A client's NSX Global Manager and Local Manager stopped trusting each other's API calls a few weeks after a routine certificate rotation. No alert pointed at the actual cause. Global-to-Local authentication just quietly started failing. SDDC Manager had rotated the Local Manager's certificate right on schedule, exactly as it's supposed to. What SDDC Manager doesn't do is update the Global Manager's stored copy of that certificate's thumbprint — that has to be recaptured manually on the Global Manager every single time a Local Manager cert rotates through SDDC Manager, and it isn't documented anywhere obvious enough for most teams to know to check it.

That's what "what breaks" actually means during a VCF migration involving NSX. Not a failed wizard, not an error dialog. A component that keeps working right up until something routine — a scheduled cert rotation — exposes a dependency nobody wrote down.

What NSX Support Depends On Is the Patch Level, Not the Version

Whether NSX comes along automatically during import or convert depends on exactly which patch level the source environment is on, and the answer genuinely isn't consistent even within a single major version. VCF 9.0.0 supports NSX for import only. 9.0.1 and later, along with 9.1, support NSX for both import and convert. Checklists labeled simply "9.0.x" can give opposite guidance about whether NSX should even be registered before starting a convert operation, because they were captured at different patch levels without the label being updated to reflect it.

The fix isn't picking a checklist and trusting the label on it. It's confirming the exact source build before assuming either answer applies to the environment actually in front of you.

What Doesn't Come Along Automatically

What gets imported alongside the environment matters just as much as what's supported. Importing an existing vCenter 8.x environment into VCF does not bring network virtualization along with it — that piece has to be handled separately, as its own workstream. Import a vCenter 9.x environment instead, and network virtualization does come along as part of the same operation. Same import process, same-sounding source, a genuinely different outcome depending on which vCenter version is actually running underneath it.

Brownfield NSX deployment as part of a VLAN-to-NSX import also isn't fully automated as of VCF 9.0 — it's called out explicitly as a manual step, one that later releases are expected to close but haven't yet. Planning that step as automatic because most of the rest of the flow is automated is exactly the assumption that turns a scheduled cutover into an unscheduled one.

Prerequisites That Don't Change By Version

A few requirements stay fixed no matter which patch level the environment happens to be on. Standard vSwitch and Cisco Nexus 1000V are disallowed for import, full stop — DVS only, in every version reviewed. Any vCenter running in Enhanced Linked Mode has to have its SSO ring split into single-vCenter domains before import can proceed. ELM doesn't survive the transition as-is, and there's no supported path where it does.

NSX-Specific Items Worth Checking Before the Migration Plan Locks

  • Deploy NSX Edge nodes through SDDC Manager, not directly from NSX. An Edge deployed outside SDDC Manager still receives lifecycle patches, but its password can't be centrally managed afterward — a gap that gets more annoying the further it is from the original deployment.
  • Active/Active Tier-0 gateways can't provide stateful services like NAT. If ECMP and NAT or load balancing are both needed, a standalone Tier-1 gateway has to sit alongside the Active/Active Tier-0 — one gateway type can't do both jobs.
  • Global Manager certificate thumbprints need manual attention every time a Local Manager cert rotates through SDDC Manager, or Global-to-Local authentication fails without an obvious pointer back to the cause.
  • NSX Global Manager sits outside SDDC Manager's automated lifecycle and certificate management entirely — both have to be planned and executed as separate, manual tracks rather than assumed to be covered by the rest of the platform's automation.

Check First

Confirm the exact NSX and VCF patch level of the source environment before trusting generic "9.0.x" guidance about import versus convert support.

Confirm which vCenter version is actually underneath the environment being imported — 8.x and 9.x behave differently regarding what comes along with network virtualization.

Audit for Standard vSwitch, Nexus 1000V, or Enhanced Linked Mode configurations that will block import outright until remediated.

Build a manual process for NSX Global Manager certificate and lifecycle management before cutover — it will not inherit SDDC Manager's automation after the fact.

None of this makes NSX the wrong thing to bring into a VCF migration — for most environments running it already, there's no real alternative. It just means the plan has to account for a few things that don't automate the way the rest of the platform does, and finding that out during the migration instead of before it is the expensive way to learn it.

If NSX is part of an upcoming VCF migration, it's worth a direct conversation before the plan gets finalized around assumptions that only hold for a specific patch level.