Virter
Virter is a command line tool for simple creation and cloning of virtual
machines.
Virter supports VMs running standard general purpose distributions such as
CentOS and Ubuntu. It is especially useful for development and testing of
projects which cannot use containers due to kernel dependencies, such as
DRBD and
LINSTOR.
Quick Start
First install and set up libvirt. Then
download one of the releases.
Virter is packaged as a single binary, so just put that into /usr/local/bin
and
you are ready to use Virter:
virter image pull alma-8 # also would be auto-pulled in next step
virter vm run --name alma-8-hello --id 100 --wait-ssh alma-8
virter vm ssh alma-8-hello
virter vm rm alma-8-hello
In the above example, the virter "ID" is the index into the possible IPs
within the IP range of the network definition. For a typical 192.x.y.z/24
net definition, the virter "ID" would be z, or 100 in this case.
Depending on your host distribution and libvirt installation, you may need to:
Usage
For usage just run virter help
.
Additional options that may be helpful when using virter vm run
:
-
List all available images: virter image ls --available
-
Using a provisioning file: --provision /root/alma8.toml
-
Adding additional disk(s):
--disk "name=disk1,size=20GiB,format=qcow2,bus=virtio"
-
Adding a bridged interface
--nic "type=bridge,source=br0,mac=1a:2b:3c:4d:5e:01"
-
Adding a NAT'ed interface:
--nic "type=network,source=default,mac=1a:2b:3c:4d:5e:01"
Other examples are provided in the examples directory. See the
README files for the individual examples.
There is also additional documentation for the provisioning
feature. This is useful for defining new images. These
images can be used to start clusters of cloned VMs.
Installation Details
Virter requires:
- A running libvirt daemon on the host where it is run
- When container provisioning is used: A container runtime.
Currently, Virter supports
docker
and podman
.
Configuration is read by default from ~/.config/virter/virter.toml
.
When starting Virter for the first time, a default configuration file will be
generated, including documentation about the various flags.
Container runtime
Select the container runtime by setting container.provider
to either docker
or podman
.
podman
Virter communicates with podman
via it's REST-API. Make sure the API socket is available.
This may be done by:
- Starting podman via systemd:
systemctl --user start podman.socket
(use systemctl --user enable --now podman.socket
to make this permanent)
- Start podman manually:
podman system service
Network domain
If you require DNS resolution from your VMs to return correct FQDNs, add the
domain
to your libvirt network definition:
<network>
...
<domain name='test'/>
...
</network>
By default, Virter uses the libvirt network named default
.
Check out doc/networks.md
for more details on VM networking.
Connecting with ssh
SSH can be configured for convenient access to virtual machines created by Virter.
See doc/ssh.md
for details.
DHCP Leases
Libvirt produces some weird behavior when MAC or IP addresses are reused while
there is still an active DHCP lease for them. This can result in a new VM
getting assigned a random IP instead of the IP corresponding to its ID.
To work around this, Virter tries to execute the dhcp_release
utility in
order to release the DHCP lease from libvirt's DHCP server when a VM is
removed. This utility has to be run by the root user, so Virter executes
it using sudo
.
If execution fails (for example because the utility is not installed or the
sudo rules are not set up correctly), the error is ignored by Virter.
So, to make Virter work more reliably, especially when you are running lots
of VMs in a short amount of time, you should install the dhcp_release
utility
(usually packaged as dnsmasq-utils
). Additionally, you should make sure that
your user can run dhcp_release
as root, for example by using a sudo rule like
this:
%libvirt ALL=(ALL) NOPASSWD: /usr/bin/dhcp_release
This allows all users in the group libvirt to run the dhcp_release
utility
without being prompted for a password.
Console logs
The --console
argument to virter vm run
causes serial output from the VM to
be saved to a file. This file is created with the current user as the owner.
However, it is written to by libvirt, so it needs to located on a filesystem to
which libvirt can write. NFS mounts generally cannot be used due to
root_squash
.
In addition, when the VM generates a lot of output, this can trigger virtlogd
to roll the log file over, which creates a file owned by root (assuming
virtlogd
is running as root). To prevent this, increase max_size
in
/etc/libvirt/virtlogd.conf
.
libvirt storage pool
Virter requires a libvirt storage pool for its images and VM volumes. By
default, it expects a pool named default
. Some libvirt distributions
configure this pool automatically. Some, such as the standard packages on
Ubuntu Focal, do not. If the pool does not exist, create it like this:
virsh pool-define-as --name default --type dir --target /var/lib/libvirt/images
virsh pool-autostart default
virsh pool-start default
AppArmor
On some distributions, AppArmor denies access to /var/lib/libvirt/images
by default.
This leads to messages in dmesg along the lines of:
[ 4274.237593] audit: type=1400 audit(1598348576.161:102): apparmor="DENIED" operation="open" profile="libvirt-d84ef9d7-a7ad-4388-bd5d-cfc3a3db28a6" name="/var/lib/libvirt/images/centos-8" pid=14918 comm="qemu-system-x86" requested_mask="r" denied_mask="r" fsuid=64055 ouid=64055
This can be circumvented by overriding the AppArmor abstraction for that directory:
cat <<EOF >> /etc/apparmor.d/local/abstractions/libvirt-qemu
/var/lib/libvirt/images/* rwk,
# required for QEMU accessing UEFI nvram variables
/usr/share/OVMF/* rk,
owner /var/lib/libvirt/qemu/nvram/*_VARS.fd rwk,
owner /var/lib/libvirt/qemu/nvram/*_VARS.ms.fd rwk,
EOF
systemctl restart apparmor.service
systemctl restart libvirtd.service
Architecture
Virter connects to the libvirt daemon for all the heavy lifting. It supplies
bootstrap configuration to the VMs using cloud-init
volumes, so that the
hostname is set and SSH access is possible.
Building from source
If you want to test the latest unstable version of Virter, you can build the
git version from sources:
git clone https://github.com/LINBIT/virter
cd virter
go build .
virsh
Virter is good for starting and cloning cloud-init
based VMs. virsh
is
useful for more detailed libvirt management. They work well together.
virt-install
virt-install
is built for images that use conventional installers. Virter
uses cloud-init
, making it simpler to use and quicker to start a fresh VM.
Running VMs in AWS/GCP/OpenNebula
Virter is local to a single host making snapshot/restore/clone operations very
efficient. Virter could be thought of as cloud provisioning for your local
machine.
Vagrant
Virter and Vagrant have essentially the same goal. Virter is more tightly
integrated with the Linux virtualization stack, resulting in better
snapshot/restore/clone support.
Multipass
Virter and Multipass have similar goals, but Multipass is Ubuntu specific.
Docker
Virter is like Docker for VMs. The user experience of the tools is generally
similar. Docker containers share the host kernel, whereas Virter starts VMs
with their own kernel.
Kata Containers
Virter starts VMs running a variety of Linux distributions, whereas Kata
Containers uses a specific guest that then runs containers.
Weave Ignite
Ignite has very strong requirements on the guest, so it cannot be used for
running standard distributions.
Development
Virter is a standard go project using modules.
Go 1.13+ is supported.