Basic KVM operations

SEE ALSO!

• virtualisation

• Creation of a new VM

• Access your VM

• Cloning existing VMs

List all the VMs, and see which ones are running

# sudo virsh list --all

Start a VM

You can list the VMs first, then:

# sudo virsh start <name of your vm>

Shutdown a VM properly

Note: this require the VM to have support for ACPI - at the moment it doesn't seem to work with windows VMs (seems OK to me- Jo).

# sudo virsh shutdown <name of the vm>

Shutdown a VM

Same thing as taking the plug off. Use with care.

# sudo virsh destroy <name of the vm>

Delete a VM

THIS REMOVES A VM. NO RESTORE POSSIBLE. PLEASE USE WITH SPECIAL CARE.

# sudo virsh shutdown <name of the vm>
# sudo virsh undefine <name of the vm>
# sudo rm -rf /var/lib/kvm/<name of the vm>

Back up a VM

After shutting down:

# cd /var/lib/kvm
# cp -R <name of the vm> /to/safe/place

Copy or move a VM to another host

• Using scp, we find that a qcow2 disk will expand to its full allocated size when moved or copied to another host (eg a 20gb VM may only take up 5gb of actual disk space, but when copied it takes up the full 20gb)

• Eduardo found a solution using rsync. Example:

 rsync /var/lib/kvm/vanilla_templates/w2k3-vanilla3/w2k3-vanilla3.qcow2 ecoria@leibniz.goo.thehumanjourney.net:/home/ecoria/test/

Second method

In case you have some permission problem you could use this method. Copy the VM image to remote server in /tmp dir:

rsync /var/lib/kvm/<vm-name>/<vm-image> <remote-server-ip>:/tmp/

On the remote server create a dir for the VM and copy it from /tmp to that dir:

mkdir /var/lib/kvm/<vm-name>
mv /tmp/<vm-image> /var/lib/kvm/<vm-name>/

Now you also need the xml file for that VM. Since it is just text you can copy it, create a file on remote server and paste it there. The xml file is found in this dir:

/etc/libvirt/qemu

After you copied the VM image and xml file you can define and start it:

virsh define /etc/libvirt/qemu/<vm-name>.xml
virsh start <vm-name>

Rename a VM

Shutdown the VM, dump the xml file and edit it.

virsh shutdown oldname
virsh dumpxml oldname > /etc/libvirt/qemu/newname.xml
vi /etc/libvirt/qemu/newname.xml

In newname.xml change "<name>oldname</name>" to "<name>newname</name>" and "

" to "

". Undefine the VM and define the new one.

virsh undefine oldname
mv /var/lib/kvm/oldname /var/lib/kvm/newname
virsh define /etc/libvirt/qemu/newname.xml
virsh start newname

ssh to VM and follow the steps from here: changing the hostname

VM-disk deleted - resolution

Solution 1

# lsof | grep 101
...
kvm       3649       root   18u      REG              253,2 2147483648     524290  (deleted)/var/lib/vz/images/101/vm-101-disk-1.raw
...

Important is the pid (3649) and the filediscriptor (18) do an copy (before that, it's a good idea to stop important services with open files inside the VM, like databases)

cp /proc/3649/fd/18 /var/lib/vz/images/101/vm-101-disk-1.raw

Voila, the disk-file (of an open-VM!) is copied back.

Solution 2

To conver the image from raw to qcow2

qemu-img convert -p -f qcow2 -O raw /proc/2850/fd/19 vm-100-disk-1.raw

Test the copied file with an dummy-VM before shutdown the original VM!

List all the VMs with the autostart option activated

# sudo virsh list --autostart

Snapshoot operations

Create snapshoot

virsh shutdown vm_name
virsh snapshot-create-as --domain vm_name \
--name "snapshot_name" \
--description "Snapshpot"

Restore snapshoot

virsh snapshot-list --domain vm_name
virsh snapshot-revert --domain vm_name snapshot_name

Delete snapshoot

virsh snapshot-delete --domain vm_name --snapshotname snapshot_name_to_be_deleted

Adding a disk to VM

On the host, create an image

qemu-img create -f qcow2 /path/to/image.qcow2 10G

Attach the disk to the VM (check before if vda is used; if yes use vdb, vdc, etc.)

virsh shutdown vm-name
virsh attach-disk vm-name --source /path/to/image.qcow2 --target vda --persistent
virsh start vm-name

On the VM, check disk vda has been attached

# ls -l /dev/disk/by-path/
lrwxrwxrwx 1 root root  9 Oct 23 13:44 virtio-pci-0000:00:05.0 -> ../../vda

Create a partition with parted or other tool

parted /dev/vda

If it shows that the disk size is only a few KB, it's an error and it can be resolved by converting the image, detach the image and attach the converted one

qemu-img convert -O qcow2 input-image.qcow2 output-image.qcow2

Format it

mkfs.ext4 /dev/vda1

If the disk will be used for data storage then reduce the space reserved for root operations

tune2fs -m 1 /dev/vda1

Find the UUID of the partition (the groups of numbers/letters before the arrow -> )

#ls -l /dev/disk/by-uuid/
lrwxrwxrwx 1 root root 10 Oct 23 13:45 936de7fd-084f-4456-898f-483341df1cf1 -> ../../vda1

Mount the partition using the UUID

mount /dev/disk/by-uuid/youruuid /path/to/mymountpoint

Append a line to /etc/fstab so your mount can survive a reboot

UUID=youruuid       /path/to/mymountpoint ext4    defaults        0       0

Detach disk from a VM

virsh detach-disk vm-name vda --config

Converting

Info about image

qemu-img info image.vmdk 

From .vdi to .qcow2

qemu-img convert -f vdi -O qcow2 from.vdi to.qcow2

From qcow2 to .vdi

qemu-img convert -O vdi from.qcow2 to.vdi

From .img to .qcow2

qemu-img convert -f raw from.img -O qcow2 to.qcow2

From .qcow2 to .raw

qemu-img convert -O raw from.qcow2 to.raw