virt-v2v-output-openstack - Man Page

When virt-v2v is converting to OpenStack, it is unusual in that virt-v2v must be running inside a virtual machine running on top of the OpenStack overcloud.  This virtual machine is called the "conversion appliance".  Note this virtual machine is unrelated to the guest which is being converted.

The reason for this is because to create Cinder volumes that will contain the guest data (for the converted guest) we must attach those Cinder volumes to an OpenStack virtual machine.

The openstack command must be installed in the appliance.  We use it for communicating with OpenStack.

When virt-v2v is running in the conversion appliance, you must supply the name or UUID of the conversion appliance to virt-v2v, eg:

 $ openstack server list
 +--------------------------------------+-----------+--------+
 | ID                                   | Name      | Status |
 +--------------------------------------+-----------+--------+
 | bbb0147a-44b9-4d19-9a9d-10ca9a984744 | test1     | ACTIVE |
 +--------------------------------------+-----------+--------+

 # virt-v2v [...] \
       -o openstack -oo server-id=bbb0147a-44b9-4d19-9a9d-10ca9a984744

or:

 # virt-v2v [...] -o openstack -oo server-id=test1

You can run many parallel conversions inside a single conversion appliance if you want, subject to having enough resources available. However OpenStack itself imposes a limit that you should be aware of: OpenStack cannot attach more than around 25 disks [the exact number varies with configuration] to a single appliance, and that limits the number of guests which can be converted in parallel, because each guest's disk must be attached to the appliance while being copied.

Converting to OpenStack requires access to the tenant (non-admin) API endpoints.  You will need to either set up your $OS_* environment variables or use output options on the virt-v2v command line to authenticate with OpenStack.

For example:

 export OS_USERNAME=admin

or:

 virt-v2v [...] -o openstack -oo os-username=admin

are equivalent, and have the same effect as using --os-username on the command line of OpenStack tools.

Normally there is a file called overcloudrc or keystonerc_admin which you can simply source to set everything up.

If you need to copy overcloudrc to another machine, check if it references the OS_CLOUD environment variable.  If so, you may have to copy the clouds.yaml file from /etc/openstack/ or $HOME/.config/openstack/ to the other machine as well.  See the section "CLOUD CONFIGURATION" in the openstack CLI manual.

Because virt-v2v must access Cinder volumes which are presented as /dev devices to the conversion appliance, virt-v2v must usually run as root in -o openstack mode.

If you use sudo to start virt-v2v and you are using environment variables for authentication, remember to use the sudo -E option to preserve the environment.

 virt-v2v [...] -o openstack -oo guest-id=123-456-7890

You may optionally specify -oo guest-id=... on the command line. The ID (which can be any string) is saved on each Cinder volume in the virt_v2v_guest_id volume property.

This can be used to find disks associated with a guest, or to associate which disks are related to which guests when converting many guests.

Using -oo verify-server-certificate=false you can tell the openstack client to ignore the server certificate when connecting to the OpenStack API endpoints.  This has the same effect as passing the --insecure option to the openstack command.

The final command to convert the guest, running as root, will be:

 # virt-v2v [-i options ...] \
       -o openstack -oo server-id=NAME|UUID [-oo guest-id=ID]

If you include authentication options on the command line then:

 # virt-v2v [-i options ...] \
       -o openstack -oo server-id=NAME|UUID -oo os-username=admin [etc]

Guests are converted as Cinder volume(s) (one volume per disk in the original guest).  To boot them use the openstack server create --volume option:

 $ openstack volume list
 +--------------------------------------+---------------+-----------+
 | ID                                   | Name          | Status    |
 +--------------------------------------+---------------+-----------+
 | c4d06d15-22ef-462e-9eff-ab54ab285a1f | fedora-27-sda | available |
 +--------------------------------------+---------------+-----------+
 $ openstack server create \
       --flavor x1.small \
       --volume c4d06d15-22ef-462e-9eff-ab54ab285a1f \
       myguest
 $ openstack console url show myguest

To specify the Cinder volume type, use -os.  If not specified then no Cinder volume type is used.

The following options are not supported with OpenStack: -oa, -of.

Glance image upload doesn't appear to correctly handle sparseness. For this reason, using qcow2 will be faster and use less space on the Glance server.  Use the virt-v2v -of qcow2 option.

If the guest has a single disk, then the name of the disk in Glance will be the name of the guest.  You can control this using the -on option.

Glance doesn't have a concept of associating multiple disks with a single guest, and Nova doesn't allow you to boot a guest from multiple Glance disks either.  If the guest has multiple disks, then the first (assumed to be the system disk) will have the name of the guest, and the second and subsequent data disks will be called guestname-disk2, guestname-disk3 etc.  It may be best to leave the system disk in Glance, and import the data disks to Cinder.