Choosing an HCX migration type for VCF 9.0

Each migration type trades downtime, parallelism and prerequisites differently, so a migration plan normally uses several of them.

• Bulk migration runs a full initial synchronisation, then delta synchronisation with a two-hour recovery point until the switchover, which can be scheduled into a maintenance window. At switchover the source powers off, a final sync runs and the replica powers on; the original is renamed and kept in a Migrated VMs folder for recovery. Guest customisation (hostname, IP, SID) adds a reboot. Encrypted VMs migrate only if the same key provider exists at the target: a KMS with matching name and IP, or a Native Key Provider exported and imported out of band. RDMs in virtual compatibility mode become VMDKs; physical-mode RDMs cannot move.
• HCX vMotion carries memory, execution state, IP and MAC; duration depends on bandwidth and latency. Only serial vMotion is supported between a site pair; use Bulk or RAV for parallel moves.
• Replication Assisted vMotion (RAV) replicates many VMs at once, keeps them in sync until a switchover window, then performs serial live switchovers. It tolerates latency and varying conditions during sync, does not support vVol datastores or independent disks, and leaves two folders per VM at the destination, which some backup tools dislike (Storage vMotion consolidates them).
• HCX Assisted vMotion (Direct) orchestrates native cross-vCenter vMotion over a direct host-to-host path; the IX carries only control traffic. HCX diagnostics are not available, and several NSX segment and VPC source combinations are unsupported.
• OS Assisted Migration (OSAM) needs Sentinel software in each guest. Multi-site mode deploys SGW at the source and SDR at the destination; single-site mode deploys SRG at the vSphere site. Each SGW or SRG supports 50 active replica disks, and four service meshes reach 200 concurrent disk migrations. Sentinel appliances run only on vSphere, the target datastore must be visible to the SDR or SRG, migrated VMs get the default storage policy, and the target reboots twice during switchover.
• Supervisor onboarding imports traditional VMs into a Supervisor namespace (see Chapter 14: Designing consumption: VCF Automation, self-service and modern applications). HCX 9.0 must run at both ends; IPv6-dependent workloads are excluded, and VMs that fail import wait in the staging area. For VMs already in the same vCenter, the vSphere Client can also import a VM into a namespace so that VM Service manages it.

Some restrictions apply across types: vCenter, NSX Manager and other VMware appliances cannot be migrated with HCX; virtualisation-based security is not supported; VMs using shared SCSI buses, multi-writer disks, Fault Tolerance or shared VMDKs cannot move; snapshots are not migrated; vSphere VM Encryption is not supported for HCX vMotion and RAV.

Common trap: Choosing HCX vMotion for a large wave and expecting parallel live moves - only serial vMotion is supported between sites; RAV delivers parallel replication with live switchover.

Common trap: Treating Bulk migration as zero downtime - its switchover is equivalent to a reboot, plus another reboot if the guest is customised.

Get the whole book

This note is one section of Ultra Transcenders 2V0-13.25: VMware Cloud Foundation 9.0 Architect, an independent study guide that explains every topic the exam covers by technology, with comparison tables, diagrams and the common traps, plus a glossary linked to Broadcom TechDocs.

About the book · 2V0-13.25 terms in the glossary · All 2V0-13.25 study notes