Academy/VCP-VCF 9.0 Administrator (2V0-17.25)/Workload Domain Provisioning — Traditional vs Stretched
This lab targets VCF 9.0

Workload Domain Provisioning — Traditional vs Stretched

VCF 9.0Advancedadminarchitectvcdx⏱ 180 min

Workload domain provisioning in VCF 9.0.x runs from VCF Operations (Inventory > Detailed View > VCF Instance > Add Workload Domain > Create New); hosts are commissioned first (vSphere Client > Global Inventory Lists > Hosts > Unassigned Hosts > Commission Hosts, or SDDC Manager). Every workload domain gets its own vCenter deployed in the management domain; NSX Manager can be new or shared. Clusters and hosts are added afterwards in the vSphere Client. Comparison with VCF 5.2 in Extension 3. Focus is on traditional single-site WLD vs stretched cluster topology with witness host and fault domains. Sources: VCF 9.0.1 docs pp.496 (Workload Domain Model), 1498-1507 (Commission ESX Hosts, Create a New Workload Domain Using VCF Operations), 1530 (Add an ESX Host to an SDDC Cluster).

Objectives

  • Plan a workload domain deployment: network pools, host commissioning prerequisites, licensing, and resource capacity analysis
  • Provision a traditional (single-site) workload domain from VCF Operations (Add Workload Domain > Create New) with its own vCenter, a new or shared NSX Manager, vSAN principal storage and a cluster image, then deploy the NSX Edge cluster and Tier-0 gateway as the post-deployment step
  • Understand stretched cluster architecture: witness host deployment, fault domains, quorum model, and cross-site network requirements
  • Provision a stretched workload domain with multi-site replication and validate witness host quorum
  • Execute workload domain lifecycle operations in the vSphere Client: commission hosts, add hosts to the SDDC cluster, expand the cluster, and validate vSAN placement
  • Compare traditional and stretched domain trade-offs (RTO, complexity, failure domain, cost)

Prerequisites

VCF 9.0.x management domain deployed and operational with VCF Operations (fleet management) and SDDC Manager accessible. 8+ ESX 9.0.x hosts prepared for commissioning in two physical sites (4 per site for the stretched extension, or 4+ in a single site for the traditional lab). A vSphere Lifecycle Manager cluster image imported into the VCF Instance (VCF Operations > Fleet Management > Lifecycle > Image Management). A network pool with free addresses and a static IP pool or DHCP for the host overlay (TEP) VLAN. DNS entries for the new vCenter and NSX Manager FQDNs. Lab account must hold the VCF Operations Administrator role (and SDDC Manager Administrator).

Prior labs: holodeck-02

Required skills:

  • VCF management domain architecture (SDDC Manager, vCenter, NSX Manager)
  • ESX host commissioning and network configuration (VMkernel, vMotion, vSAN)
  • NSX fundamentals: Tier-0 and Tier-1 gateways, segments, overlay networking
  • vSAN design: fault domains, stretched cluster quorum, witness host role
  • Network topology design: IP planning, VLAN allocation, BGP peering with NSX edges
  • VCDX-level understanding of failure domains, availability zones, and disaster recovery models

Lab Environment

VCF 9.0.x management domain with 8-12 ESX 9.0.x hosts ready for commissioning. For traditional WLD lab: 4 hosts in single site. For stretched WLD lab (extension): 4 hosts in Site A, 4 hosts in Site B (both sites need the same number of hosts), 1 witness host at a third site. All hosts have: iLO/IPMI for out-of-band management, 10Gbps+ network connectivity, vSAN-capable storage (local NVMe), and configured for vMotion/vSAN trunk port groups.

Traditional WLD
Mermaid source
graph TB
  SDDC[Management Domain]
  SDDC -->|dedicated WLD vCenter deployed in mgmt domain| WLD-A[Workload Domain A: 4 hosts, vSAN]
  SDDC -->|Network| T0A[NSX T0 Gateway]
  T0A -->|Uplink| PAS[Physical Access Switch]
Stretched WLD
Mermaid source
graph TB
  SDDC[Management Domain]
  SDDC -->|vCenter Stretched| WLD-S[Stretched WLD: Site A 4h + Site B 4h + Witness]
  SDDC -->|Network L3| Site-A[Site A Network]
  SDDC -->|Network L3| Site-B[Site B Network]
  Site-A -->|Witness traffic| Witness[Witness Host - third site]
  Site-B -->|Witness traffic| Witness
  Site-A -->|vSAN Replication| Site-B

IP Addressing

NetworkPurposeVLAN
ManagementManagement domain (SDDC Manager, vCenter, NSX Manager)1644
Workload Host ManagementWLD ESX host VMkernel (one VLAN per host group)1650-1659
vMotionvMotion VMkernel (for live migrations within WLD)1660-1669
vSANvSAN VMkernel traffic (storage traffic between the hosts of the cluster)1670-1679
NSX Host TEPHost Tunnel End Point (TEP) for NSX overlay1680-1689
NSX Edge TEPEdge node TEP for edge-based services1690-1699
Witness Isolation (Stretched only)Witness host at the third site (lab simulation); the witness vmk0 management network, routed at L3 from the data-site host management networks for vSAN witness traffic (VCF 9.0.1 docs p.1074, p.1077); carries witness components (metadata) only, no VM data or I/O1710

Credentials

SystemUsernamePassword
SDDC Manageradministrator@vsphere.localSet during VCF bring-up
ESX Hosts (to be commissioned)rootSet during host onboarding
vCenter (new WLD vCenter if applicable)administrator@vsphere.localSet during WLD provisioning
NSX ManageradminSet during management domain bring-up

Tasks

Task 1 Plan and prepare for workload domain provisioning

manageability

Pre-deployment planning is where VCDX candidates demonstrate rigor. You must validate that the infrastructure is ready before initiating a 2+ hour provisioning operation. This mirrors production change management: plan, validate, execute, document.

Step 1

Commission the hosts first — commissioning is a prerequisite, not a provisioning milestone. In the vSphere Client (management vCenter) click Global Inventory Lists > Hosts > Unassigned Hosts and review the hosts already in the free pool. Commission the four hosts for this lab if they are not there yet: click Commission Hosts, confirm the checklist (Select All > Proceed), add the hosts one at a time or by JSON (FQDN, root credentials, storage type VSAN or VSAN_ESA, network pool), confirm fingerprints, Validate All, Next, Commission. The platform activates SSH on each host only to read the host key and then deactivates it (p.1498). Hosts commissioned for vSAN ESA can only be used for vSAN ESA clusters.

Unassigned Hosts shows esxi-05 to esxi-08 commissioned with the target principal storage type (for example VSAN_ESA) and Active
If hosts show 'Disconnected' or 'Commissioning in Progress', wait for commissioning to complete (or manually commission them: Administration > Inventory > Host Actions > Commission).
SDDC Manager UI (https://<sddc-manager-fqdn>) is the documented alternative for commissioning when the vSphere Client path is not available (p.1498).
Step 2

Verify network infrastructure readiness for the WLD. For each proposed ESX host in the WLD, validate: (a) Port group for host management (VMkernel) exists and is on dedicated VLAN, (b) Port group for vMotion exists, (c) Port group for vSAN exists, (d) Port group for NSX Host TEP exists, (e) All port groups are VLAN-enabled and do NOT have vLAN trunking misconfigured. Document the VLAN and IP range for each function.

Network readiness checklist completed. Example: 'Host Management VLAN 1650 (10.2.0.0/24), vMotion VLAN 1660 (10.2.1.0/24), vSAN VLAN 1670 (10.2.2.0/24), NSX TEP VLAN 1680 (10.2.3.0/24). All port groups configured on physical switches. VLAN trunking verified.'
VLAN misconfiguration is the #1 reason WLD provisioning fails at the host commissioning step. Verify this before starting.
Step 3

Create a workload domain specification document. Define: (a) WLD Name (3-20 characters, e.g., 'Workload-Domain-01'; resource names derive from it), (b) Cluster Name (e.g., 'WLD-Cluster-01') and, if vSphere Supervisor stays enabled, a vSphere Zone name, (c) Host list (esxi-05, esxi-06, esxi-07, esxi-08) and their commissioned storage type, (d) vSAN configuration (OSA with Failures to Tolerate, or ESA HCI/Storage cluster with auto-policy), (e) network pool and the TEP IP pool/DHCP for the host overlay VLAN, (f) the new workload domain vCenter: FQDN (DNS-resolvable) and whether it joins the management SSO domain or an isolated new SSO domain — every workload domain has a dedicated vCenter deployed in the management domain (p.496), (g) NSX Manager: new (Standard or High-Availability, appliance size, three node FQDNs + cluster FQDN) or shared with an existing workload domain, (h) network connectivity: Centralized (Edge cluster and Tier-0 deployed afterwards; required for Supervisor) or Distributed (VLAN, gateway CIDR, external and transit IP blocks), (i) distributed switch profile (Default, Storage Traffic Separation, NSX Traffic Separation, both, or Custom), (j) cluster image to assign.

Workload domain specification document (1-2 pages) with all wizard parameters defined. Example snippet: 'WLD Name: Prod-Zone-A, Host Count: 4 (VSAN_ESA), vCenter: vc-prodzonea.vcf.sddc.lab in the management SSO domain, NSX: new, High-Availability, Medium; Connectivity: Centralized; Image: esx-9.0.x-prodzone.'
Step 4

Validate licensing in VCF Operations. In VCF 9.0 licenses are managed in VCF Operations and assigned to vCenter instances; hosts and components connected to a licensed vCenter are licensed automatically, and there are only two licenses: VMware Cloud Foundation (cores) and VMware vSAN (TiBs) (VCF 9.0.1 docs p.77). Confirm the VMware Cloud Foundation (cores) license has enough unassigned cores for the new hosts (per-core licensing with a minimum of 16 cores per CPU, pp.80 and 1462) and that the vSAN (TiB) entitlement covers the planned vSAN capacity. The new workload domain's vCenter is licensed by assigning the license to it. Document any licence constraint (e.g., 'unassigned cores cover the 4 hosts of WLD-01 only; WLD-02 needs additional cores').

Licensing status (example): 'VMware Cloud Foundation (cores): 256 cores, 128 assigned to the management domain vCenter, 128 unassigned. WLD-01 needs 128 cores (4 hosts x 2 CPUs x 16 cores) — OK. vSAN (TiBs) entitlement covers the planned capacity.'
Step 5

Validate infrastructure capacity. Calculate resource overhead: (a) vCenter Server VM (Small: 4 vCPU, 21 GB RAM) per WLD or shared with management, (b) NSX Edge cluster (typically 2-4 edge nodes × 4 vCPU, 8 GB RAM each), (c) vSAN witness appliance (if stretched, 2 vCPU, 8 GB RAM). Total the resources and verify against available physical capacity. In SDDC Manager, Inventory > System Capacity, check total host resources available.

Capacity calculation: 'WLD vCenter: 4C/16G, NSX Edges (optional): 8C/32G, Total workload capacity on 4 hosts: 128 vCPU, 384 GB RAM (gross). After SDDC Manager, vSAN, and overhead: ~90 vCPU, 250 GB RAM available for workload VMs. Document this for sizing discussions.'
Step 6

Create a provisioning checklist: (1) All 4 ESX hosts commissioned with the right storage type and Active in Unassigned Hosts, (2) Network VLANs verified on the physical switches; TEP IP pool or DHCP available on the host overlay VLAN; network pool has free addresses, (3) Licensing capacity confirmed, (4) vSAN layout planned (OSA FTT or ESA cluster type), (5) NSX decision made (new HA cluster or shared) and Edge cluster/Tier-0 plan for the post-deployment step (uplink VLAN, BGP peers), (6) New vCenter FQDN resolvable in DNS and SSO domain decision (joined or isolated), (7) Cluster image imported in VCF Operations Image Management, (8) Risk mitigation: snapshot the management domain before provisioning. Walk through each item. Check off as complete.

Provisioning checklist with 8 items, all verified and signed off. Example final item: 'Pre-provisioning snapshot taken: md-before-wld-01-provisioning. Rollback path clear.'

Validation Gate

Check: Planning documents complete: WLD specification, networking validation, licensing check, capacity calculation, provisioning checklist. All items signed off.

Expected: Pre-provisioning readiness confirmed. No blockers identified. Ready to proceed to Task 2 (traditional WLD provisioning).

Common Errors

VLAN mismatch during commissioning: ESX-05 cannot reach vCenter on the planned management VLAN
Cause: Port group on the physical switch is not trunked for the planned VLAN, or VLAN ID is misconfigured
Fix: Before WLD provisioning, verify each VLAN is operational: ping from physical switch to the vCenter management IP on each VLAN. If VLAN is missing, add port group to physical switch and configure VLAN trunking.
Licence shortfall found during planning: not enough unassigned cores (or vSAN TiB) for the new workload domain
Cause: The VMware Cloud Foundation (cores) or VMware vSAN (TiBs) license is fully assigned or the subscription has expired
Fix: In VCF Operations licensing, check the license's assigned and unassigned capacity and its expiry date. If it has expired or is exhausted, obtain additional subscription capacity from Broadcom (VCF Business Services console), or scale back the WLD host count or defer WLD-02 provisioning. The new WLD's vCenter is licensed by assigning the license to it (VCF 9.0.1 docs p.77).
Cannot commission hosts: 'Host esxi-05 network connectivity failed'
Cause: ESX host does not have IP connectivity to SDDC Manager on the management network
Fix: SSH to the physical ESX host. Verify: (1) Management VMkernel has an IP in the expected range (e.g., 10.2.0.5), (2) Default gateway is set and reachable, (3) Ping to SDDC Manager IP succeeds (ping 10.0.0.4). If not, reconfigure VMkernel networking on the host.

Task 2 Provision a traditional (single-site) workload domain

availability

Execute the workload domain creation from VCF Operations. The wizard deploys a dedicated vCenter into the management domain, creates or joins an NSX Manager, builds the first SDDC cluster with the cluster image and principal storage, and (if left enabled) activates vSphere Supervisor. Understand the wizard order and what happens afterwards (Edge cluster and Tier-0 for Centralized connectivity, clusters and hosts in the vSphere Client).

Step 1

In VCF Operations select Inventory > Detailed View, expand VCF Instances and browse to your instance. Click Add Workload Domain > Create New. Review the prerequisites, click Select All and Proceed. Enter the General Information: Workload Domain Name (3-20 characters), Enable vSphere Supervisor (on by default — leave on for this lab and plan the Edge cluster afterwards, or switch off for a plain vSphere cluster), SSO Domain Name (a new name creates an isolated SSO domain; otherwise the domain joins the management SSO), Password Creation (auto-generate, or set the SSO administrator password). Click Next.

General Information page accepted; wizard moves to vCenter
VCF 9.0.1 docs pp.1500-1501. Auto-generated passwords are retrieved later through the SDDC Manager API.
Step 2

Enter the vCenter details: vCenter FQDN (must resolve in DNS) and vCenter root password (unless auto-generated). Every workload domain gets its own vCenter, deployed as an appliance in the management domain; there is no option to reuse the management vCenter (p.496 'Has a dedicated vCenter'; p.1502). Click Next.

vCenter page accepted
Do not plan around a shared vCenter: adding a cluster to the management vCenter creates a management-domain cluster, not a workload domain.
Step 3

Enter the Cluster details: Cluster Name (e.g., 'WLD-Cluster-01') and, if Supervisor is enabled, the vSphere Zone Name. Click Next. Select the cluster image imported into the VCF Instance (hosts whose vendor add-on or components differ are remediated to the image during creation, p.1502). Click Next.

Cluster name and cluster image set
Step 4

Enter the NSX Manager details: create a new NSX Manager instance (Deployment Size Standard = single node, or High-Availability = three-node cluster, recommended; Appliance Size Medium/Large/Extra Large; three appliance FQDNs and the cluster FQDN; administrator and auditor passwords unless auto-generated) or join an existing NSX Manager instance from another workload domain. Choose the network connectivity: Centralized Connectivity (configure Edge cluster and Tier-0 after creation; required for vSphere Supervisor) or Distributed Connectivity (VLAN ID, Gateway CIDR, External IP Block, Private Transit Gateway IP Block — VPC-ready at once). For this lab choose a new High-Availability NSX Manager and Centralized Connectivity. Click Next.

NSX Manager and connectivity model set
Only NSX Manager may be shared between workload domains; vCenter never is (p.496).
Step 5

Enter the Storage details: vSAN (optionally Enable vSAN ESA), NFS, VMFS on FC or vVol (deprecated from 9.0). For vSAN OSA set Failures to Tolerate (determines the minimum host count) and Space efficiency; for vSAN ESA choose vSAN HCI or vSAN Storage cluster type (auto-policy management). Click Next. Select the ESX hosts — they must be commissioned with the same storage type as this cluster (p.1506). Click Next.

Storage and hosts selected: 4 commissioned hosts matching the storage type
Step 6

Enter the Distributed Switch details: Default (one vDS for all traffic, recommended), Storage Traffic Separation, NSX Traffic Separation, both, or Custom Switch Configuration (management, vMotion, vSAN and NSX traffic types configured once per cluster). If Supervisor is enabled, enter the vSphere Supervisor details (Supervisor name, Service CIDR, management network settings). Review the summary and click Finish. Follow the task in Tasks.

Workload domain creation task running: vCenter deployment, NSX Manager deployment, cluster creation with the image, Supervisor activation (if enabled)
Step 7

Post-deployment network step for Centralized Connectivity: deploy an NSX Edge cluster with a Tier-0 gateway for the new workload domain (see 'Configure Centralized Network Connectivity with Edge Clusters'); an active-standby Tier-0 is required to complete Supervisor activation unless you joined an existing NSX Manager that already has a compatible Edge cluster (p.1501). Record uplink VLAN, BGP neighbours and ASNs in your specification.

Edge cluster and Tier-0 gateway deployed for the workload domain; Supervisor activation (if enabled) completes

Validation Gate

Check: VCF Operations Inventory > Detailed View shows Workload-Domain-01 active with its own vCenter and NSX Manager. Log in to the new workload domain vCenter (its FQDN, not the management vCenter) and verify cluster 'WLD-Cluster-01' with 4 hosts connected, vSAN healthy, and the cluster compliant with the assigned image. NSX Manager (new or shared) shows 4 host transport nodes. If Centralized Connectivity: Edge cluster and Tier-0 deployed.

Expected: Traditional WLD fully provisioned: dedicated vCenter in the management domain, NSX Manager cluster, 4-host vSAN cluster on the cluster image, network connectivity model in place.

Common Errors

The wizard does not list the hosts you want to use
Cause: Hosts are not commissioned, or were commissioned with a different principal storage type than the cluster you are creating (for example vSAN OSA hosts for a vSAN ESA cluster)
Fix: Commission the hosts from Global Inventory Lists > Hosts > Unassigned Hosts with the matching storage type (p.1498-1499); hosts commissioned for ESA can only join ESA clusters.
Prerequisite or validation failure on vCenter or NSX FQDNs
Cause: The new vCenter or NSX Manager FQDNs/IPs do not resolve in DNS, or the network pool has no free addresses
Fix: Create forward and reverse DNS records for the vCenter FQDN and the NSX node and cluster FQDNs; add addresses to the network pool; rerun the wizard.
Workload domain created but Supervisor activation is pending
Cause: Centralized Connectivity requires an NSX Edge cluster with an active-standby Tier-0 gateway after creation
Fix: Deploy the Edge cluster and Tier-0 for the domain (Configure Centralized Network Connectivity with Edge Clusters), then Supervisor activation completes (p.1501).
Looking for a 'Create New Workload Domain' page in SDDC Manager
Cause: In VCF 9.0 workload domain creation runs from VCF Operations (Inventory > Detailed View)
Fix: Use VCF Operations > Inventory > Detailed View > VCF Instances > <instance> > Add Workload Domain > Create New (pp.1500-1501).

Task 3 Validate workload domain placement and network connectivity

manageability

A provisioned domain is not yet production-ready. You must validate that workloads can be placed, networks are segmented correctly, and edge nodes can route to external networks. This task covers the post-deployment checklist.

Step 1

Log in to the workload domain's own vCenter (the FQDN you entered in the wizard) and navigate to Hosts and Clusters. Click 'WLD-Cluster-01'. Verify: (a) 4 hosts are connected and green, (b) No alarms on the cluster, (c) Resource summary shows total CPU and memory available, (d) Updates > Image shows the cluster compliant with the assigned image.

Cluster summary: '4 hosts connected, 128 vCPU total, 384 GB RAM total (gross, same example host size as Task 1). vSAN ESA: 400 GB raw capacity per host (example), 1.6 TB raw across 4 hosts. DRS: Enabled. HA: Enabled with admission control.'
Step 2

Validate vSAN health. In vSAN, navigate to Monitor > vSAN > Cluster > Health. Check: (a) Object health: all objects Compliant (no degraded), (b) Disk health: all disks Healthy (vSAN ESA storage pool, or OSA disk groups), (c) Network latency: <1ms between hosts (shown in latency matrix).

vSAN health status: 'Cluster: Healthy, 4/4 nodes, 16 components, 100% compliance. Latency: <1ms. Disk space: 1.6 TB raw capacity (4 x 400 GB example), most of it free.'
Step 3

Test workload placement. Deploy a test VM into the new cluster. In vCenter, create a new VM: (a) Name: 'wld-test-01', (b) Datastore: select vSAN datastore of WLD-Cluster-01, (c) Network: select an NSX segment (if Tier-1 was created during provisioning) or default VLAN, (d) Size: 2 vCPU, 8 GB RAM. Power on the VM.

VM is created, powered on, and receives an IP address. VM status in vCenter: 'Running' on one of the 4 WLD hosts.
Step 4

Validate NSX connectivity. In NSX Manager, navigate to Fabric > Nodes > Host Transport Nodes. Verify all 4 WLD hosts appear and show Configuration State: 'Success' and Transport Node Status: 'Up'. Click on each host to view: (a) TEP IP address (should be from the NSX TEP pool, e.g., 10.2.3.5), (b) VLAN associations (Management, vMotion, vSAN, TEP), (c) No connectivity warnings.

NSX fabric: '4 host transport nodes, all Success/Up. TEP IPs: 10.2.3.5-10.2.3.8. No warnings.'
Step 5

Validate the NSX Tier-0 gateway you deployed in the post-deployment step (Centralized Connectivity). In NSX Manager, navigate to Networking > Tier-0 Gateways > WLD-T0-01. Verify: (a) Gateway status: 'Realized' (not Error), (b) Uplink segment: has an IP and is connected to the physical network, (c) BGP status: neighbour relationship 'Established' (if BGP is enabled), (d) Tier-1 gateways (or the Supervisor's VPC gateways) are attached and have routes. With Distributed Connectivity, validate the VPC external IP block and transit gateway instead.

NSX T0 gateway: 'Status: Realized. Uplink: 10.100.0.2 on VLAN 1700. BGP Neighbor 10.100.0.1 AS 65001: Established. 2 T1 gateways attached.'
Step 6

Test end-to-end connectivity. From the test VM (wld-test-01), ping an external IP (e.g., SDDC Manager 10.0.0.4 or a physical router 10.100.0.1). If the ping succeeds, NSX routing and the WLD network integration are working end-to-end.

Ping from wld-test-01 to 10.0.0.4: replies received. Round-trip latency <5ms. Network path: VM -> NSX segment -> T1 -> T0 -> physical uplink -> external network.

Validation Gate

Check: vCenter: WLD-Cluster-01 shows 4 hosts, resources available, no alarms. vSAN: all objects compliant, network latency <1ms. NSX: 4 transport nodes Success/Up, T0 gateway Realized, test VM connectivity working.

Expected: WLD is validated and production-ready for workload placement. All infrastructure components (compute, storage, network) verified.

Common Errors

Test VM cannot reach external network — ping to 10.100.0.1 times out
Cause: NSX T0 gateway is not connected to the physical uplink, or BGP neighbor is down
Fix: In NSX Manager, check T0 gateway: (1) Uplink segment status is 'Realized', (2) BGP neighbor is 'Established'. If neighbor is not established, check: (a) Physical router at 10.100.0.1 is reachable from NSX edge node, (b) BGP AS numbers match the configuration, (c) Edge node TEP IP is on correct VLAN.
vSAN cluster shows 1 host in 'degraded' state
Cause: One host lost network connectivity or has disk issues
Fix: Navigate to vSAN health and click the degraded host. Check: (1) Host is connected in vCenter, (2) Network latency to other hosts is normal, (3) Disk health is green. If degraded, host may need to be removed and re-provisioned.

Task 4 Understand stretched cluster architecture and plan multi-site deployment

availability

Stretched clusters are the VCDX-level topology: multiple sites, witness quorum, fault domains, cross-site replication. This task is architecture and planning; the actual stretched WLD deployment is Extension 1. Understanding the design constraints is critical for VCDX panelists.

Step 1

Review the vSAN stretched cluster failure model (VCF 9.0.1 documentation, vSAN Stretched Clusters and Table 855). In your lab notes, document: (a) Components: preferred site Site A (4 ESX hosts), secondary site Site B (4 ESX hosts — both data sites need the same number of hosts), witness host at a third site: three fault domains; the witness holds witness components (metadata) only. (b) Object layout: with Site disaster tolerance (site mirroring) each object has one full copy per data site plus a witness component on the witness host; accessibility is decided per object by the votes of its components, not by a count of hosts. (c) Failure scenarios: a data-site failure leaves the surviving site plus the witness with a majority of each object's votes, so the objects stay accessible and vSphere HA restarts the failed site's VMs on the surviving site. An inter-site link failure makes the witness side with the preferred site; the secondary site loses access to the objects and its VMs are restarted on the preferred site by vSphere HA. A witness failure leaves objects noncompliant but accessible, with running VMs unaffected. Losing hosts inside one site (for example two of Site A's four) is a local failure handled by the local FTT of the policy, not a quorum event. (d) Record the preferred-site choice and why (the site that must keep running when the link breaks). There is no read-only mode. Document this in a table.

Failure table: 'Fault domains: Site A preferred (4 hosts), Site B secondary (4 hosts — the sites must be symmetric), witness at a third site. (1) Site A down: Site B + witness hold the majority of each object's votes; Site B serves I/O; Site A VMs restart on Site B. (2) Site B down: Site A + witness; Site B VMs restart on Site A. (3) Inter-site link down: witness joins the preferred Site A; Site B loses object access and its VMs restart on Site A; resync when the link returns. (4) Witness down: objects noncompliant, data available, VMs unaffected. (5) Hosts lost inside one site: local FTT applies; no site-level event.'
Step 2

Understand witness host role and placement. The witness host does NOT participate in I/O — it only votes in quorum decisions. Placement: must be in a third location (third data center) or isolated network to avoid being partitioned along with one of the primary sites. In the lab, the witness can be on the same physical host as management, but on an isolated VLAN to simulate geographic separation. Document: (a) Witness host role: storage quorum voting, (b) Witness network isolation: separate VLAN with no data traffic, (c) Witness disk: typically a small disk (10-50 GB) used only for quorum metadata, not data storage.

Witness host architecture documented: 'Witness is a single vSAN witness host (appliance) at a third site, in the lab on isolated VLAN 1710 to simulate the third location. It stores witness components (metadata) only, no data, and never serves VM I/O. If the witness goes down, objects become noncompliant but remain accessible and running VMs are unaffected (Table 855); restore or replace it via Configure > vSAN > Fault Domains > Change witness host.'
Step 3

Design the stretched cluster network topology. For a 2-site stretched cluster: (a) Intra-site network latency: <1ms (within a data center, same network switch), (b) Inter-site network latency: less than 5 ms RTT between the data sites (VCF 9.0.1 docs p.1609, Table 825), (c) Inter-site bandwidth: at least 10 Gbps dedicated (25 Gbps recommended); witness to each data site less than 200 ms RTT for up to 10 hosts per site (less than 100 ms for 11-15 hosts per site; VCF 9.0.1 docs p.1609, Table 825). (d) Failure scenario: if the inter-site link fails, the witness host and the preferred site continue to service storage operations; the secondary site loses object access and vSphere HA restarts its VMs on the preferred site; site-tolerant objects resync when the link returns (Table 855). Document the network design.

Network topology: 'Site A (10.2.0.0/24 management, 10.2.3.0/24 vSAN) <--L3 Gateway--> Site B (10.3.0.0/24 management, 10.3.3.0/24 vSAN). Inter-site links are redundant (2x 10 Gbps). The witness runs at a third site, reachable from both data sites over routed L3: with witness traffic separation the data-site hosts send witness traffic from their management VMkernel adapters, routed to the witness appliance's vmk0 management network at the third site (VCF 9.0.1 docs p.1074, p.1077, p.1611; in this lab the witness sits on its isolated VLAN 1710), so whichever site keeps contact with it retains quorum. vSAN replication bandwidth: minimum 10 Gbps dedicated, RTT less than 5 ms.'
Step 4

Understand fault domains. In a stretched cluster, fault domains are used to ensure replicas are spread across sites. Configure: (a) Site A hosts assigned to Fault Domain A, (b) Site B hosts assigned to Fault Domain B, (c) the witness host is the third fault domain (it is added as the witness host of the stretched cluster, not as a member of a data fault domain). (d) Storage policy: Site disaster tolerance = Dual site mirroring (RAID-1 across the two data sites, witness component on the witness host) plus a local Failures to tolerate per site. Document the placement policy.

Fault domain configuration: 'WLD-Stretched-01: FD-A (4 hosts Site A), FD-B (4 hosts Site B), witness host at the third site. Storage policy: dual-site mirroring, one copy in FD-A and one in FD-B, witness component on the witness host; local FTT=1 within each site. If FD-A entire site fails, FD-B still has all data.'
Step 5

Compare traditional vs stretched tradeoffs. Create a comparison table: Traditional (single-site WLD) vs Stretched (2-site WLD). Rows: RTO (recovery time objective if site fails), RPO (recovery point objective — data loss), complexity (operational overhead), cost (extra site, WAN bandwidth), failure scenarios.

Comparison table: Traditional WLD RTO: Re-provision from snapshot if site is destroyed (2-4 hours), RPO: depends on backup frequency (0-24 hours). Stretched WLD RTO: immediate failover to Site B (< 1 minute), RPO: zero (synchronous replication). Complexity: Traditional = low, Stretched = high (quorum management, fault domains, witness). Cost: Traditional = low, Stretched = high (2x infrastructure, WAN, witness). Document which use case each is appropriate for.
Step 6

Document the provisioning steps for stretched WLD (you will execute in Extension 1 if time permits). Steps: (1) Plan fault domains and network topology, (2) Commission hosts from both sites, (3) Configure inter-site network (L3 reachability, latency validate), (4) Deploy witness host on isolated VLAN, (5) Provision stretched WLD specifying fault domain assignments, (6) Configure vSAN stretch sync settings (e.g., change rebuild timeout from 30 min to 60 min for remote replication), (7) Validate the partition behaviour (disable the inter-site link in a maintenance window: confirm the preferred site plus witness keep serving I/O and the secondary site's VMs restart on the preferred site; re-enable the link and watch the resync complete).

Stretched WLD provisioning checklist with 7 steps documented and ready for execution

Validation Gate

Check: Quorum model explained with failure scenarios. Witness host role and placement documented. Network topology designed with latency/bandwidth specs. Fault domains defined. Traditional vs stretched comparison table complete. Stretched provisioning checklist ready.

Expected: Comprehensive understanding of stretched cluster architecture and design constraints. Ready to architect or deploy a stretched WLD in production.

Common Errors

Confusion: does the witness participate in data storage or just voting?
Cause: Witness role is often misunderstood — it's easy to think it replicates data
Fix: Remember: Witness = quorum voting only (no data). It has a minimal disk (metadata only). This is why it can be deployed anywhere without capacity concerns.
Network partition scenario unclear: what happens if the inter-site link fails?
Cause: Expecting a host-count majority or a read-only state
Fix: When the inter-site link is down the witness host sides with the preferred site: Site A plus witness keep servicing storage; Site B loses access to the objects and vSphere HA restarts its VMs on Site A. The witness must be at a third site; co-locating it with Site A would take it down with that site (Table 855).

Task 5 Execute workload domain lifecycle operations: add hosts, expand cluster, and validate placement

manageability

Post-deployment, workload domains need day 2 operations: adding hosts to scale out, removing hosts, adding clusters. In VCF 9.0 these run in the vSphere Client against the SDDC cluster (Actions > Add Hosts > Add Unassigned Hosts), with SDDC Manager UI as the alternative; hosts must be commissioned into the free pool first. This task covers the operations and validation.

Step 1

Plan to add a 5th host to the traditional WLD. Commission esxi-09 into the free pool if needed (vSphere Client > Global Inventory Lists > Hosts > Unassigned Hosts > Commission Hosts) with the same principal storage type as WLD-Cluster-01. Verify the host is Active and matches the configuration of the existing hosts (unbalanced hardware raises a dismissible warning). If the cluster uses a static TEP IP pool, confirm it has a free address (p.1530).

esxi-09 commissioned, Active, in Unassigned Hosts with matching storage type
Step 2

In the vSphere Client browse to WLD-Cluster-01, click Actions > Add Hosts > Add Unassigned Hosts, select esxi-09 and click Next. On the vSphere Distributed Switch page review the cluster's vDS and NSX networking and assign the host's free vmnics to the existing vDS uplinks. Click Next, review host, switch configuration and licence details, and click Finish. The host uses the cluster's existing vDS for management, vMotion, vSAN and host overlay traffic. Monitor the task (vSphere Client Recent Tasks / VCF Operations Tasks). Estimated time: 15-30 minutes.

Add host task completes. esxi-09 appears in the cluster, connected; vSAN cluster grows from 4 to 5 nodes; NSX shows a 5th host transport node
Adding a host does not move existing data at once: components stay on the original 4 hosts until vSAN rebalances them. With Automatic rebalance enabled (Configure > vSAN > Services > Advanced Options > Edit), vSAN redistributes components onto the new host; on large clusters this can take hours (VCF 9.0.1 docs p.1718).
You can add or remove hosts on several SDDC clusters in parallel (p.1530). SDDC Manager UI remains the documented alternative.
Step 3

In vCenter, navigate to Hosts and Clusters > WLD-Cluster-01. Verify the 5th host (esxi-09) is now connected and green. Check vSAN health: it should show 5 nodes active; existing components are still on the original 4 hosts until vSAN rebalances them (next step).

vCenter cluster shows 5 hosts. vSAN: 5 nodes active; existing components still on the original 4 hosts until automatic rebalance moves some of them.
Step 4

Let vSAN rebalance the data across all 5 hosts. In the workload domain vCenter select the cluster > Configure > vSAN > Services > Advanced Options > Edit and confirm Automatic rebalance is enabled (VCF 9.0.1 docs p.1718: 'After you expand the vSAN cluster capacity, enable automatic rebalance'; p.1724); vSAN then moves components to the new host when the variance threshold is exceeded. Monitor progress under Monitor > vSAN > Resyncing Objects. Fault tolerance is maintained throughout.

Resyncing Objects shows rebalance activity moving components to esxi-09; disk usage converges across the 5 hosts
Step 5

While the rebalance runs, review VM storage policy compliance. In the workload domain vCenter open Policies and Profiles > VM Storage Policies and review the policy applied to the test VM (vSAN default or a custom policy such as 'vSAN-FTT1' with RAID-1). vSAN places components by policy, not by host preference; confirm the test VM (wld-test-01) is compliant under Monitor > vSAN > Virtual Objects.

Storage policy reviewed; test VM compliant; components may now use esxi-09
Step 6

Validate the expanded cluster. In vSAN Monitor > Skyline Health confirm: (a) all 5 nodes active, (b) no compliance violations, (c) resync complete (Resyncing Objects = 0), (d) capacity distribution across the 5 hosts is within the vSAN variance threshold. Document the final state.

Final validation: 'WLD-Cluster-01 now has 5 hosts. vSAN health green. All objects distributed across 5 hosts. Raw capacity increased from 1.6 TB to 2.0 TB (5 x 400 GB per host example). Ready for additional workloads.'

Validation Gate

Check: Workload domain vCenter: 5 hosts connected in WLD-Cluster-01. VCF Operations Inventory shows 5 hosts in the domain. vSAN: 5 nodes active, automatic rebalance complete, Skyline Health green.

Expected: WLD scaled out successfully. Day 2 operations (commission, add host, vSAN rebalance) demonstrated and validated.

Common Errors

Add Unassigned Hosts does not list esxi-09
Cause: The host is not commissioned, is not Active, or was commissioned with a different principal storage type
Fix: Commission esxi-09 from Global Inventory Lists > Hosts > Unassigned Hosts with the same storage type as the cluster; verify it is Active before retrying (p.1530).
vSAN rebalance is stuck at 45% for 30+ minutes
Cause: Network congestion on vSAN VLAN or insufficient I/O bandwidth
Fix: Check vSAN VLAN network latency: ping between hosts should be <1ms. If >5ms, check for network congestion (other large transfers). Rebalance is I/O heavy — if cluster is running production workloads, they may be slowing rebalance. Consider running rebalance during off-hours.
After adding the 5th host some VMs show non-compliant storage policy
Cause: A policy change or a resync still in progress
Fix: Wait for Resyncing Objects to reach zero, then re-check compliance under Monitor > vSAN > Virtual Objects; reapply the policy if still non-compliant.

Final Validation

You have planned and provisioned a traditional workload domain, validated network connectivity and placement, understood stretched cluster architecture including witness quorum and fault domains, and executed day 2 operations (host addition and rebalance). You can now architect workload domain topologies at the VCDX level, making data-driven decisions between traditional and stretched topologies based on RTO, RPO, cost, and complexity tradeoffs.

✓ Pre-provisioning Planning → Spec document, network validation, licensing check, capacity calculation, provisioning checklist — all complete

✓ Traditional WLD Provisioning → Workload-Domain-01 created from VCF Operations with its own vCenter and NSX Manager; 4 hosts, vSAN healthy, cluster image compliant; Edge cluster and Tier-0 deployed for Centralized connectivity

✓ WLD Validation → vSAN cluster compliant, NSX transport nodes Success/Up, test VM deployed and connected, T0 gateway realized, end-to-end connectivity verified

✓ Stretched Cluster Architecture → Quorum model documented, witness host role explained, fault domains designed, traditional vs stretched comparison table completed

✓ Day 2 Operations → 5th host commissioned and added from the vSphere Client, vSAN automatic rebalance completed, storage policy compliance verified, cluster validated at new size

Cleanup / Restore

Snapshot: vcp-admin-08-complete

• Power off test VM (wld-test-01): vCenter > VMs > right-click > Power Off > Remove from Inventory

• Document the WLD configuration: workload domain vCenter FQDN and SSO domain, cluster name, host list, cluster image, vSAN settings, NSX Manager (new/shared) and connectivity model, Edge cluster and Tier-0 configuration, network pool used. Save to your lab notebook.

• Take a final snapshot: 'vcp-admin-08-complete — Traditional WLD fully provisioned and validated, 5 hosts after expansion, vSAN rebalanced, Day 2 operations completed.'

• If performing Extension 1 (stretched WLD), do NOT revert yet. Keep management domain and traditional WLD intact for stretched deployment.

Design Reflection (VCDX)

A VCDX panelist examining your WLD design would ask: Why choose traditional vs stretched topology for a given business case? What is the actual RTO/RPO if Site A fails (all 4 hosts down)? How does the witness host quorum work, and what if the witness host fails? Why is the inter-site RTT limit (less than 5 ms between the data sites) critical? Can you expand the cluster from 4 to 5 hosts without downtime? How does NSX scale with WLD cluster size? What is the blast radius of a vSAN failure? Be prepared to architect a multi-site WLD that balances availability, cost, and operational complexity for a real enterprise use case.

Requirements

  • Workload domain must provide dedicated compute for tenant workloads with performance and availability guarantees isolated from the management domain
  • Cluster must scale from initial size (4 hosts) to larger size (8-16 hosts) without downtime or data loss
  • Availability: single-site WLD must tolerate 1 host failure (FTT=1); stretched WLD must tolerate entire site failure
  • Network connectivity: WLD hosts must reach external networks via NSX T0 gateway with redundant uplinks

Constraints

  • vSAN stretched cluster needs about 1 ms RTT within a site and less than 5 ms RTT between the two data sites (witness: less than 200 ms RTT for up to 10 hosts per site, less than 100 ms for 11-15) — cannot stretch a WLD across distant data centers (VCF 9.0.1 docs p.1609, Table 825)
  • Witness host for stretched cluster must be in third location (or isolated network) to avoid being partitioned with primary site
  • vSAN rebalance generates resync I/O; it runs online while VMs keep running, but on busy clusters schedule large expansions for quieter periods and keep enough free capacity for the resync
  • NSX scale: Tier-0 gateway capacity depends on edge node resources; edge nodes compete for compute in the cluster
  • Licensing: VCF 9.0 licenses are per core (VMware Cloud Foundation cores, minimum 16 cores per CPU) plus vSAN TiB, managed in VCF Operations and assigned to each WLD vCenter (VCF 9.0.1 docs pp.77, 1462). A stretched topology needs a witness appliance at a third site and inter-site bandwidth; its site failover is a vSphere HA restart. Site Recovery Manager is only relevant for a separate replication-based DR design between independent sites.
  • Every workload domain has a dedicated vCenter deployed in the management domain (SSO joined or isolated); only NSX Manager can be shared between workload domains (p.496)

Assumptions

  • All ESX hosts have vSAN-certified local NVMe storage for the chosen architecture (vSAN ESA storage pool, or vSAN OSA disk groups) — example sizing: at least 200 GB per host for the lab, 1+ TB for production
  • Network infrastructure supports dedicated VLANs for management, vMotion, vSAN, NSX TEP, and uplink traffic with adequate bandwidth (10 Gbps minimum per host for vSAN)
  • Operator has the VCF Operations Administrator role (and SDDC Manager Administrator) and understands vSphere, NSX, and vSAN administration
  • For stretched topology: inter-site network link is stable and redundant (2+ independent paths); witness host has guaranteed isolation from primary sites

Risks

  • Both data sites of a stretched cluster lost, or one site and the witness lost together — IMPACT: objects lose vote majority and become inaccessible until a site or the witness returns, MITIGATION: three independent locations, redundant links, monitored witness
  • Inter-site network partition on stretched cluster — IMPACT: the secondary site loses object access and its VMs restart on the preferred site (preferred site plus witness keep serving I/O); resync when the link returns, MITIGATION: redundant L3 inter-site links, correct preferred-site choice, HA capacity on the preferred site
  • vSAN rebalance triggers cascading failures if hosts are near capacity — IMPACT: cluster becomes unhealthy, MITIGATION: maintain 20-30% free space on vSAN datastore
  • Witness host failure in stretched cluster — IMPACT: objects noncompliant but accessible, running VMs unaffected; a second failure (a data site) during that window makes objects inaccessible, MITIGATION: witness at a true third location, monitored, replaceable quickly (Change witness host)
  • NSX T0 uplink failure isolates WLD from external networks — IMPACT: workloads lose external connectivity, MITIGATION: use redundant edge nodes and multiple uplink connections

Self-Assessment Discussion Prompts

  1. You are designing a WLD for an e-commerce platform. Business requires <1 minute RTO if a single host fails and <1 hour RTO if entire data center fails. Which topology (traditional or stretched) would you recommend, and why?
  2. In a stretched WLD with 4+4+1 (witness) configuration, the inter-site link fails for 2 hours. What is the state of each site? Which site keeps serving I/O and why (preferred site plus witness)? What happens to the VMs that were running on the secondary site, and what happens when the link returns?
  3. You plan to add 8 new hosts to an existing 4-node WLD (doubling capacity). The vSAN rebalance is projected to take 8 hours. How would you execute this to minimize workload disruption?
  4. Compare NSX scale in a traditional 4-node WLD vs stretched 7-node WLD. How many edge nodes would you deploy in each case, and where would they run?
  5. A security audit requires all VMs in the WLD to be separated by fault domain (no two replicas on the same host, no two replicas in same failure domain for stretched). How would you design vSAN FTT and storage policies to meet this requirement?
  6. Stretched WLD witness host has a hardware failure and cannot be replaced for 24 hours. What is the state of the WLD? Can VMs still run (yes, objects are noncompliant but accessible)? What is the exposure if a data site fails in that window, and can new VMs be created?

Extensions

Provision a Stretched Workload Domain (2-site topology)

Extend the lab to provision a stretched WLD using the architecture and planning from Task 4. Deploy: Site A with 4 ESX hosts, Site B with 4 ESX hosts, witness host at a third site. Configure fault domains and vSAN replication across sites. Execute a network partition test (disable inter-site link) to validate quorum behavior. This is the advanced topology that VCDX candidates must master.

harder

Design and Validate a Multi-Tier NSX Network for Workload Domain

Create a multi-tier NSX network architecture for the WLD: Tier-0 gateway for external routing, multiple Tier-1 gateways for tenant segmentation, overlay segments for different application tiers (web, app, db). Create workloads on different segments, enforce micro-segmentation policies, and validate east-west traffic. This demonstrates VCDX-level NSX design.

harder

Compare VCF 5.2 vs VCF 9.0 WLD Provisioning Workflows

If you have access to a VCF 5.2 lab, repeat the WLD provisioning (Task 2) in VCF 5.2, where the wizard ran from SDDC Manager (Inventory > Workload Domains > + Workload Domain). Document differences: creation from VCF Operations in 9.0, Supervisor enabled by default, Centralized vs Distributed (VPC) connectivity, cluster image selection in the wizard, NSX VIBs in the ESX image, clusters and hosts added from the vSphere Client. Write a 2-page comparison addressing what was improved in VCF 9.0, what was deprecated, and how the provisioning experience evolved.

same

Design and Implement WLD Placement Policies for Multi-Tenant Isolation

Design a placement policy strategy for a multi-tenant environment: Tenant A VMs must not run on the same physical host as Tenant B, each tenant has dedicated storage quota, each tenant has dedicated NSX segment. Use vCenter VM groups, host groups, and affinity rules to enforce these policies. Validate that workload placement complies with multi-tenant constraints.

harder

References

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