I ran ESXi at home for years and was perfectly happy with it. The free hypervisor did everything I needed: a few always-on VMs, some throwaway ones, a single host I rarely thought about. It was boring in the best way. So this isn't a post about ESXi being bad. It's a post about the writing on the wall getting too large to ignore, and about the surprise that the alternative turned out to be better for what I do rather than merely adequate.
The trigger was the obvious one. Broadcom finished acquiring VMware, the free ESXi licence got noticeably less certain, and the general direction of travel was towards everything being a subscription bundle aimed at people with a procurement department. None of that is aimed at me. I'm one bloke with three Mini PCs and a NAS. When the platform you depend on starts treating your entire use case as a rounding error, it's a good moment to look around before you're forced to.
what i was actually running
Three nodes, nothing exotic. Two small Intel boxes and an older tower doing the heavy lifting, a Synology for bulk storage over NFS, and around a dozen VMs across the lot. A pfSense VM, a couple of Debian boxes running containers, a Windows VM I keep purely so I can remember why I don't run Windows, and the usual sprawl of things I spun up "just to test" eighteen months ago and never killed.
Under ESXi these were three standalone hosts. I never paid for vCenter, so there was no live migration, no shared cluster view, nothing clever. Each host was its own island and I managed them through three separate web UIs. It worked, but it was three of everything.
the migration itself
The honest version: it was less painful than I'd talked myself into expecting, but it was not a single afternoon.
I did it one node at a time so I never lost the whole lab. Evacuate a host, rebuild it on Proxmox VE 8, move workloads onto it, repeat. The VMs themselves came across reasonably cleanly. Proxmox has qm importovf and qm importdisk, and for most Linux guests the process was: export from ESXi as OVF, copy the disk over, import it, fix the boot disk type, attach a network device, boot it.
# import a disk straight onto a Proxmox storage pool
qm importdisk 110 /var/lib/vz/import/debian-app.vmdk local-lvm
# then attach it as the boot disk on VM 110
qm set 110 --scsihw virtio-scsi-pci --scsi0 local-lvm:vm-110-disk-0
qm set 110 --boot order=scsi0
The gotchas were the ones you'd guess. VMware Tools had to come off and qemu-guest-agent had to go on. A couple of guests came up with their network interface renamed, because the underlying virtual NIC changed from vmxnet3 to virtio, so anything pinning a config to ens192 needed a nudge. The Windows VM needed the VirtIO driver ISO mounted before it would see its own disk, which is a rite of passage rather than a surprise.
pfSense was the one I planned around carefully, because it's the router and getting it wrong means doing the rest of the work with no internet and a lot of swearing. I rebuilt it fresh on the new host from a config backup rather than importing the VM, which was cleaner anyway. Twenty minutes of downtime, scheduled for when nobody in the house would notice, and done.
what's better
Clustering, for free, is the big one. I joined all three nodes into a single Proxmox cluster and now I have one web UI that shows everything. That alone justified the move. Migrating a VM from one node to another is a right-click, and because my storage is shared over NFS, plenty of it is a live migration with no downtime. I never had that under free ESXi and I hadn't realised how much I wanted it until I had it.
The other quiet win is that it's just Debian underneath. When something misbehaves I can SSH in and use tools I already know. ZFS support is built in and sane. LXC containers sit right next to the full VMs, so the lightweight things that didn't really need a whole VM now don't have one. Backups go to the NAS through the built-in scheduler, and Proxmox Backup Server is there if I want deduplicated incrementals, which I now do.
what i miss
I'll be fair about this, because the fanboy version of these posts is tiresome.
The ESXi web client was more polished. Proxmox's UI is functional and dense and occasionally makes you feel like you're operating machinery rather than using an app, and that's a fair trade for me but it is a real difference. vMotion across hosts with different CPUs was smoother in the VMware world than fiddling with KVM CPU type flags to get live migration happy across my two slightly different Intel generations. And there's a maturity to vSphere's handling of edge cases that comes from a couple of decades and a lot of paying customers finding the sharp corners first.
None of that outweighs "the free tier exists, is generously featured, and isn't about to be taken away from me." For a homelab, that's the whole game.
would i recommend it
If you're a home user on free ESXi and feeling nervous, yes, do it now while it's calm rather than later in a hurry. Budget a weekend, do it node by node, keep the router migration for last and rehearse it. The conversion is mechanical and well-documented, the destination is genuinely good, and you come out the other side owning your platform rather than renting it on someone else's sufferance. I should have done it a year earlier.