A VCF 9.1 deployment fails at the pre check stage far more often than it fails during bring up, and almost always for reasons you could have cleared a day earlier. This part is the full readiness checklist, so you open the VCF Installer with every host, record and address already in place. Work through it top to bottom and tick each item off before you download a single bundle.

You will confirm host counts and ESXi builds, prepare forward and reverse DNS, set NTP on every host, plan a management IP block, lay out VLANs and MTU, reserve the ranges that VCF Management Services and VCF Automation need, and record all of it in the Planning and Preparation Workbook. Two DNS servers and two NTP servers are required, and both must live outside the VCF instance you are about to build. Whichever path you take, greenfield, converge or import, these prerequisites gate the start.

Prerequisites

Group the requirements by area so nothing slips through. Each area below has a target value you can verify today. Anything you cannot confirm now is a task to close, not a footnote to skip, because the installer will stop on it later.

flowchart TD
A[Confirm hosts and ESXi build] --> B[Prepare forward and reverse DNS]
B --> C[Set NTP on hosts and appliance]
C --> D[Plan management IP block]
D --> E[Lay out VLANs and MTU]
E --> F[Reserve Management Services and Automation ranges]
F --> G[Complete the Planning Workbook]
G --> H[Run installer pre checks]
H --> I[All green, ready for bring up]

Step 1, size and prepare the ESX hosts

Confirm host count, a single vendor, clean ESXi, and fleet sizing.

1. Provide a minimum of three ESX hosts for a vSAN management domain, whether you run vSAN OSA or ESA, or two hosts if you use external storage (Fibre Channel VMFS or NFS v3).
2. Use hosts from a single hardware vendor, since mixed vendor hardware in the management domain is not supported.
3. Install ESXi fresh on each host and confirm the physical health status reads healthy.
4. Make sure each host carries a non-expired license.
5. Size for the whole fleet: a highly available medium deployment needs four hosts giving roughly 220 vCPU, 728 GB RAM and about 10.7 TB of usable disk.

Step 2, match the ESXi build to the bill of materials

Match every host to the exact build in the bill of materials.

1. Run the exact ESXi build listed in the VCF 9.1 bill of materials, not merely a matching major version.
2. Download the build called out in the BOM, install it cleanly, and record the build number next to each host in your workbook.
3. Avoid a newer or older patch than the BOM specifies, because the host is refused at commission time.
4. If converging or importing, confirm the estate floor: vSphere 8.0 Update 3a or later for converge, and vSphere 8.0 Update 3 with NSX 4.2 or later for import.

Step 3, prepare forward and reverse DNS

Create matching forward and reverse records for every name.

1. Create an A record and a matching PTR record for every component FQDN: the VCF Installer, vCenter, SDDC Manager, each NSX Manager, VCF Operations, and each VCF Management Services node.
2. Point at two DNS servers hosted outside the VCF instance, so a half built fleet cannot take out its own name resolution.
3. Confirm reverse lookup returns the exact FQDN, since the installer rejects any address whose PTR is missing or wrong.
4. Use fully qualified names everywhere, because short names in the inventory will stop you.
5. Test each name both ways before you move on.

Step 4, set NTP on every host and appliance

Time is not optional in VCF, because certificate trust and login tokens across the fleet depend on clocks staying in sync. Configure two NTP servers reachable from both the ESXi management network and the VM management network, and host them outside the instance for the same reason as DNS. On each ESXi host set the NTP service policy to start and stop with host, so the service comes up with the host every time. Keep skew to a few seconds at most. A clock that drifts even slightly will break single sign-on and token exchange in ways that look like unrelated failures later, so fix time first and verify it before you trust anything else.

On a standalone ESXi host, set NTP from the ESXi Host Client like this.

1. Open a browser to the ESXi Host Client, at the host FQDN followed by /ui, and log in as root.
2. In the left pane, click Manage.
3. Click the System tab.
4. Click Time & date.
5. Click Edit NTP Settings.
6. Select Use Network Time Protocol, which enables the NTP client.
7. Set the NTP service startup policy to Start and stop with host.
8. In the NTP servers field, enter your two NTP server addresses, separated by a comma.
9. Click Save.
10. Back on Time & date, select the NTP service, click Start, and confirm it shows Running.

Step 5, plan the management IP block

Reserve a contiguous management block sized for growth.

1. Reserve a contiguous block on the management network for the mandatory VCF Management Services cluster. For exact counts by scenario, see the VCF 9.1 IP address requirements reference.
2. Use a /27 for thirty usable addresses and room to grow, the recommended size.
3. Treat a /28 with fourteen usable as the minimum, which leaves almost no headroom.
4. Place the core appliances inside the block: vCenter, SDDC Manager, NSX, VCF Operations and the installer, plus the two reserved ranges from Step 7.
5. Make sure the block does not overlap any existing network pool, since overlapping pools are rejected.

Step 6, lay out the physical network, VLANs and MTU

Give each traffic type a VLAN and set jumbo frames end to end.

1. Assign a VLAN ID and CIDR to each of Management, vMotion, vSAN and NSX Overlay, and record the gateway for each.
2. Set MTU 9000 on vMotion, vSAN and NSX Overlay so they carry jumbo frames.
3. Configure that same MTU end to end on the physical switches, not just the host, or large frames drop silently.
4. Run management traffic at 1500 or 9000 depending on your design.
5. Confirm each VLAN is trunked to every host, since the distributed switch the installer builds uses 9000.

Step 7, reserve the IP ranges for VCF Management Services and VCF Automation

Hand the installer two contiguous ranges.

1. Provide the VCF Management Services range as a range, a minimum of twelve consecutive IP addresses.
2. Provide the VCF Automation range as a range, at least five consecutive IP addresses.
3. Reserve the Automation range even if you do not deploy Automation on day one, so the block does not need re-carving later.
4. Write down the first and last address of each range and keep both inside the management block from Step 5.
5. Make sure neither range overlaps the appliance addresses or each other, since this most often pushes a /28 over its limit.

Step 8, complete the Planning and Preparation Workbook

Capture every wizard value in the Planning and Preparation Workbook.

1. Open the Planning and Preparation Workbook, the spreadsheet that captures every value the installer wizard asks for.
2. Fill in networks and CIDRs, VLAN IDs, MTU, IP pools and ranges, gateways, host FQDNs, management addresses and the credentials for each component.
3. Treat the workbook as the single source of truth, so a blank or guessed field does not become a wrong value typed under time pressure.
4. Cross check every cell against DNS and your addressing plan before you open the wizard.

Step 9, run the installer pre checks

Deploy the installer and clear its prechecks before you commit.

1. Deploy the VCF Installer appliance at 4 vCPU, 16 GB RAM and 914 GB disk, on a network that can reach the ESX hosts, VM management, DNS and NTP.
2. Run the full validation set: forward and reverse DNS, NTP reachability, NIC speed, disk eligibility, IP availability and required ports.
3. Fix anything that is not green and run the checks again.
4. Treat a clean pass as the signal your prerequisites are complete, since clearing a red check now is far cheaper than stopping a bring up halfway.

Verify readiness before bring up

Confirm each prerequisite with a test rather than a glance. For DNS, run a forward and a reverse lookup on every FQDN and check that the reverse result matches the forward name exactly. For NTP, compare the clock on each host and appliance and confirm skew is within a few seconds. For MTU, send a jumbo frame across each network with the do not fragment bit set and a payload of 8972 bytes, which proves 9000 end to end. For addressing, confirm the reserved ranges are free and do not overlap. Then read the installer pre check page, where DNS, NTP, NIC speed, disk eligibility and IP availability should all show green. When every one of those passes, you are ready to move to the network setup in Part 3 and the installer deployment in Part 4.

Common errors and fixes

Reverse DNS is missing or wrong.
A pre check fails when a PTR record returns nothing or a name that does not match the A record. Add or correct the PTR so the reverse lookup returns the exact FQDN, then retest both directions.

NIC speed validation fails.
Validation flags NICs slower than 10GbE. Use certified 10GbE adapters for a production build. For a lab, apply the documented VCF Installer workaround that relaxes the NIC speed check, and treat it as lab only.

Disk eligibility fails.
Disks that still hold partitions, or devices not recognised for vSAN, are rejected. Clear existing partitions, present empty certified devices, and confirm the cache and capacity tiers are correct before you retry.

IP range is too small.
A /28 with fourteen usable addresses cannot hold twelve for Management Services, five for Automation and the core appliances at once. Move the management block to a /27 and re-enter the ranges.

Clock skew breaks sign on.
Certificate and token errors that look unrelated often trace back to time drift. Correct NTP on the affected host or appliance, confirm skew is within a few seconds, then run the checks again.

Common questions

How many hosts does the management domain need
Three ESX hosts if you use vSAN, whether OSA or ESA, or two hosts if you use external storage such as Fibre Channel VMFS or NFS v3. All management domain hosts must come from a single hardware vendor.

What subnet size should I reserve
A /27 with thirty usable addresses is recommended and leaves room to grow. A /28 with fourteen usable is the documented minimum, and it is tight once you reserve twelve IPs for Management Services and five for Automation.

How large is the VCF Installer appliance
Size it at 4 vCPU, 16 GB of RAM and 914 GB of disk, and give it network access to the ESX hosts, the VM management network, DNS and NTP.

Do converge and import use the same prerequisites
Forward and reverse DNS, NTP and fully qualified names apply to every path. Converge additionally needs vSphere 8.0 Update 3a or later, and import needs vSphere 8.0 Update 3 with NSX 4.2 or later.

References

• Broadcom TechDocs: Preparing ESX Hosts for VCF or vSphere Foundation
• Broadcom TechDocs: Commission ESX Hosts (VCF 9.1)
• Broadcom TechDocs: Deployment, Convergence, and Upgrade (VCF 9.1)
• William Lam: VCF 9.1 VCF Installer and SDDC Manager Configuration Workarounds

About The Author

Dr. Pranay Jha

Dr. Pranay Jha is a Cloud and AI Consultant with 18+ years of experience in hybrid cloud, virtualization, and enterprise infrastructure transformation. He specializes in VMware technologies, multi-cloud strategy, and Generative AI solutions. He holds a PhD in Computer Applications with research focused on Cloud and AI, has published multiple research papers, and has been a VMware vExpert since 2016 and a VMUG Community Leader.

See author's posts