How to Upgrade to VMware Cloud Foundation 9.0

Upgrading to VMware Cloud Foundation has been a persistent source of confusion, and VCF 9.0 changed the model rather than just the version number. It is a unified platform spanning traditional, modern, and AI workloads, with consistent operations and governance across the private cloud — and getting there requires more planning than a routine point upgrade.

Establish where you are actually starting from

The upgrade path depends on the exact combination of components deployed, not just the VCF version on the box. Before planning anything, inventory the current version of every component in the estate, the management domain topology, which Aria components are in play, and any deployment types with known path restrictions — stretched clusters, two-node, and satellite node deployments have historically waited on later releases for supported paths.

Understand what consolidated in 9.x

The most significant structural change is the move of management functions into a unified layer. Aria components transition into VCF Operations, and a Management Services layer now hosts services including the license server and software depot. This is an operational model change, and the runbooks your team wrote against 5.x will need revision before the first maintenance window, not after it.

Sequence and prechecks

Component upgrade order is not flexible, and the prechecks exist because skipping them is expensive. Plan for:

1
Inventory the estateEvery component version, the management domain topology, and any deployment type with a restricted path.
2
Back up everythingSDDC Manager file-based backups, full VM backups of management appliances, cold snapshots where the window allows.
3
Clear the prechecksResource headroom, certificates, identity sources, and hardware and firmware compatibility across the full stack.
4
Upgrade management firstSDDC Manager and the management domain lead. Order is not flexible.
5
Then workload domainsOne at a time, with validation between each, so a failure has a known blast radius.
6
Revalidate operationsBackups, monitoring, and automation all touch the new Management Services layer. Confirm each before closing the window.

The sequence most teams get wrong in one place: treating prechecks as a formality rather than the gate they are.

Backups first, without exception. File-based backups configured in SDDC Manager, full VM backups of the management appliances, and cold snapshots where the maintenance window allows.
Resource headroom. The 9.x management components carry different footprints. Verify capacity before the upgrade window, not during it.
Certificate and identity dependencies. These surface late and stall an otherwise clean upgrade.
Hardware and firmware compatibility. Validate the full stack against the compatibility guide, including NIC and storage firmware.

Know what came after

VCF 9.1 has since arrived with more flexible upgrade paths, meeting environments closer to where they already are — standalone vSphere with VCF Operations, various vSAN and NSX combinations, or a full-stack VCF estate. If you are still planning the move off 5.2.x, it is worth evaluating both destinations before committing, and running the published upgrade planning tool to turn a general guide into an ordered plan for your specific environment.

Then decide whether the destination is still right

Licensing has changed materially since the Broadcom acquisition, and the honest version of this conversation includes the alternatives. For some workloads VCF remains the strongest platform available and the upgrade is straightforward to justify. For others — particularly smaller estates or workloads with modest feature requirements — the licensing math now favors evaluating Proxmox, Hyper-V, or another hypervisor.

We run the migration analysis before the upgrade plan, not after. Upgrading into a platform you are going to leave in eighteen months is the most expensive outcome available.

Upgrade makes sense whenYou run at scale, depend on NSX or vSAN capabilities, have staff depth in the platform, and the licensing math still clears against the alternatives.

Migration deserves a look whenThe estate is modest, feature requirements are basic, renewal pricing moved sharply, or you are already carrying a second hypervisor.

Run this comparison before the upgrade plan, not after it.

How we help

South Bay Coders assesses the current estate, models both the upgrade path and the migration alternatives against real workload and licensing data, and delivers a sequenced plan your team can execute. Where a migration off VMware makes better sense, we plan and run that too — including the storage architecture underneath it.

Sources & further reading

← All articles