h1. Libvirt

h2. Resources/Liens

* Virtualisation du routage IPv4 pour les VM http://lists.tetaneutral.net/pipermail/technique/2013-April/000810.html

* libvirt

http://builder.virt-tools.org/artifacts/libvirt-virshcmdref/html/index.html

http://docs.fedoraproject.org/en-US/Fedora/13/html/Virtualization_Guide/chap-Virtualization-Managing_guests_with_virsh.html

* virt-install et preseed sans server web et pxe:

http://honk.sigxcpu.org/con/Preseeding_Debian_virtual_machines_with_virt_install.html

h2. Installation VM sans bridge avec routage static (en /32 pour l'ipv4)

Note: tous les fichiers utilisés et créés ne sont peut être pas à jour, voir ici pour les dernières version: http://git.tetaneutral.net/libvirt.git/

Le répertoire /etc/libvirt/ttnn cité plustard est géré dans ce git.

h3. Configuration de la libvirt

On créé un réseau libvirt qui fait rien (déjà fait sur n7):

<pre>

cat > libvirt_dump_network.xml <<EOF

<network ipv6='yes'>

<name>dump-net</name>

<bridge name='dumpbr0' />

</network>

EOF

virsh net-define libvirt_dump_network.xml

virsh net-autostart dump-net

virsh net-start dump-net

</pre>

On créé un hook libvirt (déjà fait sur n7):

<pre>

mkdir -p /etc/libvirt/ttnn/hooks

ln -s /etc/libvirt/ttnn/hooks /etc/libvirt/hooks

vi /etc/libvirt/hooks/qemu

----8<--------------->8----

#!/bin/bash

name="$1"

operation="$2"

suboperation="$3"

[ "$operation" != "start" ] && exit 0

[ "$suboperation" != "begin" ] && exit 0

# get interface from the libvirt xml

interface=$(cat - | sed -ne "/vnet/s/.*dev='\([^']*\)'.*/\1/gp")

[ -f /etc/libvirt/ttnn/${name}.conf ] && source /etc/libvirt/ttnn/${name}.conf

[ "$ip" -o "$ip6" ] && brctl delif dumpbr0 $interface

[ "$ip" ] && ip route add $ip/32 dev $interface

[ "$ip6" ] && ip route add $ip6/56 dev $interface

[ "$ip6" ] && ip -6 addr add fe80::31/64 dev $interface

exit 0

----8<--------------->8----

</pre>

h3. Preparation de l'installer

Pour une squeeze on met à jour virtinst  (déjà fait sur n7):

<pre>

wget http://ftp.us.debian.org/debian/pool/main/v/virtinst/virtinst_0.600.3-3_all.deb

dpkg -i virtinst_0.600.3-3_all.deb

</pre>

On install le preseed.cfg attaché à la page dans /etc/libvirt/preseed.cfg  (déjà fait sur n7)

h4. Installation d'une VM

Creation du disk (déjà fait sur n7):

<pre>

lvcreate -n picttn -L 50G /dev/raid5

</pre>

Configuration du routage static:

<pre>

vi /etc/libvirt/ttnn/picttn.conf 

# le /32 ipv4

ip="91.224.148.252"

# le /56 ipv6

ip6="2a01:6600:8080:FC00::"

</pre>

Lancement de l'installation via le réseau natter de libvirt

<pre>

virt-install -n picttn -r 2048 --os-type=linux \

    --autostart \

    --disk /dev/raid5/picttn,bus=virtio,cache=none,io=native \

    --network network=default,model=virtio \

    --nographics \

    --noreboot \

    --wait -1 \

    --location http://ftp.debian.org/debian/dists/wheezy/main/installer-amd64/ \

    --serial pty \

    --initrd-inject /etc/libvirt/ttnn/preseed.cfg \

    --initrd-inject /etc/libvirt/ttnn/picttn.conf \

    --initrd-inject /etc/libvirt/ttnn/postinstall.sh \

    --extra-args "console=ttyS0,9600n8 DEBIAN_FRONTEND=text auto priority=critical hostname=picttn domain=le-pic.org"

</pre>

Une fois l'installation terminée, on change la conf réseau:

<pre>

EDITOR="sed -ie s/default/dump-net/g" virsh edit picttn

</pre>

On peux démarrer la VM

<pre>

virsh start picttn

virsh console picttn

</pre>

Ensuite on peux se connecte à la VM

<pre>

ssh root@2a01:6600:8080:FC00::1

</pre>

Le /etc/network/interfaces IPv4 après installation

<pre>

iface eth0 inet static

    address 91.224.148.xxx

    netmask 32

    up ip route add 91.224.148.0/32 dev eth0

    up ip route add default via 91.224.148.0

</pre>

h2. Misc

<H4Mm3r> si tu lance la création de la VM sans avoir créé le disk --disk /vmroot/testnat.qcow2,format=qcow2,size=12,bus=virtio,cache=writeback

<H4Mm3r> le disk se créé mais dans le XML le format est à RAW

<H4Mm3r> si tu le créé à l'avance avec qemu-img create -f qcow2

<H4Mm3r> et que tu lance ta commande --disk /vmroot/testnat.qcow2,format=qcow2

<guerby> ok donc c'etait bien ça :)

<guerby> bbl (miam)

<H4Mm3r> dans le fichier XML le format est à RAW

<H4Mm3r> il faut en effet dumper un XML, modifier la valeur à QCOW2

<H4Mm3r> et créer la VM depuis ce XML

<H4Mm3r> virsh create file.xml

<H4Mm3r> pourqu'enfin il comprenne que c'est du QCOW...

<H4Mm3r> et là mon disk fait bien 265ko, mais est détecté comme 12G :D

h2. HOWTO by H4Mm3r (under contruction)

h3. Usefull links

#### Misc

#OS Type

http://rhevdup.blogspot.fr/2012/05/list-available-os-types-for-kvm-on-rhel.html

#Libvirt TTN

http://chiliproject.tetaneutral.net/projects/tetaneutral/wiki/Libvirt

#### KVM management

#Install QEMU KVM + LIBVIRT

http://www.howtoforge.com/virtualization-with-kvm-on-a-debian-squeeze-server

#Edit XML file

http://vladnevzorov.com/2011/02/11/how-to-change-configuration-of-kvm-virtual-machine/

#Clone a KVM

https://help.ubuntu.com/community/KVM/CreateGuests#Cloning_a_virtual_machine

http://linux.die.net/man/1/virt-clone

#### Network

#NAT

http://www.linuxquestions.org/questions/linux-networking-3/kvm-qemu-and-nat-on-the-host-machine-mini-tutorial-697980/

http://wiki.libvirt.org/page/Networking

http://libvirt.org/formatnetwork.html#examplesNAT

https://help.ubuntu.com/community/KVM/Networking

#Bridge

http://toast.djw.org.uk/qemu.html

http://www.mediaonfire.com/en/tech_tips/QEMUNetworking.html (bridge, nat, arp)

#### Improve performance

http://blog.bodhizazen.net/linux/improve-kvm-performance/

http://www.linux-kvm.org/page/Tuning_KVM

http://www.ezunix.org/index.php?title=Boost_performance_of_KVM_guests

#### Storage

http://doc.opensuse.org/products/draft/SLES/SLES-kvm_sd_draft/cha.libvirt.storage.html

h3. HOWTO

<pre>

[H] = Host

[G] = Guest

[XML] = XML content

"command"

</pre>

#### Diskfile Storage

#Create qcow2 disk

<pre>

[H] "qemu-img create -f qcow2 /path/to/disk.qcow2 $SIZE" (ex: 20G)

[H] "chmod +x /path/to/disk.qcow2"

</pre>

#Convert raw to qcow2

<pre>

[H] "qemu-img convert -O qcow2 disk.img disk.qcow2"

</pre>

#Disable cache to improve performance 

<pre>

[XML] <driver name='qemu' type='qcow2' cache='none'/>

</pre>

#### LVM Storage

<pre>

[H] Create "/etc/libvirt/storage/vmdata.xml" file :

      <pool type="logical">

        <name>my_pool</name>

        <source>

          <device path="/path/to/dev"/>

        </source>

        <target>

          <path>/target/path/path> (LV will be created there)

        </target>

      </pool>

[H] "virsh pool-define /etc/libvirt/storage/my_pool.xml"

[H] "virsh pool-start my_pool"

[H] "virsh pool-autostart my_pool"

[H] "virsh pool-info my_pool"

[H] "virsh vol-create-as my_pool my_logical_volume 12G"

[H] You must create a LV to be able to give it to a KVM "virsh vol-create-as my_pool my_vol 12G --format=lvm2"

</pre>

h3. Create KVM

#Bridged with TAP + LVM

<pre>

[H] "virt-install -d --hvm --vnc --name=my_kvm --vcpus=4 --cpuset=0-4 --ram 2048 --disk vol=my_pool/my_vol,bus=virtio,cache=none --network bridge=my_bridge,model=virtio -c /path/to/distrib.iso --keymap=fr --os-type=linux --noautoconsole"

</pre>

#TAP (if you want to rename the default "vnetX" interface by "tapX"

<pre>

[H] "virsh dumpxml my_guest > my_guest.xml

[H] "virsh undefine my_guest"

[H] Edit the my_guest.xml file to rename VNETX by TAPX

[H] "virsh define my_guest.xml"

[H] "virsh start my_guest"

</pre>

#NATed + LVM

<pre>

[H] "virt-install -d --hvm --vnc --name=my_kvm --vcpus=4 --cpuset=0-4 --ram 2048 --disk vol=my_pool/my_vol,bus=virtio,cache=none --network network=my_network,model=virtio -c /path/to/distrib.iso --keymap=fr --os-type=linux --noautoconsole"

</pre>

#Clone VM

<pre>

[H] Create the LVM dest

[H] Pause the KVM "virsh suspend my_kvm"

[H] virt-clone -o my_orig_kvm -n my_cloned_kvm -f=/path/to/lvm/dev

[G] Go and change the hostname and resolv.conf

</pre>

h3. Command explanation

#Bridged with tap

<pre>

virt-install -d --hvm --vnc --name=testlvm \

    --vcpus=4 --cpuset=0-4 \

    --ram 2048 \ 

    --disk vol=vmdata/testlvm,bus=virtio,cache=none \

    --network bridge=br0,model=virtio \

    -c /data/isos/debian-testing-amd64-netinst.iso \

    --keymap=fr --os-type=linux --os-variant=debiansqueeze --noautoconsole

</pre>

#NATed

<pre>

virt-install -d --hvm --vnc --name=testlvm \

    --vcpus=4 --cpuset=0-4 \

    --ram 2048 \ 

    --disk vol=vmdata/testlvm,bus=virtio,cache=none \

    --network network=frp_network,model=virtio \

    -c /data/isos/debian-testing-amd64-netinst.iso \

    --keymap=fr --os-type=linux --os-variant=debiansqueeze --noautoconsole

</pre>

[OLD QCOW2]

<pre>

[H] "virt-install --connect qemu:///system -n puppetnode1 -r 1024 --vcpus=4 --cpuset=4-7 --disk path=/vmroot/puppetnode1.qcow2 -c /data/isos/debian-6.0.6-amd64-netinst.iso --vnc --keymap=fr --noautoconsole --os-type linux --os-variant debiansqueeze --accelerate --network network=default --hvm"

[H] "virt-install --connect qemu:///system -n testnat -r 1024 --vcpus=2 --cpuset=2-3 --disk /vmroot/testnat.qcow2,format=qcow2,size=12,bus=virtio,cache=writeback -c /data/isos/debian-6.0.6-amd64-netinst.iso --vnc --keymap=fr --noautoconsole --os-type linux --os-variant debiansqueeze --accelerate --network bridge=br0 --hvm"

</pre>

#Start a KVM

<pre>

[H] "virsh start my_kvm"

</pre>

#Stop node

<pre>

[H] Soft : "virsh shutdown my_kvm"

[H] Hard : "virsh destroy my_kvm"

</pre>

#Suppr node del

<pre>

[H] "virsh undefine node" (+ remove files)

</pre>

#Rename node

<pre>

[H] "virsh dumpxml my_kvm > foo.xml"

[H] <edit foo.xml, change the name, move storage>

[H] "virsh undefine my_kvm"

[H] "virsh define foo.xml"

</pre>

# Connect to virsh qemu

<pre>

[H] "virsh --connect qemu:///system"

[H] "virsh # help"

</pre>

#Start a network

<pre>

[H] "virsh net-start my_network"

</pre>

#Configure VNC

<pre>

[H] Edit /etc/libvirt/qemu.conf and set vnc_listen = "your.ip.addd"

[H] You need to restart service "/etc/init.d/libvirtd restart"

[H] "ps aux | grep -i kvm" (find vnc display for your KVM)

</pre>

#Connect to a KVM through VNC

<pre>

tightvncviewer -compresslevel 7 -encodings "copyrect tight hextile zlib corre rre raw" x.xxx.xxx.xxx:x

</pre>

#Keyboard FR for KVM

<pre>

[H] "virsh edit my_kvm"

[XML] <graphics type='vnc' port='-1' autoport='yes' keymap='fr'/>

</pre>

#Create your own NAT network

<pre>

[H] Create your_nat.xml file :

[XML] <network>

  <name>frp_network</name>

  <forward mode='nat'/>

  <bridge name='br1' stp='on' delay='0' />

  <ip address='10.16.42.1' netmask='255.255.255.0'>

    <dhcp>

      <range start='10.16.42.2' end='10.16.42.50' />

    </dhcp>

  </ip>

</network>

[H] "virsh net-define your_nat.xml"

[H] "virsh net-start your_nat.xml"

[H] "virsh net-autostart your_nat.xml"

[XML] Network type "interface type = network, source network=your_nat"

</pre>

#IBM advisements to improve performance

<pre>

[H] Use LVM (or other direct access to HDD) for VMs storage (from 45 to 3' to install a VM)

[H] Set the swappiness value to 0 ("echo 0 > /proc/sys/vm/swappiness" && edit "/etc/sysctl.conf" and add "vm.swappiness=0")

[H] Disable zone_reclaim ("echo 0 > /proc/sys/vm/zone_reclaim_mode" && edit "/etc/sysctl.conf" and add "vm.zone_reclaim_mode=0")

[H] Enable KSM ("echo 1 > /sys/kernel/mm/ksm/run") and set the check time to 200ms ("echo 200 > /sys/kernel/mm/ksm/sleep_millisecs")

</pre>