VMware Cloud Foundation 9.1 Upgrade - Part 4: NSX and vCenter

🗓️ October 7, 2026 · Home Lab, How-To's

With the management components handled in Parts 1 to 3, Part 4 moves to the core infrastructure: NSX Manager and vCenter. The series overview has the environment details and prerequisites.

Everything here is driven from the VCF Operations Lifecycle UI, the same place I’ve worked since Part 2. Broadcom’s flow is to plan the domain upgrade, run the prechecks, then work through the sequence one component at a time, NSX Manager first and then vCenter (Upgrading vCenter and NSX Manager).

Where I’m starting from

Before touching anything, I checked where everything stood. Both domains were ACTIVE, and both vCenters and both NSX Managers were still on 9.0.2. Each domain has a single NSX Manager node.

What has to be in place first

Broadcom’s prerequisites call for backups of SDDC Manager and every vCenter and NSX Manager instance, either file-based or image-based, plus a cold snapshot of SDDC Manager (upgrade prerequisites).

The prerequisites page lists a few more things. None of the vCenters should be joined to an Active Directory domain, and there shouldn’t be any expired or soon-to-expire passwords or certificates on the components being upgraded. The vSAN HCL database has to be updated by hand, and the hardware has to be on the HCL. There should be no failed workflows, nothing in an activating or error state, and no active alarms. ESX hosts with TPM modules in use need the latest 2.0 firmware, and if they aren’t in use they must be disabled in the BIOS (KB 312159 covers TPM support). The vCenters also have to be licensed through the new License Server from Part 2, and KB 440471 documents the licensing errors that follow the vCenter upgrade when they aren’t.

Getting ready

I made sure both vCenters were on UTC, ran Backup Now from Administration → Backup in SDDC Manager, ran a file-based backup of each vCenter from its VAMI page, and backed up the NSX Managers too. Then I took the SDDC Manager snapshot, with the same no-memory, no-quiesce settings as Parts 1 and 2:

Get-VM -Name 'vcf-m01-sddcm01' -ErrorAction Stop |
    New-Snapshot -Name 'Pre-Part4-SDDC-Manager-9.1.1' `
        -Description 'Before NSX and vCenter upgrade to 9.1.1' `
        -Memory:$false -Quiesce:$false

PowerCLI creating the SDDC Manager snapshot

Downloading the core component binaries

The target-version binaries have to be downloaded before the upgrade. This time I used the Binary Management tab on my VCF instance in VCF Operations, under Build → Lifecycle → VCF Instances, with VCF Version set to 9.1 and the Upgrade Binaries tab selected. I ticked ESX, NSX, and vCenter at 9.1.1.0 and downloaded them. SDDC Manager already showed Downloaded from Part 2.

The Binary Management tab with the ESX, NSX, and vCenter 9.1.1.0 upgrade binaries selected

The three rows went from Pending to Scheduled and then to Downloaded.

The three upgrade binaries scheduled for download

ESX, NSX, and vCenter 9.1.1.0 showing Downloaded

Planning the domain upgrade

I selected vcf-m01 under my instance in Build → Lifecycle → VCF Instances and opened the Upgrades tab (plan an upgrade). It has Plan Domain Upgrade and Plan Component Upgrade buttons, and I used Plan Domain Upgrade, which is a four-step wizard.

The Upgrades tab for the vcf-m01 management domain

  1. Target Version. I picked 9.1.1.0. A note on the page points to KB 90074 for turning the compatibility checks on the target version on or off. The 9.1.1.0100 patch covers Fleet Lifecycle, SDDC Lifecycle, and the migration service engine, not vCenter or NSX (patch releases).

    The Target Version step with 9.1.1.0 selected

  2. Customize Target Version. The wizard listed the components in the domain with their current and target builds, and I left the proposed targets alone.

    The Customize Target Version step with the current and target builds

  3. vCenter and NSX Upgrade Options. Optimized Maintenance Window Upgrade does the vCenter reduced downtime preparation first, with no downtime, and then upgrades NSX Manager and does the vCenter switchover inside the maintenance window. It requires the vCenter target version to be the same as SDDC Manager’s current version. Sequential Upgrade does everything in the maintenance window, NSX Manager first and then vCenter, and the vCenter can go in place or with reduced downtime. I chose Sequential Upgrade, and the wizard says I can revisit that choice until I start the NSX Manager and vCenter upgrades.

    The vCenter and NSX Upgrade Options step with Sequential Upgrade selected

  4. Plan Overview. The overview laid out the order: NSX Manager, vCenter, ESX, then the NSX Edge clusters and the NSX Upgrade Finalize.

    The Plan Overview step ending in Submit Plan

I selected Submit Plan. The Upgrades tab then showed the plan with Edit Plan and Cancel Plan buttons, and the NSX Manager tile offered Upgrade Now, Schedule, and Run Prechecks, while the vCenter and ESX tiles were still grayed out.

The Upgrades tab after submitting the plan, with the upgrade sequence

Running the precheck

I clicked Run Prechecks and left it on Target version precheck for 9.1.1.0 with every component in scope, which covered vCenter, the vcf-mgmt-cl cluster, and NSX. The dialog warns that the NSX Upgrade Coordinator is upgraded to the target version during the precheck and then rolled back to restore the original state.

The Run Prechecks dialog with all three assets selected

The precheck ran for about 11 minutes.

The precheck at 92% completed

It came back with 3 errors, 5 warnings, and 1 silenced. An upgrade can be attempted even when a precheck fails, but it’s recommended to review the errors and warnings first (perform an upgrade precheck).

The precheck results with 3 errors, 5 warnings, and 1 silenced

All three errors were the same vSAN data migration what-if check, one for each ESX host, with the same result. The one I opened rates the impact High, says the what-if evacuation with “ensure access” failed, and tells me to run the what-if test manually and remediate what it reports.

The three vSAN data migration what-if errors

So I ran it manually, from the vSphere Client under vcf-mgmt-cl → Monitor → vSAN → Data Migration Pre-check, for vcf-m01-esx01 with Ensure accessibility. The host can enter maintenance mode, but all three tabs carried a warning. One native trace object would become non-compliant.

The Objects state tab with the native trace object

The Cluster Capacity tab before and after evacuating esx01

The predicted health listed one object with reduced availability and no rebuild, and storage space over the warning threshold.

The predicted vSAN object health

The predicted storage space against the thresholds

Those are predictions for a host in maintenance mode, which doesn’t happen until the ESX upgrade in Part 5.

The five warnings split between the hosts and NSX. Each host had a third-party VIB warning for William Lam’s Nested vSAN ESA Mock HW VIB. The warning says it can be ignored when a custom ISO is used to upgrade a VUM-based cluster, and asks that any custom VIBs be present in the new image.

The ESXi third party VIBs warning for esx01

The two NSX warnings, Check for data inconsistencies in DB and IP reputation outbound rules disabled warning, were about the system-created FPROT context profile and the default malicious IP rule. The context profile contains an App ID that vDefend 9.1 no longer supports, and KB 410572 says no action is needed beyond acknowledging the warning. After the upgrade is finalized, the unsupported App ID has to be removed from the profile before the profile is used in a rule.

The FPROT context profile warning

The other one said the upgrade will disable the default ‘Malicious IP at Destination Rule.’ That’s the 9.1.1 default, and the outbound rule can be activated manually if I want it (Default Outbound Rule, Activate Malicious IP Filtering Feed).

The IP reputation outbound rules warning

The silenced result was the vSAN disk group mode check on vcf-mgmt-cl, skipped because that vSAN health test is silenced in vSAN. vCenter had no errors or warnings.

The silenced vSAN disk group mode check

Updating NSX Manager

The Upgrade Sequence on the Upgrades tab lists the NSX Manager, vCenter, ESX, and NSX Edge and Finalize tiles in that order. The ESX hosts, the Edge clusters, and the finalize are all in Part 5.

The Upgrades tab and its upgrade sequence

Broadcom’s sequential procedure and the planner’s NSX Local Managers phase both start with Configure under VMware NSX Manager Nodes Upgrade (sequential upgrade).

  1. In VCF Operations, select Build → Lifecycle, expand VCF Instances, and select the management domain.
  2. On the Upgrades tab, find VMware NSX Manager Nodes Upgrade.
  3. Optionally, click Run Prechecks on the tile to check NSX on its own.
  4. Click Upgrade Now, or click Schedule to pick another time.
  5. Follow it with View Status and in the Tasks panel.
  6. When all the NSX Manager nodes are upgraded, a green banner appears on the status page. Back on the Upgrades tab, the Upgrade Sequence step count goes up by one.

I ran the tile’s precheck just to see what it would add, and it found nothing new. It’s labeled a Configuration precheck, covers NSX alone, and came back with 0 errors, 2 warnings, 32 passed, and 0 silenced. The two warnings were the same two from the target version precheck, and the vSAN what-if errors aren’t in it, since they’re ESX checks.

The NSX Manager tile after its own precheck

The NSX configuration precheck report with 0 errors, 2 warnings, and 32 passed

Then I clicked Upgrade Now, and the tile switched to “Upgrade is in progress.” In the Tasks panel it showed up as “Upgrade DOMAIN - [vcf-m01] using BUNDLE - NSX_T_MANAGER:9.1.1.0.25691509,” with three subtasks: proceeding with the schedule once the precheck is complete, upgrading the upgrade coordinator, and upgrading the NSX Manager.

The NSX upgrade task and its three subtasks in the Tasks panel

View Status breaks the work down further. The upgrade coordinator goes first, with a cluster precheck, a backup, setting the upgrade payload, accepting the EULA, and updating the coordinator itself. The Manager node follows with its own precheck, the upgrade, and a postcheck.

It finished in about 42 minutes. The status page ended on a green banner reading “Upgrade completed with success,” with the upgrade coordinator, the NSX management cluster at 9.1.1.0.25691509, and the Manager node all marked Upgraded.

The finished NSX upgrade with every stage complete

Back on the Upgrades tab, the Upgrade Sequence count had gone from 1 of 5 to 2 of 5 steps completed and the NSX Manager tile was gone from the list. The Configure button on the vCenter tile was active now, and the ESX and NSX Edge and Finalize tiles were still grayed out.

The Upgrades tab with the NSX Manager step complete and Configure active on the vCenter tile

Updating vCenter

The vCenter tile’s Configure button was grayed out until the NSX Manager upgrade finished, and then it became active. The wizard offers two mechanisms. The vCenter Reduced Downtime Upgrade brings up a second appliance, migrates the configuration over, and cuts the downtime to around 5 minutes, but it needs additional capacity to run both appliances at the same time. The vCenter Regular Upgrade takes vCenter down for the whole upgrade. I chose the Regular Upgrade.

The Configure Upgrade wizard with vCenter Regular Upgrade selected

  1. Under VMware vCenter Upgrade, click Configure.
  2. Choose vCenter Regular Upgrade and click Next.
  3. Select a backup option and click Next. A file-based backup of the vCenter appliance is Broadcom’s best practice, and I took mine under Getting ready. The vCenter inventory and settings shouldn’t change between that backup and the end of a successful upgrade.
  4. Review the configuration and click Finish.
  5. On the tile, click Upgrade Now.
  6. Follow the upgrade with View Status and in the Tasks panel.

After Finish, the tile said “You configured the in place upgrade for the VMware vCenter appliance,” with Upgrade Now, Schedule, Run Prechecks, and Cancel Upgrade. I clicked Upgrade Now, and the tile switched to “Upgrade is in progress.”

The vCenter tile after the wizard

View Status broke the upgrade into ten stages, starting with pre-validation, unstage, set repo, and a precheck of its own. It finished in about 1 hour and 6 minutes, with “Upgrade completed with success” and vcf-m01-vc01.demo.lab marked Upgraded at 9.1.1.0.25712839.

The finished vCenter upgrade

I rechecked log collection for vcf-m01-vc01 in VCF Operations, and vCenter Log Collection was still turned on.

Configuration upgrades

Configuration upgrades can show up after a software upgrade. One can be applied right away or after all the software upgrades are done, it has to happen in a maintenance window, and pending ones don’t block later upgrades (apply configuration upgrades).

After the vCenter upgrade, one showed up under Available Configuration Upgrades: Services config in vCenter, type FEATURE, resource type DOMAIN. The Upgrade Sequence count read 3 of 5, and Configure on the ESX tile was active. I’m leaving it until the other upgrades are done, following Broadcom’s upgrade order, which lists the configuration upgrades last (upgrade order). I apply it in Part 6.

The Upgrades tab after the vCenter upgrade, with a configuration upgrade available

Deleting the SDDC Manager snapshot

I did this later than the rest of Part 4, once the ESX precheck in Part 5 flagged storage on the cluster. The NSX Manager and vCenter upgrades are done, so the Pre-Part4-SDDC-Manager-9.1.1 snapshot from Getting ready has served its purpose. The first line below lists the snapshot and its size, and the second removes it.

Get-VM -Name 'vcf-m01-sddcm01' | Get-Snapshot | Select-Object VM, Name, SizeGB
Get-VM -Name 'vcf-m01-sddcm01' | Get-Snapshot -Name 'Pre-Part4-*' | Remove-Snapshot -Confirm:$false

Removing the SDDC Manager snapshot with PowerCLI

What’s next

In Part 5, I take on the ESX hosts and finish the NSX upgrade. If the NSX or vCenter stages went differently for you, drop a comment and let me know.

-virtualex-

Comments