, ,

What Is New in VCF 9.1 and the Case for Upgrading (VCF 9.1 Upgrade Series, Part 1)

VCF 9.1 adds a mandatory VCF Management Services cluster and reshuffles the management plane. Here is what changed, why to upgrade from 9.0.x, and the order the upgrade runs in.

VCF 9.1 Upgrade · Part 1 of 14

TL;DR · Key Takeaways

  • VCF 9.1 adds VCF Management Services, a container cluster that hosts fleet lifecycle, SDDC lifecycle, the software depot and the new license server. It is mandatory, not optional.
  • The 9.0 Fleet Management Appliance is gone. Its job moves into two services, Fleet Lifecycle and SDDC Lifecycle, inside VCF Management Services, and the old appliance is powered off and decommissioned.
  • The sequence starts with VCF Operations, not SDDC Manager. Ops first, then SDDC Manager, then deploy VCF Management Services, then NSX, vCenter and ESX.
  • The standalone Identity Broker cluster from 9.0 folds into VCF Management Services, and licensing moves to a central VCF License Server.
  • Plan for a /27 on the management network and a Cloud Proxy in the instance that hosts VCF Operations before you start.
  • Start with the VCF 9.1 Upgrade Planning Tool. It generates a phase-by-phase plan for your exact starting point.

I have upgraded enough VCF fleets to know the hard part of a 9.1 upgrade isn’t clicking buttons. It’s the surprise halfway through, when a component you forgot about blocks the whole plan. 9.1 moves more furniture than a normal dot release, so before any bundle gets downloaded, the job is to get the mental model right.

This first part is the map for the rest of the series. I’ll cover what actually changed, why it’s worth the maintenance window, and the exact order the upgrade runs in. Read this one before you touch anything, then follow the planning parts that come next. If you want the component-level detail on Operations itself, my VCF 9 Operations series covers it, and for a side by side of the two releases see my 9.0 vs 9.1 breakdown.

What actually changed in 9.1

Most of what’s new in 9.1 is a reshuffle of the management plane. The workload side (vSphere, vSAN, NSX) upgrades much like before. The part that’s different is how the fleet manages itself.

VCF Management Services, the new center of gravity

The headline change is a component called VCF Management Services. It’s a container cluster that Broadcom describes as a common runtime for the fleet’s lifecycle and operational pieces, and it is mandatory. You cannot upgrade the rest of the stack to 9.1 until it’s deployed. Inside it live the parts that used to be spread across separate appliances: fleet lifecycle, SDDC lifecycle, the software depot that feeds updates, and the new license server.

If you ran 9.0, think of this as where the Fleet Management Appliance’s job went, except broken into services and run as containers instead of a single VM.

What got decommissioned

Three things from 9.0 go away during the upgrade. The standalone Fleet Management Appliance is replaced by two services, Fleet Lifecycle and SDDC Lifecycle, that now run inside VCF Management Services. Once its data is transferred, the old appliance is powered down and ready to decommission. The external Identity Broker cluster (vIDB) folds into VCF Management Services and its standalone VMs are powered off. And per component or Aria based licensing gives way to a central VCF License Server. None of these are optional. The upgrade path assumes them.

VCF 9.0VCF 9.1SDDC ManagerFleet Management ApplianceIdentity Broker cluster (standalone)Licensing per component / AriaupgradeSDDC ManagerVCF Management ServicesFleet LifecycleSDDC LifecycleSoftware DepotLicense ServerIdentity Broker (folded in)
Four separate 9.0 pieces collapse into one mandatory VCF Management Services cluster in 9.1.

Why upgrade, the honest case

Faster patches

The reason I’d push a 9.0.x shop to move sooner than later is the patch cadence. Once you’re on 9.1 with the software depot and fleet lifecycle in place, updates and patches flow through one mechanism instead of you chasing bundles per product. Broadcom has been clear that getting to 9.1 is what lets you take the faster patch releases.

One license server

Licensing in the 9.0 world was spread out and easy to get wrong. The central license server in 9.1 gives you one place to hold and assign entitlements, with local control and room to scale. If you have ever fought a license assignment failure on a vCenter, this is the change that removes that class of problem, once it’s set up correctly.

Security and staying current

There’s a plainer reason too. Running an older VCF means running a larger attack surface. Each release folds in fixes across vSphere, NSX and the management plane. Staying a version behind on infrastructure this central isn’t a position I’d want to defend to an auditor.

FunctionVCF 9.0VCF 9.1
Fleet lifecycleFleet Management Appliance (VM)Fleet Lifecycle service in VCF Management Services
SDDC lifecycleSDDC Manager drivenSDDC Lifecycle service in VCF Management Services
Update sourcePer product bundlesSoftware Depot in VCF Management Services
LicensingPer component / AriaCentral VCF License Server
IdentityStandalone Identity Broker cluster (vIDB)Folded into VCF Management Services

Which component upgrades first?

Here’s the part people get wrong. The sequence does not start with SDDC Manager. It starts with VCF Operations.

VCF Operations is mandatory in 9.x, and on a 9.0.x fleet it’s the component that owns fleet management. The upgrade uses the Operations upgrade to move that fleet management data into the new VCF Management Services cluster, then decommissions the old Fleet Management Appliance. If you tried to upgrade SDDC Manager or deploy Management Services first, you’d be moving furniture before the truck arrived.

The order that works: run prechecks, upgrade VCF Operations, upgrade SDDC Manager to 9.1, deploy VCF Management Services and the license server, then upgrade the management domain core as NSX, vCenter, ESX, and finish the NSX Edge cluster.

One detail that trips people: NSX is two steps

In 9.x the NSX VIBs ship inside ESX, so NSX upgrades in two passes. You upgrade the NSX management stack first, then finalize by upgrading the Edge cluster after vCenter and ESX are done. If you treat NSX as one step, you’ll wonder why the workload domain won’t show as finished.

Before you start: back up SDDC Manager, vCenter, NSX Manager and VCF Operations, and confirm you can actually restore them, not just that a backup job ran green. Any vCenter upgrade and any host remediation is production affecting. Do it in a change window and know your rollback before you click Upgrade.
1. Run prechecks and remediate2. Upgrade VCF Operations (first)3. Upgrade SDDC Manager to 9.14. Deploy VCF Management Services + license5. Upgrade NSX (management stack)6. Upgrade vCenter7. Upgrade ESX hosts, then finalizethe NSX Edge cluster
The mandatory 9.0.x to 9.1 sequence. Operations first, Edge last.

Where you can upgrade from

Not every environment takes the same road to 9.1. You can come from 9.0.x, which is the cleanest path and the one this series follows step by step. You can come from 5.2.x, which is a longer trip with more remediation and, in most cases, an intermediate stop before 9.1. And you can come from a standalone vSphere Foundation or a vSphere 8 with Aria Operations 8 footprint. Part 2 breaks down each supported path and the version floors that go with them. For this part, the thing to hold onto is that your starting point decides how much of the upgrade is remediation versus a straight version bump.

The prerequisites that actually bite

Prechecks in SDDC Manager catch most problems, but a handful of prerequisites are the ones I see stall real upgrades. Handle these before upgrade day, not while the clock is running.

A temporary IP for every vCenter

The vCenter upgrade needs a temporary IP address on the management network for each vCenter server while it migrates. One free address per vCenter, reachable and not already claimed by something else. On a tight subnet this is easy to forget until the workflow asks for it and you don’t have one ready.

Certificates and passwords that haven’t quietly expired

The upgrade validates certificate and password validity across components. An expired or about to expire certificate, or a password that rotated out from under SDDC Manager, will fail a precheck. Check these a week ahead so you have time to fix them calmly instead of during a window. If certificates are a recurring pain in your environment, that’s its own project worth doing first.

The vSAN hardware compatibility database

If you run vSAN, the hardware compatibility database needs to be current before the ESX upgrade. A stale HCL database can flag hosts that are actually fine, or miss ones that aren’t. Refresh it as part of prep so the host remediation step has accurate data to work from.

Bundles and the Operations version floor

Download the upgrade bundles ahead of the window so you’re not waiting on transfers mid-upgrade. And if you’re coming from 5.2.x, VCF Operations has to be on 8.18 before it can move to 9.1, which can mean an intermediate hop. From 9.0.x you don’t hit that floor, but you still want Operations fully current so the 9.1 bundles show up in the depot at all.

What VCF Management Services needs before you start

Two requirements catch people who skip the planning workbook.

Network space

VCF Management Services wants a block of contiguous IPs on the management network. The minimum works out to a /28, but Broadcom recommends a full /27, which is 30 usable addresses, so you have headroom. If your management subnet is already tight, sort this out before upgrade day, not during it.

A cloud proxy

The integration between VCF Operations, SDDC Manager and VCF Management Services needs a Cloud Proxy in the VCF instance where VCF Operations lives. If you don’t have one there, you’ll deploy it as part of the upgrade. Knowing that in advance saves a mid-upgrade scramble.

VCF Management Services (container cluster)Fleet Lifecyclefleet wide updatesSDDC Lifecycledomain updatesSoftware Depotbundle sourceVCF License Servercentral entitlement assignmentIdentity Brokermigrated from standalone vIDBMandatory. Must be deployed before NSX, vCenter and ESX upgrade to 9.1.
The pieces that live inside VCF Management Services in 9.1.
ComponentRequired for the 9.1 core upgrade?When
VCF OperationsYes, mandatory and firstStep 2
SDDC ManagerYesStep 3
VCF Management Services + License ServerYes, mandatoryStep 4
NSX, vCenter, ESXYesStep 5
VCF Automation (ex Aria Automation)No, only if deployedAfter core, Day-N
Workload domainsNo, optional Day-NLater windows

A worked example: planning a 9.0.2 management domain

Here’s how I’d frame a real one. Say you’re on 9.0.2, one management domain, NSX and vSAN in play, VCF Operations already deployed, no Aria Automation. The first thing I do is not open SDDC Manager. I open the VCF 9.1 Upgrade Planning Tool, tell it I’m on 9.0.2, and let it generate the phased plan plus the resource and networking requirements for that exact footprint.

Then I map maintenance windows to phases. Prechecks and the Operations upgrade in one window, allowing a couple of hours and expecting the fleet management transition to take time. SDDC Manager and the Management Services deployment in the next. NSX, vCenter and ESX across one or more windows depending on host count and how long you can drain clusters. The Edge finalize at the end. A single small management domain can land in a long day. A fleet with several workload domains is a multi-day plan, and that’s fine, because workload domains can be upgraded later as a Day-N job.

Planworkbook + toolPhase 1Ops + SDDC Mgr~2 to 3 hrsPhase 2Mgmt Services~1 hr deployPhase 3NSX + vCenter + ESX + Edgehours to days, by host countPhase 4
Rough phase shape for a small management domain. Get your own timings from a lab run.
Seen this go wrong: on one 9.0.2 fleet the 9.1 upgrade binaries simply didn’t appear in VCF Operations. No error, just an empty list. It was the classic wrong-path problem: Operations wasn’t on the level the 9.1 bundles expected, so the depot had nothing to show. Broadcom even has a KB for exactly this. The fix was boring. Get Operations current first, and the bundles showed up.
What I’d do: upgrade a lab or the smallest management domain end to end before you schedule the fleet. Not for the practice, for the timings. You want your own numbers for how long the Operations transition and the Management Services deploy take on your hardware before you promise a window to the business.
Signs it worked: SDDC Manager shows the management domain fully on 9.1, the old Fleet Management Appliance and standalone Identity Broker VMs are powered off, and VCF Management Services is up with the license server assigning entitlements. If any of those are still half in and half out, the upgrade isn’t done.

Common questions

Do I have to deploy VCF Management Services, or can I skip it?
You can’t skip it. It’s mandatory, and the rest of the 9.1 upgrade is gated on it being deployed. Plan the network space for it up front.

Can I upgrade SDDC Manager first to get it out of the way?
No. VCF Operations goes first because it carries the fleet management data that moves into VCF Management Services. SDDC Manager is the third step, after Operations and before the Management Services deploy.

What happens to my old Fleet Management Appliance and Identity Broker?
Both are retired during the upgrade. Their functions move into VCF Management Services, and the old VMs are powered down and ready to decommission. Don’t delete them until you’ve confirmed the new services are healthy.

How long does a 9.0.x to 9.1 upgrade take?
It depends on host count and how many workload domains you have. A single small management domain can finish in a day. Larger fleets are multi-day. Use the Upgrade Planning Tool to get phase estimates for your footprint, and upgrade workload domains later as Day-N work.

Where should I actually start?
With the Planning and Preparation Workbook and the VCF 9.1 Upgrade Planning Tool, before you download a single bundle. Parts 3 and 4 of this series go deep on both.

What the rest of this series covers

The plan for the next thirteen parts follows the upgrade itself. Parts 2 through 4 stay in planning: supported paths to 9.1, the Planning and Preparation Workbook, and the Upgrade Planning Tool. Parts 5 through 7 are prepare: prechecks and fixing what they flag, backups and a rollback plan before you touch anything, and the full sequence with why the order matters. Parts 8 through 12 are execute, one component at a time, from VCF Operations through the workload domains. Parts 13 and 14 land it, with post-upgrade validation and troubleshooting a stalled upgrade. If you’re upgrading a real fleet, read the planning parts before you schedule anything, and keep the general VCF 9 series handy for the component details this upgrade touches.

VCF 9.1 Upgrade · Part 1 of 14
Complete Guide  |  Next: Part 2

References

About The Author


Discover more from Journal of Intelligent Infrastructure – By Dr Pranay Jha

Subscribe to get the latest posts sent to your email.

Leave a Reply

Your email address will not be published. Required fields are marked *

Architect’s Toolkit

About the Author

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.

Discover more from Journal of Intelligent Infrastructure - By Dr Pranay Jha

Subscribe now to keep reading and get access to the full archive.

Continue reading