virt-v2v - Man Page

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 conversions may be done using virt-v2v-in-place(1).

Virt-v2v can also customize the guest during conversion, using the same options as virt-customize(1).  For example, injecting files using --upload, or using --firstboot-script to add additional script(s) to run at the first boot after conversion.  Read the virt-customize manual for more information on this topic.

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-release-notes-1.42(1) — Release notes for 1.42 release.

virt-v2v-release-notes-2.0(1) — Release notes for 2.0 release.

virt-v2v-release-notes-2.2(1) — Release notes for 2.2 release.

virt-v2v-release-notes-2.4(1) — Release notes for 2.4 release.

virt-v2v-release-notes-2.6(1) — Release notes for 2.6 release.

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).

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 --bridge ovirtmgmt

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

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

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).

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).

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 -oo qemu-boot

--append-line FILE:LINE

Append a single line of text to the FILE.  If the file does not already end with a newline, then one is added before the appended line.  Also a newline is added to the end of the LINE string automatically.

For example (assuming ordinary shell quoting) this command:

 --append-line '/etc/hosts:10.0.0.1 foo'

will add either 10.0.0.1 foo⏎ or ⏎10.0.0.1 foo⏎ to the file, the latter only if the existing file does not already end with a newline.

⏎ represents a newline character, which is guessed by looking at the existing content of the file, so this command does the right thing for files using Unix or Windows line endings. It also works for empty or non-existent files.

To insert several lines, use the same option several times:

 --append-line '/etc/hosts:10.0.0.1 foo'
 --append-line '/etc/hosts:10.0.0.2 bar'

To insert a blank line before the appended line, do:

 --append-line '/etc/hosts:'
 --append-line '/etc/hosts:10.0.0.1 foo'

--chmod PERMISSIONS:FILE

Change the permissions of FILE to PERMISSIONS.

Note: PERMISSIONS by default would be decimal, unless you prefix it with 0 to get octal, ie. use 0700 not 700.

--chown UID:GID:PATH

Change the owner user and group ID of a file or directory in the guest. Note:

• Only numeric UIDs and GIDs will work, and these may not be the same inside the guest as on the host.
• This will not work with Windows guests.

For example:

 virt-customize --chown '0:0:/var/log/audit.log'

See also: --upload.

--commands-from-file FILENAME

Read the customize commands from a file, one (and its arguments) each line.

Each line contains a single customization command and its arguments, for example:

 delete /some/file
 install some-package
 password some-user:password:its-new-password

Empty lines are ignored, and lines starting with # are comments and are ignored as well.  Furthermore, arguments can be spread across multiple lines, by adding a \ (continuation character) at the of a line, for example

 edit /some/file:\
   s/^OPT=.*/OPT=ok/

The commands are handled in the same order as they are in the file, as if they were specified as --delete /some/file on the command line.

--copy SOURCE:DEST

Copy files or directories recursively inside the guest.

Wildcards cannot be used.

--copy-in LOCALPATH:REMOTEDIR

Copy local files or directories recursively into the disk image, placing them in the directory REMOTEDIR (which must exist).

Wildcards cannot be used.

--delete PATH

Delete a file from the guest.  Or delete a directory (and all its contents, recursively).

You can use shell glob characters in the specified path.  Be careful to escape glob characters from the host shell, if that is required. For example:

 virt-customize --delete '/var/log/*.log'.

See also: --upload, --scrub.

--edit FILE:EXPR

Edit FILE using the Perl expression EXPR.

Be careful to properly quote the expression to prevent it from being altered by the shell.

Note that this option is only available when Perl 5 is installed.

See "NON-INTERACTIVE EDITING" in virt-edit(1).

--firstboot SCRIPT

Install SCRIPT inside the guest, so that when the guest first boots up, the script runs (as root, late in the boot process).

The script is automatically chmod +x after installation in the guest.

The alternative version --firstboot-command is the same, but it conveniently wraps the command up in a single line script for you.

You can have multiple --firstboot options.  They run in the same order that they appear on the command line.

Please take a look at "FIRST BOOT SCRIPTS" in virt-builder(1) for more information and caveats about the first boot scripts.

See also --run.

--firstboot-command 'CMD+ARGS'

Run command (and arguments) inside the guest when the guest first boots up (as root, late in the boot process).

You can have multiple --firstboot options.  They run in the same order that they appear on the command line.

Please take a look at "FIRST BOOT SCRIPTS" in virt-builder(1) for more information and caveats about the first boot scripts.

See also --run.

--firstboot-install PKG,PKG..

Install the named packages (a comma-separated list).  These are installed when the guest first boots using the guest’s package manager (eg. apt, yum, etc.) and the guest’s network connection.

For an overview on the different ways to install packages, see "INSTALLING PACKAGES" in virt-builder(1).

--hostname HOSTNAME

Set the hostname of the guest to HOSTNAME.  You can use a dotted hostname.domainname (FQDN) if you want.

--install PKG,PKG..

Install the named packages (a comma-separated list).  These are installed during the image build using the guest’s package manager (eg. apt, yum, etc.) and the host’s network connection.

For an overview on the different ways to install packages, see "INSTALLING PACKAGES" in virt-builder(1).

See also --update, --uninstall.

--link TARGET:LINK[:LINK..]

Create symbolic link(s) in the guest, starting at LINK and pointing at TARGET.

--mkdir DIR

Create a directory in the guest.

This uses mkdir -p so any intermediate directories are created, and it also works if the directory already exists.

--move SOURCE:DEST

Move files or directories inside the guest.

Wildcards cannot be used.

--no-logfile

Scrub builder.log (log file from build commands) from the image after building is complete.  If you don't want to reveal precisely how the image was built, use this option.

See also: "LOG FILE".

--no-selinux-relabel

Do not attempt to correct the SELinux labels of files in the guest.

In such guests that support SELinux, customization automatically relabels files so that they have the correct SELinux label.  (The relabeling is performed immediately, but if the operation fails, customization will instead touch /.autorelabel on the image to schedule a relabel operation for the next time the image boots.)  This option disables the automatic relabeling.

The option is a no-op for guests that do not support SELinux.

--password USER:SELECTOR

Set the password for USER.  (Note this option does not create the user account).

See "USERS AND PASSWORDS" in virt-builder(1) for the format of the SELECTOR field, and also how to set up user accounts.

--password-crypto md5|sha256|sha512

When the virt tools change or set a password in the guest, this option sets the password encryption of that password to md5, sha256 or sha512.

sha256 and sha512 require glibc ≥ 2.7 (check crypt(3) inside the guest).

md5 will work with relatively old Linux guests (eg. RHEL 3), but is not secure against modern attacks.

The default is sha512 unless libguestfs detects an old guest that didn't have support for SHA-512, in which case it will use md5. You can override libguestfs by specifying this option.

Note this does not change the default password encryption used by the guest when you create new user accounts inside the guest. If you want to do that, then you should use the --edit option to modify /etc/sysconfig/authconfig (Fedora, RHEL) or /etc/pam.d/common-password (Debian, Ubuntu).

--root-password SELECTOR

Set the root password.

See "USERS AND PASSWORDS" in virt-builder(1) for the format of the SELECTOR field, and also how to set up user accounts.

Note: In virt-builder, if you don't set --root-password then the guest is given a random root password.

--run SCRIPT

Run the shell script (or any program) called SCRIPT on the disk image.  The script runs virtualized inside a small appliance, chrooted into the guest filesystem.

The script is automatically chmod +x.

If libguestfs supports it then a limited network connection is available but it only allows outgoing network connections.  You can also attach data disks (eg. ISO files) as another way to provide data (eg. software packages) to the script without needing a network connection (--attach).  You can also upload data files (--upload).

You can have multiple --run options.  They run in the same order that they appear on the command line.

See also: --firstboot, --attach, --upload.

--run-command 'CMD+ARGS'

Run the command and arguments on the disk image.  The command runs virtualized inside a small appliance, chrooted into the guest filesystem.

If libguestfs supports it then a limited network connection is available but it only allows outgoing network connections.  You can also attach data disks (eg. ISO files) as another way to provide data (eg. software packages) to the script without needing a network connection (--attach).  You can also upload data files (--upload).

You can have multiple --run-command options.  They run in the same order that they appear on the command line.

See also: --firstboot, --attach, --upload.

--scrub FILE

Scrub a file from the guest.  This is like --delete except that:

• It scrubs the data so a guest could not recover it.
• It cannot delete directories, only regular files.

--selinux-relabel

This is a compatibility option that does nothing.

--sm-attach SELECTOR

Attach to a pool using subscription-manager.

See "SUBSCRIPTION-MANAGER" in virt-builder(1) for the format of the SELECTOR field.

--sm-credentials SELECTOR

Set the credentials for subscription-manager.

See "SUBSCRIPTION-MANAGER" in virt-builder(1) for the format of the SELECTOR field.

--sm-register

Register the guest using subscription-manager.

This requires credentials being set using --sm-credentials.

--sm-remove

Remove all the subscriptions from the guest using subscription-manager.

--sm-unregister

Unregister the guest using subscription-manager.

--ssh-inject USER[:SELECTOR]

Inject an ssh key so the given USER will be able to log in over ssh without supplying a password.  The USER must exist already in the guest.

See "SSH KEYS" in virt-builder(1) for the format of the SELECTOR field.

You can have multiple --ssh-inject options, for different users and also for more keys for each user.

--tar-in TARFILE:REMOTEDIR

Copy local files or directories from a local tar file called TARFILE into the disk image, placing them in the directory REMOTEDIR (which must exist).  Note that the tar file must be uncompressed (.tar.gz files will not work here)

--timezone TIMEZONE

Set the default timezone of the guest to TIMEZONE.  Use a location string like Europe/London

--touch FILE

This command performs a touch(1)-like operation on FILE.

--truncate FILE

This command truncates FILE to a zero-length file. The file must exist already.

--truncate-recursive PATH

This command recursively truncates all files under PATH to zero-length.

--uninstall PKG,PKG..

Uninstall the named packages (a comma-separated list).  These are removed during the image build using the guest’s package manager (eg. apt, yum, etc.).  Dependent packages may also need to be uninstalled to satisfy the request.

See also --install, --update.

--update

Do the equivalent of yum update, apt-get upgrade, or whatever command is required to update the packages already installed in the template to their latest versions.

See also --install, --uninstall.

--upload FILE:DEST

Upload local file FILE to destination DEST in the disk image. File owner and permissions from the original are preserved, so you should set them to what you want them to be in the disk image.

DEST could be the final filename.  This can be used to rename the file on upload.

If DEST is a directory name (which must already exist in the guest) then the file is uploaded into that directory, and it keeps the same name as on the local filesystem.

See also: --mkdir, --delete, --scrub.

--write FILE:CONTENT

Write CONTENT to FILE.

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)

"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 the "VIRTIO_WIN" environment variable if present.
                If the "VIRTIO_WIN" environment variable is absent
                (which is the recommended setting), then libosinfo is
                consulted first, for driver files that are locally
                available on the conversion host.

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

"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

System disk on a Dynamic Disk is not supported

If the Windows system disk (the drive containing \windows) is located on a Dynamic Disk then it cannot be converted.  Data disks — that is, disks which are part of the guest but do not contain parts of the Windows operating system — may be Dynamic Disks.

See https://bugzilla.redhat.com/2140548.

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

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.

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 $VIRT_V2V_TMPDIR (usually /var/tmp, see also "ENVIRONMENT VARIBLES" below).  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 $VIRT_V2V_TMPDIR (or /var/tmp).

-o glance

This temporarily places a full copy of the output disks in $VIRT_V2V_TMPDIR (or /var/tmp).

-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.

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. Windows may reboot 4 or more times the first time after conversion.  This is required to install the required drivers, guest agents, remove VMware Tools, and configure the network.  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.

Removing VMware Tools from Windows guests

Virt-v2v attempts to remove VMware Tools.  For Windows guests this is supposed to happen during the first boot after conversion.

We use VMware's recommended uninstallation method as that is the safest choice.  If this fails, VMware Tools must be manually removed by some other method.

One possible method is described here: https://gist.github.com/broestls/f872872a00acee2fca02017160840624 You should carefully check this script since it makes very invasive changes to the Windows Registry and filesystem.

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

Minimum free space: 100 MB

Linux /boot

Minimum free space: 50 MB

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

Windows C: drive

Minimum free space: 100 MB

We may have to copy in many virtio drivers and guest agents.

Any other mountable filesystem

Minimum free space: 10 MB

In addition to the actual free space, each filesystem is required to have at least 100 available inodes.

Minimum free space check in the host

You must have sufficient free space in the host directory used to store large temporary overlays.  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.  You can change the directory that virt-v2v uses by setting $VIRT_V2V_TMPDIR.

See also "Resource requirements" above and "Environment Variables" below.

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.

See also "Starting the libvirt system 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.

Writing to block devices

This normally requires root.  See the next section.

Some output modes write to local files.  In general these modes also let you write to block devices, but before you run virt-v2v you may have to arrange for symbolic links to the desired block devices in the output directory.

For example if using -o local -os /dir then virt-v2v would normally create files called:

 /dir/name-sda     # first disk
 /dir/name-sdb     # second disk
 ...
 /dir/name.xml     # metadata

If you wish the disks to be written to block devices then you would need to create /dir/name-sda (etc) as symlinks to the block devices:

 # lvcreate -L 10G -n VolumeForDiskA VG
 # lvcreate -L 6G -n VolumeForDiskB VG
 # ln -sf /dev/VG/VolumeForDiskA /dir/name-sda
 # ln -sf /dev/VG/VolumeForDiskB /dir/name-sdb

Note that you must precreate the correct number of block devices of the correct size.  Typically -of raw has to be used too, but other formats such as qcow2 can be useful occasionally so virt-v2v does not force you to use raw on block devices.

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>

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.

There are two ways to use this option.

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).

 Failed to connect socket to '/var/run/libvirt/virtqemud-sock': No such file or directory
 Failed to connect socket to '/var/run/libvirt/virtqemud-sock-ro': Connection refused

If you have just installed libvirt and virt-v2v, then you may see the errors above.  This is caused by libvirt daemons that provide various services not running straight after installation.  (This may depend on your distribution and vendor presets).

To fix this on systemd-based distributions, do:

 systemctl isolate multi-user.target

See also https://bugzilla.redhat.com/2182024.