Input and Output¶

You normally run virt-v2v with several -i* options controlling the input mode and also several -o* options controlling the output mode. In this sense, "input" refers to the source foreign hypervisor such as VMware, and "output" refers to the target KVM-based management system such as oVirt or OpenStack.

The input and output sides of virt-v2v are separate and unrelated. Virt-v2v can read from any input and write to any output. Therefore these sides of virt-v2v are documented separately in this manual.

Virt-v2v normally copies from the input to the output, called "copying mode". In this case the source guest is always left unchanged. In-place conversion (--in-place) only uses the -i* options and modifies the source guest in-place. (See "In-place conversion" below.)

Other virt-v2v topics¶

virt-v2v-support(1) — Supported hypervisors, virtualization management systems, guests.

virt-v2v-input-vmware(1) — Input from VMware.

virt-v2v-input-xen(1) — Input from Xen.

virt-v2v-output-local(1) — Output to local files or local libvirt.

virt-v2v-output-rhv(1) — Output to oVirt or RHV.

virt-v2v-output-openstack(1) — Output to OpenStack.

virt-v2v-copy-to-local(1) — Deprecated tool to handle Xen guests using host block device storage.

Convert from VMware vCenter server to local libvirt¶

You have a VMware vCenter server called "vcenter.example.com", a datacenter called "Datacenter", and an ESXi hypervisor called "esxi". You want to convert a guest called "vmware_guest" to run locally under libvirt.

 virt-v2v -ic vpx://vcenter.example.com/Datacenter/esxi vmware_guest

In this case you will most likely have to run virt-v2v as "root", since it needs to talk to the system libvirt daemon and copy the guest disks to /var/lib/libvirt/images.

For more information see virt-v2v-input-vmware(1).

Convert from VMware to RHV/oVirt¶

This is the same as the previous example, except you want to send the guest to a RHV Data Domain using the RHV REST API. Guest network interface(s) are connected to the target network called "ovirtmgmt".

 virt-v2v -ic vpx://vcenter.example.com/Datacenter/esxi vmware_guest \
   -o rhv-upload -oc https://ovirt-engine.example.com/ovirt-engine/api \
   -os ovirt-data -op /tmp/ovirt-admin-password -of raw \
   -oo rhv-cafile=/tmp/ca.pem -oo rhv-direct \
   --bridge ovirtmgmt

In this case the host running virt-v2v acts as a conversion server.

For more information see virt-v2v-output-rhv(1).

Convert from ESXi hypervisor over SSH to local libvirt¶

You have an ESXi hypervisor called "esxi.example.com" with SSH access enabled. You want to convert from VMFS storage on that server to a local file.

 virt-v2v \
   -i vmx -it ssh \
   "ssh://root@esxi.example.com/vmfs/volumes/datastore1/guest/guest.vmx" \
   -o local -os /var/tmp

The guest must not be running. Virt-v2v would not need to be run as root in this case.

For more information about converting from VMX files see virt-v2v-input-vmware(1).

Convert disk image to OpenStack¶

Given a disk image from another hypervisor that you want to convert to run on OpenStack (only KVM-based OpenStack is supported), you can run virt-v2v inside an OpenStack VM (called "v2v-vm" below), and do:

 virt-v2v -i disk disk.img -o openstack -oo server-id=v2v-vm

See virt-v2v-output-openstack(1).

Convert disk image to disk image¶

Given a disk image from another hypervisor that you want to convert to run on KVM, you have two options. The simplest way is to try:

 virt-v2v -i disk disk.img -o local -os /var/tmp

where virt-v2v guesses everything about the input disk.img and (in this case) writes the converted result to /var/tmp.

A more complex method is to write some libvirt XML describing the input guest (if you can get the source hypervisor to provide you with libvirt XML, then so much the better). You can then do:

 virt-v2v -i libvirtxml guest-domain.xml -o local -os /var/tmp

Since guest-domain.xml contains the path(s) to the guest disk image(s) you do not need to specify the name of the disk image on the command line.

To convert a local disk image and immediately boot it in local qemu, do:

 virt-v2v -i disk disk.img -o qemu -os /var/tmp --qemu-boot

Xen paravirtualized guests¶

Older versions of virt-v2v could turn a Xen paravirtualized (PV) guest into a KVM guest by installing a new kernel. This version of virt-v2v does not attempt to install any new kernels. Instead it will give you an error if there are only Xen PV kernels available.

Therefore before conversion you should check that a regular kernel is installed. For some older Linux distributions, this means installing a kernel from the table below:

 RHEL 3         (Does not apply, as there was no Xen PV kernel)
 
 RHEL 4         i686 with > 10GB of RAM: install 'kernel-hugemem'
                i686 SMP: install 'kernel-smp'
                other i686: install 'kernel'
                x86-64 SMP with > 8 CPUs: install 'kernel-largesmp'
                x86-64 SMP: install 'kernel-smp'
                other x86-64: install 'kernel'
 
 RHEL 5         i686: install 'kernel-PAE'
                x86-64: install 'kernel'
 
 SLES 10        i586 with > 10GB of RAM: install 'kernel-bigsmp'
                i586 SMP: install 'kernel-smp'
                other i586: install 'kernel-default'
                x86-64 SMP: install 'kernel-smp'
                other x86-64: install 'kernel-default'
 
 SLES 11+       i586: install 'kernel-pae'
                x86-64: install 'kernel-default'
 Windows        (Does not apply, as there is no Xen PV Windows kernel)

Enabling virtio¶

"Virtio" is the name for a set of drivers which make disk (block device), network and other guest operations work much faster on KVM.

Older versions of virt-v2v could install these drivers for certain Linux guests. This version of virt-v2v does not attempt to install new Linux kernels or drivers, but will warn you if they are not installed already.

In order to enable virtio, and hence improve performance of the guest after conversion, you should ensure that the minimum versions of packages are installed before conversion, by consulting the table below.

 RHEL 3         No virtio drivers are available
 
 RHEL 4         kernel >= 2.5.9-89.EL
                lvm2 >= 2.02.42-5.el4
                device-mapper >= 1.02.28-2.el4
                selinux-policy-targeted >= 1.17.30-2.152.el4
                policycoreutils >= 1.18.1-4.13
 
 RHEL 5         kernel >= 2.6.18-128.el5
                lvm2 >= 2.02.40-6.el5
                selinux-policy-targeted >= 2.4.6-203.el5
 
 RHEL 6+        All versions support virtio
 
 Fedora         All versions support virtio
 
 SLES 11+       All versions support virtio
 
 SLES 10        kernel >= 2.6.16.60-0.85.1
 
 OpenSUSE 11+   All versions support virtio
 
 OpenSUSE 10    kernel >= 2.6.25.5-1.1
 Debian 6+      All versions support virtio
 Ubuntu 10.04+  All versions support virtio
 Windows        Drivers are installed from the ISO or directory pointed
                to by "VIRTIO_WIN" environment variable if present

RHEL 4: SELinux relabel appears to hang forever¶

In RHEL ≤ 4.7 there was a bug which causes SELinux relabelling to appear to hang forever at:

 *** Warning -- SELinux relabel is required. ***
 *** Disabling security enforcement.         ***
 *** Relabeling could take a very long time, ***
 *** depending on file system size.          ***

In reality it is waiting for you to press a key (but there is no visual indication of this). You can either hit the "[Return]" key, at which point the guest will finish relabelling and reboot, or you can install policycoreutils ≥ 1.18.1-4.13 before starting the v2v conversion. See also https://bugzilla.redhat.com/show_bug.cgi?id=244636

Debian および Ubuntu¶

"warning: could not determine a way to update the configuration of Grub2"

Currently, virt-v2v has no way to set the default kernel in Debian and Ubuntu guests using GRUB 2 as bootloader. This means that virt-v2v will not change the default kernel used for booting, even in case it is not the best kernel available on the guest. A recommended procedure is, before using virt-v2v, to check that the boot kernel is the best kernel available in the guest (for example by making sure the guest is up-to-date).

"vsyscall attempted with vsyscall=none"

When run on a recent Debian host virt-v2v may fail to convert guests which were created before 2013. In the debugging output you will see a crash message similar to:

 vsyscall attempted with vsyscall=none ip:...
 segfault at ...

This is caused because Debian removed support for running old binaries which used the legacy vsyscall page to call into the kernel.

You can work around this problem by running this command before running virt-v2v:

 export LIBGUESTFS_APPEND="vsyscall=emulate"

For more information, see https://bugzilla.redhat.com/1592061

Windows¶

Windows ≥ 8 Fast Startup is incompatible with virt-v2v

Guests which use the Windows ≥ 8 "Fast Startup" feature (or guests which are hibernated) cannot be converted with virt-v2v. You will see an error:

 virt-v2v: error: unable to mount the disk image for writing. This has
 probably happened because Windows Hibernation or Fast Restart is being
 used in this guest. You have to disable this (in the guest) in order
 to use virt-v2v.

As the message says, you need to boot the guest and disable the "Fast Startup" feature (Control Panel → Power Options → Choose what the power buttons do → Change settings that are currently unavailable → Turn on fast startup), and shut down the guest, and then you will be able to convert it.

For more information, see: "WINDOWS HIBERNATION AND WINDOWS 8 FAST STARTUP" in guestfs(3).

Boot failure: 0x0000007B

This boot failure is caused by Windows being unable to find or load the right disk driver (eg. viostor.sys). If you experience this error, here are some things to check:

• First ensure that the guest boots on the source hypervisor before conversion.
• Check you have the Windows virtio drivers available in /usr/share/virtio-win, and that virt-v2v did not print any warning about not being able to install virtio drivers.

  On Red Hat Enterprise Linux 7, you will need to install the signed drivers available in the "virtio-win" package. If you do not have access to the signed drivers, then you will probably need to disable driver signing in the boot menus.

• Check that you are presenting a virtio-blk interface (not virtio-scsi and not ide) to the guest. On the qemu/KVM command line you should see something similar to this:

   ... -drive file=windows-sda,if=virtio ...
      

  In libvirt XML, you should see:

   <target dev='vda' bus='virtio'/>
      

• Check that Windows Group Policy does not prevent the driver from being installed or used. Try deleting Windows Group Policy before conversion.
• Check there is no anti-virus or other software which implements Group Policy-like prohibitions on installing or using new drivers.
• Enable boot debugging and check the viostor.sys driver is being loaded.

OpenStack and Windows reactivation

OpenStack does not offer stable device / PCI addresses to guests. Every time it creates or starts a guest, it regenerates the libvirt XML for that guest from scratch. The libvirt XML will have no <address> fields. Libvirt will then assign addresses to devices, in a predictable manner. Addresses may change if any of the following are true:

• A new disk or network device has been added or removed from the guest.
• The version of OpenStack or (possibly) libvirt has changed.

Because Windows does not like "hardware" changes of this kind, it may trigger Windows reactivation.

This can also prevent booting with a 7B error [see previous section] if the guest has group policy containing "Device Installation Restrictions".

Support for SHA-2 certificates in Windows 7 and Windows Server 2008 R2

Later versions of the Windows virtio drivers are signed using SHA-2 certificates (instead of SHA-1). The original shipping Windows 7 and Windows Server 2008 R2 did not understand SHA-2 certificates and so the Windows virtio drivers will not install properly.

To fix this you must apply SHA-2 Code Signing Support from: https://docs.microsoft.com/en-us/security-updates/SecurityAdvisories/2015/3033929 before converting the guest.

For further information see: https://bugzilla.redhat.com/show_bug.cgi?id=1624878

Networks and bridges¶

Guests are usually connected to one or more networks, and when converted to the target hypervisor you usually want to reconnect those networks at the destination. The options --network, --bridge and --mac allow you to do that.

If you are unsure of what networks and bridges are in use on the source hypervisor, then you can examine the source metadata (libvirt XML, vCenter information, etc.). Or you can run virt-v2v with the --print-source option which causes virt-v2v to print out the information it has about the guest on the source and then exit.

In the --print-source output you will see a section showing the guest’s Network Interface Cards (NICs):

 $ virt-v2v [-i ...] --print-source name
 [...]
 NICs:
     Network "default" mac: 52:54:00:d0:cf:0e

Bridges are special classes of network devices which are attached to a named external network on the source hypervisor, for example:

 $ virt-v2v [-i ...] --print-source name
 [...]
 NICs:
     Bridge "br0"

To map a specific source bridge to a target network, for example "br0" on the source to "ovirtmgmt" on the target, use:

 virt-v2v [...] --bridge br0:ovirtmgmt

To map every bridge to a target network, use:

 virt-v2v [...] --bridge ovirtmgmt

Fine-grained mapping of guest NICs

The --mac option gives you more control over the mapping, letting you map single NICs to either networks or bridges on the target. For example a source guest with two NICs could map them individually to two networks called "mgmt" and "clientdata" like this:

 $ virt-v2v [...] \
    --mac 52:54:00:d0:cf:0e:network:mgmt \
    --mac 52:54:00:d0:cf:0f:network:clientdata

Note that virt-v2v does not have the ability to change a guest’s MAC address. The MAC address is part of the guest metadata and must remain the same on source and target hypervisors. Most guests will use the MAC address to set up persistent associations between NICs and internal names (like "eth0"), with firewall settings, or even for other purposes like software licensing.

Resource requirements¶

Network

The most important resource for virt-v2v appears to be network bandwidth. Virt-v2v should be able to copy guest data at gigabit ethernet speeds or greater.

Ensure that the network connections between servers (conversion server, NFS server, vCenter, Xen) are as fast and as low latency as possible.

Disk space

Virt-v2v places potentially large temporary files in $TMPDIR (which is /var/tmp if you don't set it). Using tmpfs is a bad idea.

For each guest disk, an overlay is stored temporarily. This stores the changes made during conversion, and is used as a cache. The overlays are not particularly large - tens or low hundreds of megabytes per disk is typical. In addition to the overlay(s), input and output methods may use disk space, as outlined in the table below.

-i ova

This temporarily places a full copy of the uncompressed source disks in $TMPDIR.

-o glance

This temporarily places a full copy of the output disks in $TMPDIR.

-o local

-o qemu

You must ensure there is sufficient space in the output directory for the converted guest.

See also "Minimum free space check in the host" below.

VMware vCenter resources

Copying from VMware vCenter is currently quite slow, but we believe this to be an issue with VMware. Ensuring the VMware ESXi hypervisor and vCenter are running on fast hardware with plenty of memory should alleviate this.

Compute power and RAM

Virt-v2v is not especially compute or RAM intensive. If you are running many parallel conversions, then you may consider allocating one CPU core and 2 GB of RAM per running instance.

Virt-v2v can be run in a virtual machine.

Trimming

Virt-v2v attempts to optimize the speed of conversion by ignoring guest filesystem data which is not used. This would include unused filesystem blocks, blocks containing zeroes, and deleted files.

To do this, virt-v2v issues a non-destructive fstrim(8) operation. As this happens to an overlay placed over the guest data, it does not affect the source in any way.

If this fstrim operation fails, you will see a warning, but virt-v2v will continue anyway. It may run more slowly (in some cases much more slowly), because it is copying the unused parts of the disk.

Unfortunately support for fstrim is not universal, and it also depends on specific details of the filesystem, partition alignment, and backing storage. As an example, NTFS filesystems cannot be fstrimmed if they occupy a partition which is not aligned to the underlying storage. That was the default on Windows before Vista. As another example, VFAT filesystems (used by UEFI guests) cannot be trimmed at all.

fstrim support in the Linux kernel is improving gradually, so over time some of these restrictions will be lifted and virt-v2v will work faster.

Post-conversion tasks¶

Guest network configuration

Virt-v2v cannot currently reconfigure a guest’s network configuration. If the converted guest is not connected to the same subnet as the source, its network configuration may have to be updated. See also virt-customize(1).

Converting a Windows guest

When converting a Windows guests, the conversion process is split into two stages:

1.

Offline conversion.

2.

First boot.

The guest will be bootable after the offline conversion stage, but will not yet have all necessary drivers installed to work correctly. These will be installed automatically the first time the guest boots.

N.B. Take care not to interrupt the automatic driver installation process when logging in to the guest for the first time, as this may prevent the guest from subsequently booting correctly.

Free space for conversion¶

Free space in the guest

Virt-v2v checks there is sufficient free space in the guest filesystem to perform the conversion. Currently it checks:

Linux root filesystem or Windows "C:" drive

Minimum free space: 20 MB

Linux /boot

Minimum free space: 50 MB

This is because we need to build a new initramfs for some Enterprise Linux conversions.

Any other mountable filesystem

Minimum free space: 10 MB

Minimum free space check in the host

You must have sufficient free space in the host directory used to store temporary overlays (except in --in-place mode). To find out which directory this is, use:

 $ df -h "`guestfish get-cachedir`"
 Filesystem        Size  Used Avail Use% Mounted on
 /dev/mapper/root   50G   40G  6.8G  86% /

and look under the "Avail" column. Virt-v2v will refuse to do the conversion at all unless at least 1GB is available there.

See also "Resource requirements" above.

Running virt-v2v as root or non-root¶

Nothing in virt-v2v inherently needs root access, and it will run just fine as a non-root user. However, certain external features may require either root or a special user:

Mounting the Export Storage Domain

When using -o rhv -os server:/esd virt-v2v has to have sufficient privileges to NFS mount the Export Storage Domain from "server".

You can avoid needing root here by mounting it yourself before running virt-v2v, and passing -os /mountpoint instead, but first of all read the next section ...

Writing to the Export Storage Domain as 36:36

RHV-M cannot read files and directories from the Export Storage Domain unless they have UID:GID 36:36. You will see VM import problems if the UID:GID is not correct.

When you run virt-v2v -o rhv as root, virt-v2v attempts to create files and directories with the correct ownership. If you run virt-v2v as non-root, it will probably still work, but you will need to manually change ownership after virt-v2v has finished.

Writing to libvirt

When using -o libvirt, you may need to run virt-v2v as root so that it can write to the libvirt system instance (ie. "qemu:///system") and to the default location for disk images (usually /var/lib/libvirt/images).

You can avoid this by setting up libvirt connection authentication, see http://libvirt.org/auth.html. Alternatively, use -oc qemu:///session, which will write to your per-user libvirt instance.

Writing to Openstack

Because of how Cinder volumes are presented as /dev block devices, using -o openstack normally requires that virt-v2v is run as root.

Writing to Glance

This does not need root (in fact it probably won’t work), but may require either a special user and/or for you to source a script that sets authentication environment variables. Consult the Glance documentation.

Minimal XML for -i libvirtxml option¶

When using the -i libvirtxml option, you have to supply some libvirt XML. Writing this from scratch is hard, so the template below is helpful.

Note this should only be used for testing and/or where you know what you're doing! If you have libvirt metadata for the guest, always use that instead.

 <domain type='kvm'>
   <name> NAME </name>
   <memory>1048576</memory>
   <vcpu>2</vcpu>
   <os>
     <type>hvm</type>
     <boot dev='hd'/>
   </os>
   <features>
     <acpi/>
     <apic/>
     <pae/>
   </features>
   <devices>
     <disk type='file' device='disk'>
       <driver name='qemu' type='raw'/>
       <source file='/path/to/disk/image'/>
       <target dev='hda' bus='ide'/>
     </disk>
     <interface type='network'>
       <mac address='52:54:00:01:02:03'/>
       <source network='default'/>
       <model type='rtl8139'/>
     </interface>
   </devices>
 </domain>

In-place conversion¶

It is also possible to use virt-v2v in scenarios where a foreign VM has already been imported into a KVM-based hypervisor, but still needs adjustments in the guest to make it run in the new virtual hardware.

In that case it is assumed that a third-party tool has created the target VM in the supported KVM-based hypervisor based on the source VM configuration and contents, but using virtual devices more appropriate for KVM (e.g. virtio storage and network, etc.).

Then, to make the guest OS boot and run in the changed environment, one can use:

 virt-v2v -ic qemu:///system converted_vm --in-place

Virt-v2v will analyze the configuration of "converted_vm" in the "qemu:///system" libvirt instance, and apply various fixups to the guest OS configuration to make it match the VM configuration. This may include installing virtio drivers, configuring the bootloader, the mountpoints, the network interfaces, and so on.

Should an error occur during the operation, virt-v2v exits with an error code leaving the VM in an undefined state.

Machine readable output¶

The --machine-readable option can be used to make the output more machine friendly, which is useful when calling virt-v2v from other programs, GUIs etc.

このオプションを使用するには 2 通りの方法があります。

Firstly use the option on its own to query the capabilities of the virt-v2v binary. Typical output looks like this:

 $ virt-v2v --machine-readable
 virt-v2v
 libguestfs-rewrite
 colours-option
 vdsm-compat-option
 input:disk
 [...]
 output:local
 [...]
 convert:linux
 convert:windows

A list of features is printed, one per line, and the program exits with status 0.

The "input:" and "output:" features refer to -i and -o (input and output mode) options supported by this binary. The "convert:" features refer to guest types that this binary knows how to convert.

Secondly use the option in conjunction with other options to make the regular program output more machine friendly.

At the moment this means:

1.

Progress bar messages can be parsed from stdout by looking for this regular expression:

 ^[0-9]+/[0-9]+$
    

2.

The calling program should treat messages sent to stdout (except for progress bar messages) as status messages. They can be logged and/or displayed to the user.

3.

The calling program should treat messages sent to stderr as error messages. In addition, virt-v2v exits with a non-zero status code if there was a fatal error.

Virt-v2v ≤ 0.9.1 did not support the --machine-readable option at all. The option was added when virt-v2v was rewritten in 2014.

It is possible to specify a format string for controlling the output; see "ADVANCED MACHINE READABLE OUTPUT" in guestfs(3).