Move VMware workloads to Red Hat OpenShift Virtualization: see the target architecture, size the bare-metal cluster, build the platform, then migrate one Linux and one Windows VM with the Migration Toolkit for Virtualization. No real systems are touched.
Incident
Bare-metal nodes with Intel VT-x or AMD-V enabled in BIOS. Each VM runs as a QEMU/KVM process inside a virt-launcher pod, scheduled like any other pod.
A Linux bridge on a dedicated bonded trunk lets VMs attach straight to existing VLANs, so IPs, firewalls and DNS stay unchanged. Live migration gets its own VLAN.
VM disks are PVCs. Use block-mode, ReadWriteMany storage so VMs can live-migrate between nodes. ODF on local NVMe, or an existing SAN whose CSI driver supports RWX block.
MTV reads inventory from vCenter, copies disks with VMware's VDDK, and runs virt-v2v to swap VMware drivers for VirtIO. Warm migration copies while the VM runs; cold copies after shutdown.
Platform overhead to budget (Red Hat published figures): about 4 cores across infrastructure nodes and 2 cores per worker for OpenShift Virtualization; about 360 MiB per worker; per VM ≈ 0.002 × VM memory + 218 MiB + 8 MiB per vCPU; about 10 GiB local storage per node.
ODF on the same nodes needs about 10 cores and 24 GiB per storage node. With a SAN instead, drop that line and use the vendor's CSI driver (it must support RWX block for live migration).
Network: 2 × 25 GbE for management and pods, 2 × 25 GbE trunk for VM VLANs and live migration, plus BMC access for install.
Assumes memory is not overcommitted and about 85% of each node is usable after system reserves. Confirm against real VMware usage data (RVTools export or vROps) before ordering hardware.