Academy/vSphere Foundation 9.0 Administrator (2V0-16.25)/Lab: vSAN Witness Configuration (2-Node)
This lab targets VCF 9.0

Lab: vSAN Witness Configuration (2-Node)

VCF 9.0Intermediatevcp-foundation⏱ 75 min

VVF 9.0 administrator lab (2V0-16.25). Built for a Holodeck VVF instance.

Objectives

  • Deploy a vSAN witness appliance and build a 2-node cluster (Obj 2.3)
  • Set the preferred fault domain and explain partition behaviour
  • Explain protection after a host failure, including nested fault domains

Prerequisites

Holodeck VVF 9.0 instance deployed and healthy

Lab Environment

Holodeck vSphere Foundation 9.0 instance (New-HoloDeckInstance -VVF; -vSANMode OSA by default, ESA where the physical layer supports it): vCenter managing one vSAN cluster of nested ESX hosts, plus VCF Operations. VVF has no SDDC Manager and no NSX. Two data sites and a third witness site are simulated with nested hosts; witness RTT targets apply to real WAN links.

Credentials

SystemUsernamePassword
vCenteradministrator@vsphere.localHolodeck documented default (Post Deployment table)
Nested ESX hostsrootHolodeck documented default (Post Deployment table)

Tasks

Task 1 vSAN Witness Configuration (2-Node)

2-node vSAN with witness is designed for edge/ROBO sites where cost constrains host count. Understanding witness placement, quorum behavior, and failure scenarios is critical for edge architecture.
Step 1

Build 2x ESX hosts (data sites A and B) + 1x vSAN witness appliance (an ESX VM deployed from the OVA) on a third site or host outside the cluster (VCF 9.0.1 docs p.1699); an NFS share cannot act as a witness

Step 2

Add the witness appliance to vCenter as a host outside the 2-node cluster, then Cluster → Configure → vSAN → Services → Configure → Two-node vSAN cluster → select the witness host

Step 3

Witness deployed from the vSAN witness appliance OVA at a profile size: Tiny (≤750 components, ≤10 VMs — OSA only; vSAN ESA does not support Tiny), Medium (≤21,833 components, ≤500 VMs — the ESA minimum), Large or Extra Large (VCF 9.0.1 docs pp.1699-1700). The profile fixes the appliance's vCPU/RAM/disk.

Step 4

Test quorum loss: Power down one ESXi host → Remaining host + witness maintain quorum

Step 5

Verify: vSAN health dashboard → Witness health = green

Validation Gate

Check: After configuring 2-node vSAN with witness: verify witness connectivity to both hosts, confirm preferred fault domain assignment, and simulate a host failure to validate HA behavior

Expected: Witness shows connected to both data hosts. Preferred host is set. During simulated failure of non-preferred host: VMs restart on preferred host, vSAN objects degrade to single copy but remain accessible.

Common Errors

Placing the witness host on the same failure domain as one of the data hosts
Fix: The witness must be in a separate failure domain (different rack, site, or at minimum a different ESXi host). If the witness is on the same rack as Host-1 and that rack loses power, vSAN loses quorum (1 surviving host out of 3 vote components) and goes offline.
Sizing the witness host too small
Fix: The witness holds witness components only (no VM data), but it is sized by component and VM count through the OVA profile: Tiny for ≤750 components / ≤10 VMs (OSA only), Medium for ≤21,833 components / ≤500 VMs (minimum for vSAN ESA), Large for ≤45,000, Extra Large for ≤64,000. It needs network connectivity to both data hosts (two-node witness link: less than 500 ms RTT, VCF 9.0.1 docs p.1608). Pick the profile from the documented table rather than hand-sizing CPU and RAM.
Not configuring preferred fault domain
Fix: In 2-node vSAN, one host is designated as 'preferred' fault domain. During a network partition (hosts can't see each other but both see witness), the preferred host keeps serving I/O. Without this setting, behavior during partition is unpredictable. Configure in the vSphere Client: cluster → Configure → vSAN → Fault Domains (VVF has no SDDC Manager).
Assuming 2-node vSAN has no protection left after a host failure
Fix: Across the two hosts a 2-node cluster mirrors data (FTT=1 between hosts). With nested fault domains (at least three disk groups per host and a matching policy) data is also mirrored inside each host, so the surviving host can still tolerate a drive failure (VCF 9.0.1 docs p.1670). RAID-5 across hosts needs 3 hosts on ESA (2+1) or 4 on OSA (3+1), so it is not available in a 2-node cluster (p.1713). Without nested fault domains, a host failure leaves a single copy — a key risk to communicate to the customer.

Final Validation

Gate: every check below passes with the evidence recorded in the lab log.

✓ Witness → Witness host connected to both data hosts

✓ Preferred site → Preferred fault domain set in Configure → vSAN → Fault Domains

✓ Failure test → VMs restart on the surviving host and objects stay accessible

Cleanup / Restore

• Remove the 2-node test cluster and the witness appliance if they are not needed later

Design Reflection (VCDX)

2-node vSAN is relevant for edge/ROBO architectures in VCDX defense. Panelists test whether you understand the limitations (mirroring between hosts, nested fault domains for in-host protection, single host failure tolerance) and when to recommend 2-node vs 4-node clusters.

Requirements

  • Deploy 2-node vSAN with witness for edge/ROBO site
  • Configure preferred fault domain for predictable partition behavior
  • Understand single-host-failure tolerance limitations

Constraints

  • 2-node vSAN mirrors between hosts (FTT=1); nested fault domains add protection inside each host
  • Witness must be in separate failure domain
  • Network partition behavior depends on preferred fault domain config

Assumptions

  • WAN connectivity to witness site is reliable
  • Edge workloads can tolerate single-copy risk during host failure, or nested fault domains are configured

Risks

  • Both hosts failing simultaneously — total data loss
  • Witness network failure causing quorum loss and vSAN offline

⚠ Known Pitfalls

Deploying 2-node vSAN without communicating the host-failure risk — without nested fault domains a host failure leaves a single copy of the data.
Assuming 2-node vSAN can run RAID-5 across hosts — it cannot; erasure coding needs 3 hosts on ESA (2+1) or 4 on OSA (3+1).

References

Was this page useful?
Type to search. ↑ ↓ to move, Enter to open, Esc to close.