vmotion
EVC Mode & CPU Compatibility FAQ | Lazy Admin Blog
You’ve just unboxed a shiny new host with the latest Intel or AMD processor, but your current cluster is running hardware from three years ago. You try to vMotion a VM, and vSphere gives you the dreaded “CPU Incompatibility” error.
Enter Enhanced vMotion Compatibility (EVC). Here’s everything you need to know to get your mixed-hardware cluster working without the headache.
What exactly is EVC?
Think of EVC as a “lowest common denominator” filter for your CPUs. It masks the advanced features of newer processors so that every host in the cluster appears to have the exact same instruction set. This allows VMs to live-migrate between old and new hardware because the “view” of the CPU never changes.
Quick FAQ
Q: Can I mix Intel and AMD in the same EVC cluster? A: No. EVC only works within a single vendor family. You can mix different generations of Intel, or different generations of AMD, but you cannot vMotion between the two brands.
Q: Will EVC slow down my new servers? A: Technically, yes—but rarely in a way you’ll notice. It hides new instructions (like specialized encryption or AI math sets), but the raw clock speed and core count of your new CPUs are still fully utilized. Most general-purpose VMs don’t use the high-end instructions being masked.
Q: Do I need to power off VMs to enable EVC? A: It depends:
- Enabling on an empty cluster: No downtime.
- Enabling on a cluster where VMs are already running on the oldest host: Usually no downtime.
- Enabling on a cluster where VMs are running on newer hosts: You must power off those VMs so they can “re-boot” with the masked CPU instructions.
Q: What is “Per-VM EVC”? A: Introduced in vSphere 6.7, this allows you to set the EVC mode on the VM itself rather than the whole cluster. This is a lifesaver for migrating VMs across different vCenters or into the Cloud (like AWS/Azure).
How to Find Your Correct EVC Mode
Don’t guess. Use the official tool:
- Go to the VMware Compatibility Guide (CPU/EVC Matrix).
- Select your ESXi version.
- Select the CPU models of your oldest and newest hosts.
- The tool will tell you the highest supported “Baseline” you can use.
Step-by-Step: Enabling EVC on an Existing Cluster
- Select your Cluster in vCenter.
- Go to Configure > VMware EVC.
- Click Edit.
- Select Enable EVC for Intel/AMD hosts.
- Choose the Baseline that matches your oldest host.
- Validation: vCenter will check if any running VMs are currently using features above that baseline. If they are, you’ll need to shut them down before you can save the settings.
Summary Table: EVC Baselines
| If your oldest host is… | Use this EVC Mode |
| Intel Ice Lake | Intel “Ice Lake” Generation |
| Intel Cascade Lake | Intel “Cascade Lake” Generation |
| AMD EPYC Rome | AMD EPYC “Rome” Generation |
vSphere Ports & Connections: The Infrastructure Roadmap | Lazy Admin Blog
In a locked-down enterprise environment, the “Any-to-Any” firewall rule is a myth. To manage ESXi effectively, you need to poke specific holes in your hardware and software firewalls.
The Core Management Ports
These are the “must-haves” for basic connectivity between vCenter, the vSphere Client, and the Host.
| Port | Protocol | Source | Destination | Purpose |
| 443 | TCP | Management Workstation | vCenter / ESXi | vSphere Client / SDK: The primary port for the Web Client and API access. |
| 902 | TCP/UDP | vCenter Server | ESXi Host | vCenter Agent (vpxa): vCenter uses this to send data to the host and receive heartbeats. |
| 902 | TCP | Management Workstation | ESXi Host | VM Console: Required to open the “Remote Console” (MKS) to a virtual machine. |
| 80 | TCP | vCenter / Workstation | ESXi Host | HTTP: Used for redirecting to 443 and for some legacy file downloads. |
Advanced Feature Ports
If you are using specific vSphere features like vMotion, HA, or specialized storage, you need these additional ports open:
1. vMotion (Live Migration)
- 8000 (TCP): Required for vMotion traffic.
- 2049 (TCP/UDP): If using NFS storage for the virtual disks.
2. vSphere High Availability (HA)
- 8182 (TCP/UDP): Used by the Fault Domain Manager (FDM) agent for inter-host communication and election of the master host.
3. Provisioning & Deployment
- 69 (UDP): TFTP, used for PXE booting ESXi for Auto Deploy.
- 4012 (TCP): Used by the Auto Deploy service.
4. Troubleshooting & Monitoring
- 22 (TCP): SSH access to the ESXi Shell.
- 161 / 162 (UDP): SNMP polling and traps for hardware monitoring.
Troubleshooting “Host Disconnected”
If your host shows as “Not Responding” in vCenter, check these three things in order:
- Ping: Can the vCenter server ping the ESXi management IP?
- Port 902: From the vCenter server, try to telnet to the host on port 902 (
telnet <host-ip> 902). If it fails, the heartbeat can’t get through. - DNS: VMware is extremely sensitive to DNS. Ensure forward and reverse lookups work for both the vCenter and the Host.
Lazy Admin Tip 💡
Don’t memorize every port! Use the VMware Ports and Protocols Tool (the official online matrix). It allows you to select your source and destination products and generates a custom firewall rule list for you.
A high resolution pdf can be downloaded here Connections and Ports in ESX and ESXi
#VMware #vSphere #Networking #SysAdmin #Firewall #DataCenter #ESXi #ITOps #LazyAdmin #Connectivity
Top VMware ESXi & vSphere Interview Questions
Preparing for a Virtualization role? This guide covers everything from legacy ESX vs. ESXi differences to advanced HA and DRS logic.
🔄 The Evolution: ESX vs. ESXi
- Service Console: ESX had a Service Console (based on RHEL); ESXi is “thin” and has no console, leading to a smaller footprint and faster boots.
- Hardware: ESXi can be purchased as an embedded hypervisor directly on hardware.
- Health Checks: ESXi features built-in server health status monitoring.
🛡️ High Availability (HA) 5.0 Deep Dive
In vSphere 5.0, the HA architecture moved from a Primary/Secondary model to a Master/Slave concept using the FDM (Fault Domain Manager) agent.
| Role | Responsibilities |
| Master | Monitors host/VM availability, manages restarts, communicates with vCenter. |
| Slave | Monitors local VMs, sends status to Master, participates in elections if Master fails. |
Heartbeat Mechanisms:
- Network Heartbeat: Sent between Master and Slaves every second.
- Datastore Heartbeat: Used if the network heartbeat is lost to determine if a host is isolated or has actually failed.
🚀 vMotion & DRS (Distributed Resource Scheduler)
vMotion Prerequisites:
- Shared storage (required prior to 5.1).
- GigaBit Ethernet dedicated vMotion network (VMkernel port).
- Processor compatibility (EVC – Enhanced vMotion Compatibility).
- No active CD-ROM/ISO mounts or CPU affinity.
DRS Automation Levels:
- Manual: vCenter suggests migrations; admin executes.
- Partially Automated: vCenter handles initial VM placement; suggestions for migrations.
- Fully Automated: vCenter moves VMs automatically based on load.
💾 Storage & Networking Quick Hits
- vSAN: Aggregates local storage from ESXi hosts into a single shared datastore.
- iSCSI Port Binding: Used when multiple VMkernel ports are in the same subnet to allow multiple paths to an array.
- Path Selection Policies (PSP): Fixed, MRU (Most Recently Used), and Round Robin.
- Key Command Line Tools:
esxtop: Live performance data.vmkfstools: Virtual disk management.vmware-cmd: VM management and info.
📊 Hardware Version Comparison
| Feature | HW Version 4 (ESX 3.x) | HW Version 7 (vSphere 4.x) | HW Version 8 (vSphere 5.0) |
| Max vRAM | 64 GB | 256 GB | 1 TB |
| Max vCPU | 4 | 8 | 32 |
| USB Support | No | Yes | Yes (incl. 3.0) |
| NICs per VM | 4 | 10 | 10 |
🚀 Key Differences in Modern vSphere (7.0 & 8.0)
1. The Architecture Shift: Project Monterey & DPUs
Modern vSphere now supports DPUs (Data Processing Units). Instead of the CPU handling networking and security, these tasks are offloaded to the SmartNIC.
2. Tanzu (Kubernetes Integration)
The biggest change in vSphere 7/8 is that Kubernetes is built directly into the hypervisor. You no longer just manage VMs; you manage “Namespaces” and containers natively on ESXi.
3. vMotion Enhancements (vSphere 7+)
In version 5.0, vMotion would “stun” a VM briefly. Modern vMotion uses a “Claim” mechanism that makes migrating massive VMs (Monster VMs) almost instantaneous with zero performance impact.
4. Scalability Comparison (vSphere 5.0 vs. 8.0)
| Feature | vSphere 5.0 (Legacy) | vSphere 8.0 (Modern) |
| vCPUs per VM | 32 | 768 |
| RAM per VM | 1 TB | 24 TB |
| Hosts per Cluster | 32 | 96 |
| VMs per Cluster | 3,000 | 10,000 |
🆕 2026 Interview Questions: Modern Edition
Q: What is the “vSphere Distributed Services Engine”?
A: It is the feature that allows vSphere to use DPUs (SmartNICs) to offload infrastructure services like NSX and vSAN, freeing up the host’s CPU for application workloads.
Q: What is a “Lifecycle Manager” (vLCM)?
A: In vSphere 7+, vLCM replaced the old Update Manager (VUM). It uses a declarative model (Desired State) where you define an image for a cluster, and the hosts automatically maintain that version/driver level.
Q: What is “vSAN Express Storage Architecture” (ESA)?
A: Introduced in vSphere 8, ESA is a new way of processing data optimized for high-performance NVMe drives, removing the old “Disk Group” (Cache/Capacity) requirement.
Q: How does vSphere 8 handle AI/ML workloads?
A: Through vGPU and Device Groups, allowing VMs to span multiple physical GPUs and utilizing High-Bandwidth Memory (HBM) for massive AI model training.
#VMware #vSphere #ESXi #Virtualization #SysAdmin #TechInterview #vMotion #CloudComputing #LazyAdmin #DataCenter