Figuring out virtualization on Linux - Part 1, Basics, Setting up a VM.

Not so much a guide as it is a learning-log of sorts.

A head-empty-no-thoughts guide is shortlinked here.

Figuring out basics of virtualization

The goal is to understand at least a little bit of what I do.

Things I am aware of before starting this trip:

1. Virtualization can be achieved with VMWare, Virtualbox (as I have used in the past), but solutions like Hyper-V and KVM on Windows and Linux respectively offer much more in terms of hardware (CPU) acceleration.
2. I want to set it up for experimentation purposes, because computers are cool and networking seems cool and I want to explore more of both. I’m also really fascinated by DevOps and realistically will have to interact with it often enough, and want to know what it is instead of pressing random buttons prescribed by obscure internal documentation.
3. There’s more than just KVM available, but it seems like a well-supported, popular option.
4. If I want to use KVM, I will probably want to use QEMU, libvirtd, virt-manager?, virsh?, and ???.

Now the Arch Wiki page is rich in detail but not in explanation, so that won’t serve as a good starting point. Thankfully, I found octetz’ intro-level guide to virtualization with KVM and Veronica Cary’s QEMU/KVM for absolute beginners. These videos provide both a high-level overview of what is happening amongst all the moving parts, and a step-by-step guide to setting up the tools.

At this point, I know:

1. KVM is the hypervisor we will be using. It’s baked into the kernel.

2. QEMU is the actual virtual machine emulator. It emulates a bunch of things and interacts with KVM (although it can work without it too, slower).

   I am not aware of how it’s handling emulation, exactly. Storing the documentation here, for future reference.

3. libvirtd is a generic virtualization API that works on top of several technologies, that we will use for working with KVM/QEMU.
4. virt-manager/virsh are front-end applications that interact with the libvirt API when required. virt-manager also comes with virt-install and virt-viewer.
5. It seems that for networking, bridge-utils is used. I am not too clear on networking yet, however.
6. libvirtd has a concept of “system” and “session” instances. As a quick summary: in system mode, libvirtd is running as root and has more access to things in general. For a less hacky explanation, this post seems appropriate.
   • On this note, virt-manager uses system mode by default, but virsh uses user mode. We will need to configure virsh to use system mode by default, as that’s what we’re using to begin with. Why? Because that’s what the videos I saw said would be fine most of the time. I’ll make my own decisions once I’m more aware. The tradeoff seems to be worse desktop-use-case integration for better networking (with system).

Alright, it’s now time to set it up.

Setting up virt-manager and virsh to work with libvirtd, KVM, and QEMU

Ensure KVM will work

First, we need to ensure that KVM is set up and ready. I ran this command from the Arch wiki:

$ LC_ALL=C lscpu | grep Virtualization
Virtualization:                  AMD-V

I have an AMD CPU, and this confirms that KVM is ready.

Installing everything else

Again, I’m not sure exactly what is needed for the task, but this should suffice. I’m using pacman, adjust as necessary.

$ sudo pacman -Syu \
  qemu \
  qemu-arch-extra \
  libvirt \
  lxc \
  arch-install-scripts \
  iptables-nft \
  dnsmasq \
  bridge-utils \
  openbsd-netcat \
  virt-viewer \
  virt-manager \
  virt-install \
  dmidecode \
  dhclient

Things I am sure we need: qemu, libvirt, lxc, bridge-utils, virt-manager. I do not know about the rest.

Setting up permissions

From what I can tell, this has primarily two parts:

1. Configuring virsh to work with qemu:///system by default

   sudo cp -rv /etc/libvirt/libvirt.conf ~/.config/libvirt/
   sudo chown zubair:wheel ~/.config/libvirt/libvirt.conf
   

   Then, edit the file and uncomment the last line to say: uri_default = "qemu:///system"

2. Ensuring that the user has permissions to interact with libvirtd

   Here, the octetz blog I’m using specifies three ways, mentioning that if the user is in an administrator group (like wheel in arch), they probably won’t have to worry. I’m really not clear though – the wiki mentions that a password prompt is necessary, but the video/post does not allude to it. I have decided to take the “safe” option that is mentioned and add my user to the libvirt group.

   sudo gpasswd -a zubair libvirt
   

   At this point, I had to restart my computer to make networking work. Maybe not had to, but I did it anyway.

Creating a VM

I’m making an Elementary VM, using virt-manager. Easy enough: ensure libvirtd is active with sudo systemctl start libvirtd, start up virt-manager, create a new VM with an elementary ISO I have lying about. Allocate 2GB memory, 2 cores of CPU, and 20GB storage. It’s just an experimental one anyway. The storage is in my root partition, which is a bit smaller than I’d like it to be given that I now know VMs are installed to root, by default. I’ll check if I can install to home instead, while still being a qemu:///system instance.

Anyway, installation was painless.

I’m now trying the /home install with ubuntu server, and also trying to install it with virt-install instead of directly through virt-manager’s UI.

virt-install \
  --name ubuntuserver \
  --ram 2048 \
  --disk path=/home/zubair/Images/ubuntu.qcow2,size=16 \
  --vcpus 2 \
  --os-type linux \
  --os-variant generic \
  --console pty,target_type=serial \
  --cdrom /home/zubair/Downloads/ISOs/ubuntu-20.04.3-live-server-amd64.iso

I keep getting “failed to unmount cdrom”.

But it works. Installed, restarted, and logged in using SSH.

What if I want to use virt-manager with a custom storage for the image?

• Right click on the “QEMU/KVM” that should be right on top
• Click on “Details”
• Go to the “Storage” tab
• Add a pool on the left column, with the “+”. Call it what you will, and set the path to wherever you wish.
• Inside the pool, Click “+” on the “Volumes” option.
• Name it, give it the qcow2 format, and allocate the capacity.
• When installing the system, in step 4, select custom storage. Alternatively, you should be able to do all the previous steps in this step itself.

And that’s it for part one!

The no-thoughts-head-empty guide

I do not recommend this, but it should be possible to set up the whole system without knowing anything you’re doing. Kinda like VMWare, really.

1. Check that hardware virtualization is enabled:

   LC_ALL=C lscpu | grep Virtualization
   

   It should return “AMD-V” or “VT-x”. If it doesn’t, check online to see if your system can enable hardware virtualization.

2. Install everything:

   $ sudo pacman -Syu \
     qemu \
     qemu-arch-extra \
     libvirt \
     lxc \
     arch-install-scripts \
     iptables-nft \
     dnsmasq \
     bridge-utils \
     openbsd-netcat \
     virt-viewer \
     virt-manager \
     virt-install \
     dmidecode \
     dhclient
   

3. Setup permissions:

   sudo cp -rv /etc/libvirt/libvirt.conf ~/.config/libvirt/ && \
   sudo chown {user}/{group} ~/.config/libvirt/libvirt.conf && \
   echo "uri_default = \"qemu:///system\"" >> ~/.config/libvirt/libvirt.conf &&\
   sudo gpasswd -a {user} libvirt
   

   You do need some thoughts here, replace {user} and {group} with what’s relevant. For Arch users, {group} should be wheel.

And installation should be done! Restart your system, maybe.

To start libvirtd, run this:

sudo systemctl start libvirtd

If you want it to run every time you start your system, make that

sudo systemctl enable libvirtd

<-- Back