1. Download & deploy
Disklift is a ready-to-run Debian appliance - about 7 GB. Download the image for the hypervisor you will run it on and deploy it there. It starts in free evaluation mode with no license required, so you can run a real migration before buying anything.
- VMware: download vmware.tar.zst, extract it, and deploy the OVF inside it to vCenter or ESXi (Deploy OVF template)
- Proxmox: download the qcow2 image and import it as a VM disk
- Start the VM and sign in on its console as
migrationwith passwordmigration(you can change the console password from the menu) - The first login walks you through the installation: network settings and, if the appliance sits behind a firewall, a proxy
- The appliance needs a proxy or direct internet access (HTTPS, port 443) - it downloads the latest files during setup
- When setup finishes the web UI is at
https://<its-ip>:8443
Evaluation limits (no license)
A new installation runs in evaluation mode for 30 days. Each limit ends the evaluation of that kind of work when it is reached - whichever comes first.
| Migration | Replication / DR | |
|---|---|---|
| VMs | up to 5 | up to 3 |
| First full copies | up to 750 GB in total | up to 750 GB in total |
| Changed data after the first copy | - | up to 500 GB in total |
Only data that is really transferred counts. Fail-over and cut-over are never blocked by these limits, so a test can always be finished. The usage is shown on the License screen. A purchased license removes the limits - see pricing.
2. Connect your hypervisors
Add each VMware vCenter or standalone ESXi host, Hyper-V host, OpenStack endpoint, Nutanix AHV cluster, Linux KVM host, or Proxmox node from Application settings. Credentials are encrypted at rest and only ever decrypted in memory for the duration of a job.
- VMware (vCenter or ESXi): the vSphere API and SSH on the ESXi hosts that hold the VMs - to Proxmox or another VMware host
- Proxmox: the API (token or user/password) and SSH on every node - as a destination, or as a source for Proxmox → Proxmox moves
- Hyper-V: either WinRM over HTTPS or Windows OpenSSH (your choice per host) - migration and DR replication to Proxmox with change tracking, and Hyper-V-to-Hyper-V replication through Hyper-V's own native Replica
- OpenStack: the API (Keystone) and SSH to the storage host for Ceph, LVM or NFS backends - to Proxmox or another OpenStack endpoint
- Nutanix AHV (beta): the Prism Element API on port 9440 - backup, restore, and migration and DR to Proxmox. Add Prism Central (pc.2024.3 or later) for change tracking
- Linux KVM (libvirt): SSH only (
virsh,perl,qemu-img, andqemu-nbdfor qcow2 disks) - migration and DR to Proxmox or another KVM host
3. Build a policy
A policy groups VMs with a schedule, bandwidth limit, and retry behavior, and decides what kind of job they run: a one-time migration, or ongoing DR replication.
- Pick source VMs and a destination node/storage
- Choose migration (one-time cutover) or DR (continuous sync, with failover/failback later)
- Changed-block tracking and thin copy apply automatically; wire compression and deduplication (first copy to Proxmox) are optional settings per policy, both off by default
- Watch progress live on the performance dashboard, including thin-copy, compression and deduplication savings and throughput
4. License it
Buy migration credits or DR seats on the Pricing page, then activate the license one of two ways.
- From inside the app: License screen → Activate online - your installation ID is sent automatically
- From here: use Activate a license and paste in your installation ID by hand
- Your license file is also emailed to the address used at checkout, as a backup
- A migration credit is only spent the first time a VM's data actually moves - never charged twice
- A used credit is not returned if the VM or its policy is deleted
- Unused credits can be used until the license's end date (12 months from purchase for packs bought here); a VM that already used a credit can always be finished
Network ports & access
Everything Disklift does goes over standard, encrypted management channels - nothing is installed inside your VMs. Open these from the Disklift appliance to your hypervisors. The data itself is read and written through the appliance, so your source and destination hosts do not need to reach each other.
| Platform | What it uses | Ports (from the appliance) | What you need to provide |
|---|---|---|---|
| The appliance itself | Web UI (HTTPS) | TCP 8443 from your browsers to the appliance | Nothing to install - open the port on any firewall between you and the appliance. |
| Internet (appliance) | Setup downloads, online license activation | TCP 443 outbound, or through your HTTP proxy | Direct internet access, or the proxy address entered during first-login setup. |
| VMware vCenter / ESXi | vSphere API + SSH | TCP 443 to vCenter / ESXi (API) TCP 22 to each ESXi host (SSH / SFTP) | A vSphere account with rights on the VMs and datastores, and the SSH service enabled on the ESXi hosts (data is read with dd over SSH, and written over SFTP when VMware is the destination). |
| Proxmox VE | Proxmox API + SSH | TCP 8006 (API) TCP 22 to every node (SSH) | An API token or user/password with the needed privileges, and SSH as root (or a user with password-less sudo) on each node. Data is written over SSH. |
| Hyper-V | WinRM over HTTPS or Windows OpenSSH - you choose one per host | TCP 5986 (WinRM over HTTPS) or TCP 22 (Windows OpenSSH) | A dedicated account in Hyper-V Administrators (or a local administrator). Disk data is read through the same connection - no separate file share is needed. Windows OpenSSH must be enabled on the host (the appliance's Documentation has the steps). |
| Hyper-V → Hyper-V (Replica) | Hyper-V's own Replica traffic between your hosts | TCP 80 (Kerberos) or TCP 443 (certificate) between the two Hyper-V hosts, plus the management port above | Replica enabled on the replica server. Hyper-V moves the data itself. |
| Nutanix AHV Beta | Prism Element API (and optionally Prism Central for change tracking) | TCP 9440 to the cluster virtual IP TCP 9440 to Prism Central, if used | A Prism account that can list VMs, take snapshots, and create images and VMs. With Prism Central the same account must be able to create and delete recovery points. |
| Linux KVM (libvirt) | SSH only - virsh -c qemu:///system on the host | TCP 22 to each KVM host | SSH as root, or a user with password-less sudo, on a host with libvirtd, perl and qemu-img; qemu-nbd and the nbd kernel module for qcow2 disks. Nothing is installed. |
| OpenStack | OpenStack APIs, plus SSH to the storage host | The Keystone identity URL and the Nova / Cinder / Glance / Neutron endpoints published in your service catalog (the ports your cloud uses - often 5000 for identity, or 443 behind a load balancer) TCP 22 to the storage host | A project credential (an application credential is recommended). For Ceph, LVM or NFS backends: SSH to a host with the Ceph client (rbd), the cinder-volume host, or a host that mounts the Cinder shares - as root or with password-less sudo. With API only, disks are copied through the API. |
SSH host keys and TLS certificates of every hypervisor are pinned the first time you connect; a change blocks the connection until an administrator re-verifies it. If a password is rejected three times, the appliance stops trying rather than lock the account out. The app's Documentation (user menu) lists the exact privileges per platform.
A separate network for replication and backups
The appliance ships with one network adapter (ens192 on the VMware image, ens18 on the Proxmox image). Everything it copies would
otherwise travel over that adapter - the same network your hypervisors are managed on - so a large migration or backup can slow management traffic down.
Give the appliance extra adapters for the data and keep the first one for the web page and the hypervisor APIs.
Good: one extra adapter
A second adapter on a storage or replication VLAN. Enough for most sites, and the simplest to set up.
Better: two extra adapters, bonded
Bond them with LACP (802.3ad) for more bandwidth and a link that survives a failed port or cable - the switch ports must be in an LACP port-channel. No LACP on the switch? Use active-backup.
Separate backups too
A third adapter, or a second VLAN on the same bond, keeps traffic to your NFS or S3 repositories apart from replication.
- Add the adapters to the appliance VM and connect them to the data switch (as trunk ports if you use VLAN tags).
- In the appliance open Application settings → System → Network. It lists every adapter with its link, negotiated speed, duplex, traffic and errors. The management adapter is shown but cannot be changed.
- Add a network: choose Data or Backup, tick one adapter or several (LACP or active-backup), and give the VLAN tag, address, MTU (9000 for jumbo frames end to end) and any routes. Up to four networks are supported. Apply - if an address does not come up, the change is rolled back by itself.
- For each hypervisor, Configuration → right-click the host → Data network… and enter the host's (and each node's) address on the data network. Disk data then goes there; API calls stay on management. The data address must present the same SSH host key as the management address.
- For each NFS or S3 repository, choose its backup network under Protection → Repositories → Edit.
- Use Test a path on the Network page to see which adapter reaches an address and whether jumbo frames get through. Forced speed and duplex can be set per adapter if a switch port needs it.
IPv4 in this version. Open the same ports as in the table above on the data network for the hypervisors you reach there. Restores, backups and Nutanix reads stream straight through the appliance - nothing is staged on its disk.
Recommended appliance sizing
The appliance reads from the source and writes to the destination, so its CPU, memory and network set how many VMs it can move at once. These are starting points, not measured limits: deploy the tier that matches your estate, then watch the Queue wait tile on the Performance page and move up a tier (or add workers) if waits keep growing.
| Estate | VMs protected | Concurrent jobs | vCPU | Memory | Disk | Network |
|---|---|---|---|---|---|---|
| Small | up to ~25 | up to 4 | 4 | 8 GB | 32 GB (as shipped) | 1 GbE |
| Medium | ~25 – 150 | up to 16 | 8 | 16 GB | 64 GB | 1 – 10 GbE |
| Large | ~150 – 600 | up to 32 | 16 | 32 GB | 100 GB | 10 GbE |
| Enterprise | 600+ | 32+, several workers | 32 or more | 64 GB or more | 200 GB or more | 10 GbE or faster |
On VMware
- Disk: at least 32 GB, thin provisioned, on SSD-backed storage where you can
- Network adapter: VMXNET3, on a port group with a direct path to your hosts
- Guest OS: Debian Linux (64-bit); install VMware Tools / open-vm-tools
- Reserve the memory and give the VM its vCPUs without overcommit on a busy host
On Proxmox
- Disk: at least 32 GB on SSD or NVMe-backed storage, with discard enabled so the disk stays thin
- Network adapter: VirtIO (paravirtualised) on a bridge with a direct path to your hosts
- Disk controller: VirtIO SCSI single; CPU type host; QEMU guest agent on
- Do not overcommit the appliance's CPU or memory against other VMs on a busy node
- Network is usually the limit. One gigabit link moves roughly 110 MB/s at best; a Large or Enterprise estate should have 10 GbE between the appliance and the hypervisors - ideally on its own data network, or two 10 GbE adapters in an LACP bond.
- Compression and deduplication use extra CPU on the appliance. If you turn them on for a busy appliance, size one tier up.
- Enterprise estates: run several workers (
pmxctl workers N, or Application tools → Worker containers in the appliance console) and consider an external, highly available PostgreSQL. Allow roughly 2 vCPU and 2 GB of memory for every 8 concurrent jobs, on top of about 2 vCPU and 4 GB for the web service and database. - The disk holds the database, job history and logs. Old history is cleaned up automatically, and a warning appears before the disk fills - but grow it if you keep many months of activity or run very many VMs.
Not sure which tier fits? Tell us your VM count and source hosts and we will suggest one.
Frequently asked
Do I need a license to try it?
No. Disklift runs in free evaluation mode out of the box - connect real hypervisors and run a real migration with no license at all, for 30 days, up to 5 VMs / 750 GB of first copies for migration and up to 3 VMs / 750 GB of first copies plus 500 GB of changed data for replication (details). You only need a license to remove those limits for production use.
What's the difference between a Migration and a DR license?
Migration is a credit pack: each VM spends one credit the first time it's moved, one-time. A used credit is not returned if the VM or its policy is deleted. Unused credits are valid until the license's end date - 12 months from purchase for packs bought here, or the shorter term agreed for your order. DR is a per-seat annual subscription: one seat covers one VM kept under continuous replication for the term, with failover, test failover and failback included.
Can one install run both migration and DR at once?
Yes - it's the same install either way. Some VMs can run a one-time migration policy while others run an ongoing DR replication policy, side by side, licensed independently.
What do I download, and how do I install it?
A ready-to-run Debian appliance of about 7 GB - a compressed archive containing the OVF for VMware (extract it first), or a qcow2 image for Proxmox. Deploy it to
your hypervisor, open its console, and sign in as migration / migration. The first login walks you
through the installation. If the appliance is behind a firewall it needs a proxy, otherwise direct internet access,
because it downloads the latest files during setup.
Does the source VM have to be powered off?
No. Online (the recommended copy method) keeps the VM running: for VMware a snapshot freezes the disks, the first full copy and then small incremental syncs read from it, and the VM is only shut down - gracefully, if you choose - at the final cut-over, so downtime is limited to that window. Offline (migration policies) shuts the VM down before the first copy and leaves it off: the cleanest copy, with downtime from the start. Either way the original is kept after cut-over until you decide to delete it.
For extra caution a policy can also offer an optional safety clone of a VMware VM: a thin, powered-off, network-disconnected copy with new MAC addresses, made before anything is shut down or changed. It is off by default, a running VM is copied crash-consistently with no downtime, and the migration does not start if the clone fails.
How do I extract the VMware download (vmware.tar.zst)?
The VMware appliance is a Zstandard-compressed tar archive (.tar.zst) - it keeps the download small. Extract it once, then
deploy the OVF files inside. Make sure you have enough free disk space for the extracted files.
Windows: on Windows 11 you can right-click the file and choose Extract All (File Explorer reads .tar.zst
natively on recent builds). On older Windows, install 7-Zip 24.05 or newer, right-click
the file and choose 7-Zip → Extract Here - that produces a .tar file, which you extract the same way a second time.
Linux: install zstd (for example sudo apt install zstd), then run:
zstd -dc vmware.tar.zst | tar -xvf -
macOS: brew install zstd, then the same command as Linux.
Then deploy it: in the vSphere Client choose Deploy OVF Template → Local file and select the .ovf file together
with its .vmdk (and .mf) files; on a standalone ESXi host use Create / Register VM → Deploy a virtual machine from an
OVF or OVA file. Power it on, open its console and sign in as migration / migration.
Which ports and access does it need?
The web UI is on HTTPS port 8443. To your platforms it needs: VMware – 443 (API) and 22 (SSH on the ESXi hosts); Proxmox – 8006 (API) and 22 (SSH on every node); Hyper-V – 5986 (WinRM over HTTPS) or 22 (Windows OpenSSH); OpenStack – the API endpoints from your service catalog, plus 22 (SSH) to the storage host for Ceph, LVM or NFS; Nutanix AHV – 9440 (Prism); Linux KVM – 22 (SSH). See the ports table above for the full list.
Do I really need SSH?
Yes for VMware (SSH must be enabled on the ESXi hosts), Proxmox (SSH on every node), OpenStack storage backends (Ceph, LVM, NFS) and Linux KVM hosts (SSH is the only way in). That is how the disk data is read and written. Hyper-V can use either WinRM over HTTPS or Windows OpenSSH. Nutanix AHV needs no SSH: its disks move through the Prism image service.
Where's the full operational documentation?
Inside the app itself, under Documentation in the user menu - that's where the detailed per-hypervisor setup, DR runbooks, and troubleshooting guidance live, versioned alongside the build you're running.
What if I need help?
Email [email protected] or call 571-758-2754. See the Contact page for more.