David Díaz | KVM Backups, the easy way

I am now using a dedicated server to host all of my various projects. It turns out if you are willing to do a lot of the work yourself you can get quite a bit for you buck. Since I have to take care of backups myself now I had to rethink my general backup strategy.

In the name of being as lazy as possible what I've been doing now is using libvirt to create individual VMs for every project. To back them up I can just copy off the entire disk somewhere else. Once compressed these are relatively small for what they are and are easy to move around into storage.

I am also using git underneath almost everything so that actual code and assets are also backed up locally and on github/gitea etc.

Anyways, here's what I am using to back machines up:

DOMAIN="$1"

DISK="$(virsh domblklist $DOMAIN | tail -n +3 | awk '{print $2}' | tr -d '\n')";
TIME=$(date "+%Y-%m-%d--%H.%M.%S");

virsh snapshot-create-as --domain $DOMAIN "$DOMAIN-backup" \
        --diskspec vda,file=/var/lib/libvirt/images/$DOMAIN-backup.qcow2 \
        --disk-only --atomic --quiesce --no-metadata;

xz -k --verbose /var/lib/libvirt/images/$DOMAIN.qcow2;

curl --netrc -T /var/lib/libvirt/images/$DOMAIN.qcow2.xz ftp://ftp.example.org/$DOMAIN.$TIME.qcow2.xz;

virsh blockcommit $DOMAIN vda --active --verbose --pivot;
rm /var/lib/libvirt/images/$DOMAIN.qcow2.xz;
rm /var/lib/libvirt/images/$DOMAIN-backup.qcow2;

First I create a snapshot for the current machine I'll be backing up, these machines are running the qemu agent so I get to use --quiesce to freeze the file system. If you don't have the agent installed you'd probably have to make some small changes to the command.

virsh snapshot-create-as --domain $DOMAIN "$DOMAIN-backup" \
        --diskspec vda,file=/var/lib/libvirt/images/$DOMAIN-backup.qcow2 \
        --disk-only --atomic --quiesce --no-metadata;

Once that's done I'll compress it using xz. This will make the backup tiny but it takes a really long time, you could use zip, tar or nothing at all if you want the whole process to be faster.

xz -k --verbose /var/lib/libvirt/images/$DOMAIN.qcow2;

Now we can upload the compressed file to our backup location. I use curl to upload to a local network drive I have available. I also attach a timestamp to the file name as I do this.

curl --netrc -T /var/lib/libvirt/images/$DOMAIN.qcow2.xz ftp://ftp.example.org/$DOMAIN.$TIME.qcow2.xz;

Finally I will use blockcommit to pivot back to the original disk image, merging the contents of the backup and anything that's happened since the snapshot was taken, and remove the unnecessary snapshot and compressed files.

virsh blockcommit $DOMAIN vda --active --verbose --pivot;
rm /var/lib/libvirt/images/$DOMAIN.qcow2.xz;
rm /var/lib/libvirt/images/$DOMAIN-backup.qcow2;

It took me a bit to figure out how to do this which is why I am documenting it here for myself. Most of this is based of off this wiki entry.

Have a comment? Feel free to email me.
Did you enjoy this post? Buy me a coffee ☕️.

David Díaz.