FVMW - Fake VMware API for OpenShift MTV Demos

FVMW is a lightweight mock vCenter/ESXi API server that allows stock OpenShift MTV (Migration Toolkit for Virtualization) to perform end-to-end VM migration demos without a real VMware environment.

Built on govmomi/vcsim (Apache 2.0), it presents a fully functional vSphere SOAP API with configurable inventory, disk export, and VMX file serving. The default inventory matches the OpenShift Virtualization Roadshow demo so migration instructions work unchanged.

• Simulates a vCenter-compatible API on a single shared endpoint
• Full vSphere SOAP API with datacenter, cluster, host, datastore, network, folder, and VMs
• Serves flat VMDK disk images via the /folder/ endpoint for virt-v2v disk transfer
• Generates VMX files for libvirt domain XML parsing
• XML response rewriter fixes vcsim compatibility issues with libvirt's ESX driver
• Multiple users share the same VMDK images via a CephFS PVC
• Per-user VM names (e.g. database-user1) via FVMW_USER_SUFFIX env var

One shared endpoint serves all users. Each authenticated user gets their own 4 VM names in /RS00/vm/Roadshow/:

Infrastructure: Datacenter RS00, Cluster vcs-rs-00, Network segment-migrating-to-ocpvirt

# Build
make build

# Start locally (VPX or ESX mode)
./scripts/test-local.sh start        # VPX mode
./scripts/test-local.sh start esx    # ESX mode
./scripts/test-local.sh test         # Run govc tests
./scripts/test-local.sh stop

# Open read-only vCenter-like UI
# (login: user1/password, user2/password, etc.)
xdg-open "http://127.0.0.1:18443/ui/"

# 1. Configure local secrets
cp local.env.example local.env
vi local.env

# 2. One-time bootstrap (with cluster-admin)
oc login <cluster>
cd deploy/ansible
ansible-playbook bootstrap.yml -e @../../local.env

# 3. Set up build pipeline (uses SA kubeconfig from bootstrap)
ansible-playbook build.yml -e @../../local.env

# 4. Prepare disk images (one-time, see Disk Preparation below)

# 5. Deploy per-user instances
ansible-playbook deploy.yml -e @../../local.env

See deploy/ansible/README.md for full details.

For deploying fvmw itself (Deployment/Service/Route/PVC/Secret) to additional clusters, a Helm chart is available at charts/fvmw:

helm install fvmw charts/fvmw \
  --namespace fvmw --create-namespace \
  --set clusterDomain=apps.mycluster.example.com \
  --set fvmw.password=changeme

Disk preparation (steps 3-4 above) is still done separately via deploy/ansible/prepare-disks.yml, or by pointing disks.existingClaim at an already-populated PVC.

Shared Endpoint:

vcenter.<cluster-domain> Route (edge TLS)
                 |
            fvmw Pod
     +----------------------+
     | apiType: VirtualCenter
     | DC: RS00
     | Cluster: vcs-rs-00
     | VMs: *-user1, *-user2, ...
     +----------------------+

Shared CephFS PVC (/disks):
  database.vmdk, database-flat.vmdk
  winweb01.vmdk, winweb01-flat.vmdk
  winweb02.vmdk, winweb02-flat.vmdk
  haproxy.vmdk, haproxy-flat.vmdk

MTV Migration Flow:
  1. Provider -> vcenter route -> fvmw (inventory sync)
  2. virt-v2v -> vcenter route -> fvmw (VM discovery + VMX metadata)
  3. virt-v2v -> vcenter route -> fvmw (download flat VMDK via /folder/)
  4. virt-v2v converts disk -> creates KubeVirt VM

VMDK images must be on the shared PVC in monolithicFlat format. To prepare:

# 1. Export VMDKs from a real vCenter (streamOptimized format)
govc export.ovf -vm /DC/vm/database /tmp/export/

# 2. Upload to PVC via a helper pod
oc cp /tmp/export/database-disk-0.vmdk fvmw/helper-pod:/disks/database.vmdk

# 3. Convert to flat format inside the cluster
oc exec helper-pod -- qemu-img convert -O vmdk -o subformat=monolithicFlat \
  /disks/database.vmdk /disks/database-flat.vmdk

Both the original (streamOptimized) and flat VMDKs remain on the PVC. The -flat.vmdk files are served via the /folder/ endpoint for virt-v2v.

1. Create an MTV VMware Provider pointing to https://vcenter.<cluster>/sdk
2. MTV connects and syncs inventory via the vSphere SOAP API (handled by vcsim)
3. Provider shows 4 VMs in RS00/Roadshow/ with correct guest OS types
4. Create NetworkMapping (pod networking) + StorageMapping (ceph-rbd)
5. Create + execute Migration Plan (select database-{user}, winweb01-{user}, winweb02-{user})
6. virt-v2v connects to the shared endpoint and downloads VMX + flat VMDK files
7. virt-v2v converts the disk and creates a KubeVirt VM in OpenShift Virtualization

Apache 2.0 (govmomi/vcsim dependency is also Apache 2.0)