(kvm) qemu virsh
migrate guest to new host
old# virsh list
old# virsh list --all
old# virsh dumpxml Void > Void.xml

old# cp Void.xml Void.qcow2 /mnt ; sync

...

new# virsh create Void.xml   # disappears after shutdown, or
new# virsh define Void.xml   # persistent, not started

persistent domain

migrate

learn properly


note that if the source and target hypervisor differ more in terms of versions, compilation 
settings etc it might be safer to try the change first with the virt-manager GUI

pay attention to the keyboard and mouse options, e.g. PS/2 vs USB vs VirtIO...and maybe to
graphics options, e.g. Virtio vs. QXL

no default network?
shell# virsh
net-list --all
net-start default
net-autostart default
net-list --all

LVM raw disk to qcow2 file...
See Converting kvm guests from lvm to qcow2, base images and snapshots

...and back?
  cd /var/lib/libvirt/images/
  qemu-img convert -O raw alpine.qcow2  alpine.raw
  ls -l alpine.raw
  -rw------- 1 root root 1811160064 Nov  2 12:01 alpine.raw
  #                      ^^^^^^^^^^                         
  lvcreate -L 1811160064b -n alpine guests
  dd if=alpine.raw of=/dev/guests/alpine bs=4M


After migration edit xml or virsh attach-disk or switch disk
in virt-manager


squeeze images, save disk space
consider starting with BleachBit to trim out temporary files etc (often readily available in your distribution)

then guestfs-tools: virt-sparsify

  • if enough RAM available, speed up by TMPDIR=/your/path/to/tmpfs/ramdisk  ...e.g.
      mount -t tmpfs -o size=30g tmpfs /mnt

  • if not enough free space in /tmp use any TMPDIR big enough (e.g. /tmp on a small subvolume)

  • ! on SuSE also install libguestfs-appliance rpm-devel

  • in Fedora also make sure libvirt has write access to TMPDIR, or you get
      virt-sparsify: error: libguestfs error: could not create appliance through libvirt.

without (big) guestfs-tools?
cp image.qcow2 image.qcow2.previous

# without compression, faster but bigger
qemu-img convert -O qcow2 image.qcow2.previous image.qcow2

# with compression, smaller but slower
qemu-img convert -O qcow2 -c image.qcow2.previous image.qcow2

but do sparsify before compression for best (much smaller) results

many disk ops with
qemu-img

all at once
from raw LVM to sparsified then compressed QCOW2 file for back up
convenience, slight speed up with ramdisk

   mount -t tmpfs -o size=40g tmpfs /mnt
   mkdir /mnt/tmp
   cd /mnt/
   qemu-img convert -O qcow2 /dev/guests/suse suse-uncompressed.qcow2
   TMPDIR=/mnt/tmp virt-sparsify suse-uncompressed.qcow2 suse-sparsified.qcow2
   qemu-img convert -O qcow2 -c suse-sparsified.qcow2 ~/backups/suse.qcow2
   sync
   cd
   umount /mnt
   virsh dumpxml suse > ~/backups/suse.xml


Depending on the partition sizes and layout, e.g. if /tmp is big enough, simply...

  cd /var/lib/libvirt/images/
  rm -rf /tmp/tmp  &&  mkdir /tmp/tmp  &&  for i in  *2  ; do TMPDIR=/tmp/tmp virt-sparsify $i /tmp/tmp.qcow2  &&  sync  &&  qemu-img convert -O qcow2 -c /tmp/tmp.qcow2 $i  &&  sync  &&  ls -lh $i; done  &&  rm -rf /tmp/tmp /tmp/tmp.qcow2

...could be enough if you don't need the intermediate files


If /tmp is too small because of tmpfs and small RAM or small btrfs slice size use some suitable filesystem, e.g. /var/tmp instead


On new Fedora 38 install it seems you need

	# virt-sparsify: error: libguestfs error: could not create appliance through 
	# libvirt.  Try running qemu directly without libvirt using this environment variable
	export LIBGUESTFS_BACKEND=direct


Note if migrating from bleeding edge host to LTS host, xml could include too
fancy virtual hardware features.

In that case try to create new machine and just use the disk image.

diff backups
   mkdir old new
   guestmount -a /backups/week/host.qcow2 -i --ro old/
   guestmount -a /backups/day/host.qcow2  -i --ro new/

   diff -r old new | less

   # see new items only
   diff -r old new 2>/dev/null | grep '^Only in new' | less



See also:
 Virt-Manager like a Pro
 virt-sysprep
 RaspBerry ARMv8 64bit as server
 usb redir
 use your QEMU wiser with Quickemu 
 even live blockcopy