performance
  • Resource allocation: 4 vCPU, 8GB RAM minimum per edge
  • Placement constraints: Anti-affinity rules for HA

Performance Validation

Benchmark Testing

CPU Performance 

# Single-threaded performance
# Multi-threaded scaling
# Virtual machine density testing
# Comparison with Intel baseline

Memory Performance 

# Memory bandwidth testing
# Memory latency measurement
# NUMA topology validation
# Memory overcommit testing

Storage Performance 

# Local NVMe testing
# NAS connectivity (10G)
# IOPS measurement
# Latency characterization

Network Performance 

# 10G throughput testing
iperf3 -s                          # Server mode
iperf3 -c <target> -t 60 -P 4     # Client test

# Expected results:
# 10G DAC: 9.4+ Gbps
# Storage: 8+ Gbps sustained
# VM-to-VM: 9+ Gbps local

Production Validation

Application Testing

  • Web applications: Response time validation
  • Database workloads: Query performance comparison
  • Development tools: Build time improvements
  • Container workloads: Startup and scaling performance

Monitoring and Alerting 

# vCenter performance monitoring
# ESXi host resource utilization
# Network bandwidth utilization
# Storage performance metrics

Troubleshooting

Common Migration Issues

EVC Compatibility Problems 

# Symptoms: vMotion failures between Intel/AMD
# Solution: Lower EVC mode or separate clusters
# Validation: Test with representative workloads

Network Connectivity Issues 

# Symptoms: Loss of network after migration
# Check: VLAN configuration, uplink status
# Validation: Layer 2/3 connectivity testing

Performance Degradation 

# Symptoms: Slower performance on AMD
# Check: CPU scheduling, NUMA affinity
# Tuning: VM hardware version, resource allocation

Performance Optimization

AMD-Specific Tuning 

# CPU scheduler optimization
esxcli system settings advanced set -o /Cpu/PreemptionTimer -i 100000

# Memory optimization
esxcli system settings advanced set -o /Mem/ShareScanTime -i 60

# Power management
esxcli system settings advanced set -o /Power/CpuPolicy -s "High Performance"

Migration Timeline

Detailed Schedule

Week 1: Infrastructure Preparation

  • Day 1-2: Rack installation and cable management
  • Day 3-4: 10G network deployment and testing
  • Day 5-7: First MS-A2 setup and ESXi installation

Week 2: Cluster Integration

  • Day 1-2: Add MS-A2 to vSphere cluster
  • Day 3-4: Configure networking and storage
  • Day 5-7: Validation testing and performance baseline

Week 3-4: Initial Migration

  • Week 3: Migrate development and test VMs
  • Week 4: Migrate non-critical production workloads

Week 5-6: Cluster Expansion

  • Week 5: Deploy second MS-A2, continue migration
  • Week 6: Deploy third MS-A2, accelerate migration

Week 7-8: Completion

  • Week 7: Migrate remaining workloads
  • Week 8: Decommission NUCs, optimize configuration

Risk Mitigation

  • Parallel clusters: Maintain NUCs until migration complete
  • Rollback capability: Keep original environment functional
  • Extended timeline: Allow flexibility for issues
  • Support resources: Engage vendor support if needed

References

Internal Documentation

Vendor Resources

Performance Testing

Last Updated: 2025-01-11 Maintained by: Mark Alston

This project is for educational and home lab purposes.