A while ago I was explaining the difference between QEMU, KVM and LibVirt, and I ended up by emailing this nonsense. I don’t claim it’s accurate, it certainly isn’t. It’s probably not even funny. Enjoy :).

In the beginning there was QEMU, but it was slow and the people grieved.

Then KVM was forked from QEMU with a kernel module to use the CPU’s virtualisation features to work much faster and there was much rejoicing. Linus also rejoiced and welcomed KVM’s kernel module into the mainline kernel.

But the people did not rejoice, as they were mostly using Sun’s VirtualBox (also forked from QEMU).

QEMU awoke from its slumber and joined with KVM and their union caused almost no rejoicing, in fact I am not convinced anyone really noticed.

But the System Administrators were still dissatisfied and complained that there should be standardisation of commands across different hypervisors. And thus LibVirt was born and the System Administrators rejoiced.

Yet still the people used VirtualBox. But lo! The evil Oracle slew the Sun and VirtualBox moved into darkness, and there was much gnashing of teeth and wearing of sackcloth; although this was generally considered a step forwards from the t-shirts that they usually wore.

But still the people could not use QEMU-KVM, without issuing complex incantations, and so Virt-Manager was born and finally the people rejoiced, with much clicking of mice.

The End.

• Virtualization With KVM On A Debian Squeeze Server

I have used KVM, but never with libvirt.  I understand that libvirt is a standard api for managing various virtualisation technologies, e.g. KVM, QEMU, Xen and others, giving you a common way of managing your guests regardless of the technology in use.

For more information please visit:

• Virtualization With KVM On A Debian Squeeze Server

This is just an aide-memoire, please don’t try and follow it.

Creating a template

# rm -f /etc/ssh/ssh_host_*
cat << EOF > /etc/rc2.d/S15ssh_gen_host_keys
#!/bin/bash
ssh-keygen -f /etc/ssh/ssh_host_rsa_key -t rsa -N ''
ssh-keygen -f /etc/ssh/ssh_host_dsa_key -t dsa -N ''
rm -f \$0
EOF
# chmod a+x /etc/rc2.d/S15ssh_gen_host_keys

Tip re. freenx-server

You will need to reconfigure the freenx-server:

$ sudo dpkg-reconfigure freenx-server

Updating template

$ sudo vzctl enter XXX
$ sudo apt-get clean
$ exit
$ sudo vzctl set XXX --ipdel all --save
$ sudo vim /var/lib/vz/private/XXX/etc/resolv.conf
$ sudo rm -f /var/lib/vz/private/XXX/etc/hostname
$ sudo vzctl stop XXX
$ cd /var/lib/vz/private/XXX
$ tar --numeric-owner -zcf /var/lib/vz/template/cache/debian-5.0-i386-minimal.tar.gz .

Create configuration template for 10 vms

$ sudo vzsplit -n 10 -f vps.tenth

Create new vm

sudo vzctl create XXX --ostemplate debian-5.0-i386-minimal --config vps.basic
sudo vzctl create XXX --ostemplate debian-5.0-i386-kde --config vps.tenth
sudo vzctl start XXX
sudo vzctl set XXX --ipadd 192.168.0.XXX --save
sudo vzctl set XXX --nameserver 192.168.0.254 --save
sudo vzctl set XXX --onboot yes --save
sudo vzctl set XXX --hostname server-XXXXXX.example.co.uk --save
sudo vzctl set XXX --diskspace $(( 1048576*2 )):$(( 1153434*2 )) --save
sudo vzctl set XXX --diskspace 10G:11G --save
vzcfgvalidate /etc/vz/conf/XXX.conf

Other tricks

sudo vzctl start XXX
sudo vzctl exec XXX passwd
sudo vzctl exec XXX ps aux
sudo vzctl enter XXX
sudo vzctl stop XXX

Removing a vm

$ sudo vzctl destroy XXX
$ sudo rm /etc/vz/conf/XXX.conf.destroyed

List running and non-running

sudo vzlist -a

Memory use

Check memory usage for guest 103

$ sudo vzcalc -v 103

Set minimum memory 256 x memory wanted

$ sudo vzctl set vpsid --vmguarpages $((256 * 256)) --save

Set maximum memory 256 x memory wanted

$ sudo vzctl set vpsid --privvmpages $((256 * 1024)) --save

References

• http://wiki.openvz.org/Debian_template_creation
• http://wiki.openvz.org/Resource_shortage
• http://www.linux.com/archive/feature/114214

After a considerable time spent googling without any success, I decided to upgrade the Kernel from 2.6.26 to the backports 2.6.32 and normal operation was restored. I mention this only in the hope that the next person who googles qemu-kvm black screen might be saved a considerable time searching for the solution!

Solving VirtualBox Mismatch

If you install virtualbox-ose under Debian, along with virtualbox-ose-modules-2.6, there can be a mismatch between versions, which will prevent VirtualBox from running. At the time of writing virtualbox-ose is at version 1.66 and virtualbox-ose-modules-2.6 is at 1.62.

Unfortunately you need to build the kernel module yourself.

Step 1: Determine modules package name

$ dpkg -l "virtualbox-ose-modules*" | grep ii
ii  virtualbox-ose-modules-2.6.30-2-amd64 3.0.12-dfsg-1+2.6.30-8squeeze1 VirtualBox modules for Linux (kernel 2.6.30-

Take a note of this package name for the next step.

Step 2: Uninstall the Debian package

$ sudo apt-get remove virtualbox-ose-modules-2.6.30-2-amd64

Step 3: Install VirtualBox Source

$ sudo apt-get install virtualbox-ose-source
$ sudo module-assistant auto-install virtualbox-ose

This will produce a custom virtualbox-ose-modules package in /usr/src, which will then be used to install.

Look out for error messages, if it states that a deb file already exists, remove it and try again.

Step 4: Load Kernel Driver

This step was carried out automatically in the version that I have.

$ sudo modprobe -r vboxdrv
$ sudo modprobe vboxdrv

The following command will create a resizing 10G image in your current directory:

# /usr/bin/qemu-img create -f qcow2 vdisk.img 10G

Booting from CDROM

I struggled on our LTSP server to boot from CDROM, but found that I could boot easily from the downloaded IMG files (from Microsoft).

Boot from first CD image:

$ kvm -no-acpi -m 512 -cdrom /home/username/vdisks/w2k3/iso/X13-04874.img -boot d -hda w2k3.img -k en-gb -smp 2

Boot from second CD image:

$ kvm -no-acpi -m 512 -cdrom /home/username/vdisks/w2k3/iso/X13-04795.img -boot d -hda w2k3.img -k en-gb -smp 2

References

• http://kvm.qumranet.com/kvmwiki/HOWTO
• http://www.debian-administration.org/articles/545

Download Open Source Edition (OSE) from VirtualBox:

• http://www.virtualbox.org/wiki/Downloads

Pre-requisites

Install the following pre-requisites:

# apt-get install gcc g++ bcc iasl xsltproc xalan libxalan110-dev uuid-dev zlib1g-dev libidl-dev libsdl1.2-dev libxcursor-dev libqt3-headers libqt3-mt-dev libasound2-dev libstdc++5 libhal-dev

Compile

# sudo su
# ./configure
# . env.sh  << Note the space between the '.' and 'env'
# kmk all

Installation

# cd out/linux.x86/release/bin/src
# make
# make install
# cd ..

References

This document has been copied from the following, which you should read as your primary source.

• http://www.virtualbox.org/wiki/Linux%20build%20instructions

I also found the following (suitably adapted for Debian) to be most useful:

• http://www.howtoforge.com/virtualbox_ubuntu

Although I could not get subversion working from behind our proxy server, the source for 1.5.2 was downloadable at the above site.

Licensing

The full product is not open source, only free for personal and evaluation use. They do offer an open source version, but this requires compiling from source and does not include headless VRDP or automated set-up, nor USB support.

Performance

The good news is that VirtualBox is very fast indeed, and you can activate the CPU virtual threading to make it even faster.

Installing the Full Product

So simple, in comparison to the alternatives.

1. Either download from http://www.virtualbox.org/wiki/Downloads or follow instructions to add to apt sources
2. Install in the normal Debian way

Creating a virtual machine

So intuitive, you can work it out for yourself.

VirtualBox requires ISO files. If you have IMG files, they can simply be renamed

Don’t forget to turn on CPU virtual threading, if you have a suitable CPU

Simple networking

See Simple networking in KVM for more information on this method.

Simply add the two scripts below to your .VirtualBox folder (or a place of your choosing) and in the VirtualBox GUI interface, edit the settings of your virtual machine and change the networking to host networking, give the interface the name “tap0” and add the scripts below in the relevant fields.

• Host IP: 192.168.0.17
• TUN Interface IP: 192.168.0.18
• Guest Interface IP: 192.168.0.19

Create interface

#!/bin/sh
# vbox-ifup.sh

# configure tun0 device (UML and newer versions of Qemu use tap0 here!)
sudo /sbin/ifconfig tap0 192.168.0.18

# activate ip forwarding
sudo bash -c 'echo 1 > /proc/sys/net/ipv4/ip_forward'

# set up routing to the guest IP
sudo route add -host 192.168.0.19 dev tap0

# activate ARP proxy to "spoof" arp address
sudo bash -c 'echo 1 > /proc/sys/net/ipv4/conf/tap0/proxy_arp'

# set "spoofed" arp address
sudo arp -Ds 192.168.0.19 eth0 pub

Close interface

#!/bin/sh
# vbox-ifdown.sh

# delete "spoofed" arp address
sudo arp -d 192.168.0.19

# deactivate ARP proxy to "spoof" arp address
sudo bash -c 'echo 0 > /proc/sys/net/ipv4/conf/tap0/proxy_arp'

# remove routing to the guest IP
sudo route del -host 192.168.0.19 dev tap0

# deactivate ip forwarding
sudo bash -c 'echo 0 > /proc/sys/net/ipv4/ip_forward'

# remove tun0 device (UML and newer versions of Qemu use tap0 here!)
sudo /sbin/ifconfig $1 down

References

• Build VirtualBox
• Complete Host Networking
• Install SeamlessRDP
• Configure for seamless and headless operation
• Fix for multiple windows
  • http://ubuntuforums.org/showthread.php?t=224212
  • http://www.fontis.com.au/rdesktop
• VirtualBox Manual
• Concurrent Users
  • http://concurrentremotesessions.netfirms.com/
  • http://sala.pri.ee/terminal-server-patch/

Configuration

• Host IP: 192.168.0.17
• TUN Interface IP: 192.168.0.18
• Guest Interface IP: 192.168.0.19

I went into windows and provisioned the emulated ethernet interface with an IP address of 192.168.0.19; along with appropriate DNS settings and default gateway settings. I’ve also reserved the address 192.168.0.18 for use by the tunnel device.

My /etc/qemu-ifup looks like this:

#!/bin/sh
# configure tun0 device (UML and newer versions of Qemu use tap0 here!)
sudo /sbin/ifconfig $1 192.168.0.18

# activate ip forwarding
sudo bash -c 'echo 1 > /proc/sys/net/ipv4/ip_forward'

# set up routing to the guest IP
sudo route add -host 192.168.0.19 dev tap0

# activate ARP proxy to "spoof" arp address
sudo bash -c 'echo 1 > /proc/sys/net/ipv4/conf/tap0/proxy_arp'

# set "spoofed" arp address
sudo arp -Ds 192.168.0.19 eth0 pub

Thats it… The bridging is done only for a single IP address by installing a proxy ARP.

Running

$ sudo kvm -hda w2k3.img -smp 2 -m 512 -no-acpi -k en-gb -net nic,model=rtl8139 -net tap,script=/etc/qemu-ifup -usb -usbdevice tablet -snapshot

• Note that you must sudo the command, as root priority is required for the qemu-ifup script.
• Note also that for some reason under KVM you need to specify the qemu-ifup script, as it does not default to that. I suspect that it defaults to kvm-ifup, but I have not looked into this further.
• The USB switch is useful in that it enables the mouse to leave the virtual window, and also seems to synchronise the host/guest cursors, which was a problem on our installation.

References

• QEMU
• QEMU Networking
  • QEMU Setup Hints

# egrep '^flags.*(vmx|svm)' /proc/cpuinfo 

If nothing is displayed, then you do not have a CPU that supports KVM. If lines are output that include the text “vmx” then you have an Intel CPU that supports KVM, if “svm” then an AMD CPU that supports KVM.

Check Kernel

In order to run KVM you must be running 2.6.20 or higher. You can check this by running the following command:

# uname -r
2.6.26-1-686

Install Prerequisites

# aptitude install gcc-3.4 libsdl1.2-dev zlib1g-dev libasound2-dev linux-kernel-headers pkg-config libgnutls-dev

Install KVM

Make sure that you have Backports installed if you are running Etch.

# aptitude install kvm qemu

You need QEMU in order to create disk images.

Add User to KVM

# adduser $USER kvm

Load KVM Modules

Decide whether you have an Intel or AMD CPU (see Step 1).

# modprobe kvm
# modprobe kvm-intel or kvm-amd (depending on CPU type)

If this fails then try

# dmesg | grep kvm

If it states “kvm disabled in BIOS” then you will need to boot into the BIOS and switch on VT support (probably under advanced processor options).

You can check that the modules are enabled by running:

# lsmod | grep kvm

Set KVM modules to autoload

Add the modules to /etc/modules to ensure that they load upon the next reboot.

# echo "kvm" >> /etc/modules
# echo "kvm-intel" >> /etc/modules

References

• KVM Wiki Howto
• KVM on Debian Wiki

Options for Windows Virtualisation

There are several options for Windows virtualisation, including:

• QEMU
• Xen
• VirtualBox
• VMWare Server

QEMU

A true open source virtualisation product, but one that runs slowly, often too slowly to be of use. Fortunately you can optimise QEMU by the installation of one of the following:

• KQEMU
• KVM

The former is not entirely open source, but is your main option if you have a Kernel less than 2.6.20.

XEN

Is not an open source product, but is available free. It is considered to be very quick.

VirtualBox

InnoTek offer two versions of their VirtualBox product:

• Full Product - free for Personal and reasonable Evaluation
• OSE (Open Source Edition) - free but requires compiling from source and comes without Guest Additions and without Headless VRDP. The latter seems unimportant, given that Windows has an RDP server already, but it is the only way of running VirtualBox headlessly AFAIK.

I have tested on Debian with great success and the following are the links I used:

• Build VirtualBox
• Complete Host Networking
• Install SeamlessRDP
• Configure for seamless and headless operation
• Fix for multiple windows
  • http://ubuntuforums.org/showthread.php?t=224212
  • https://www.fontis.com.au/rdesktop
• VirtualBox Manual
• Concurrent Users
  • http://concurrentremotesessions.netfirms.com/
  • http://sala.pri.ee/terminal-server-patch/

VMWare Server

Free but not open source. Requires compiling from source. Excellent reliability, supported product. The author has not tested this product.