Virtualization in Grid'5000

See also: KVM | KVM on production environment | KVM contextualization

Purpose

This page presents ways to use KVM on production image. The aim is to permit the execution of virtual machines on the nodes with a "non-deploy" reservation.

First steps

Job submission

In order to test easily the kvm environment, we will use an interactive job.

jdoe@fnancy:~$ oarsub -I

Disk image, virtual machine

A disk image containing debian squeeze is available at the following path: /grid5000/images/KVM/squeeze-x64-base.qcow2

It can be use as a base for more advanced work. For the next steps of this tutorial, copy the disk image to /tmp on the node

jdoe@griffon-37:~$ cp /grid5000/images/KVM/squeeze-x64-base.qcow2 /tmp/

Network configuration

Create Tun/Tap interface for each virtual machines. This virtual interface will be attached to your virtual machine, and bridged on the production network. Therefore, the virtual machine will be able to get an IP from the DHCP server and access the network.

A script is available to create automatically this interface on the node: create_tap:

 jdoe@griffon-37:~$ sudo create_tap

• Tun/Tap interfaces are listed by issuing the command /sbin/ifconfig.

jdoe@griffon-37:~$ /sbin/ifconfig

[...]
tap0      Link encap:Ethernet  HWaddr 00:16:3e:db:c6:41
          inet6 addr: fe80::58ff:a4ff:fe97:c6a8/64 Scope:Link
          UP BROADCAST RUNNING PROMISC MULTICAST  MTU:1500  Metric:1
          RX packets:0 errors:0 dropped:0 overruns:0 frame:0
          TX packets:0 errors:0 dropped:29435 overruns:0 carrier:0
          collisions:0 txqueuelen:500
          RX bytes:0 (0.0 B)  TX bytes:0 (0.0 B)

	Note
	Create one Tun/Tap interface per guest OS.

Generate a random mac address for the Tun/Tap interface

• Use the following script random_mac. This will

generate a random mac address, starting with 00:16:3e.

jdoe@griffon-37:~$ export MACADDR=`random_mac`

	Note
	The first three octets are important (identification for the virtualized network). Don't change it.

Run the guest OS using kvm

Use the kvm command.

jdoe@griffon-37:~$ screen kvm -m 512 -hda /tmp/squeeze-x64-base.qcow2 -net nic,macaddr=$MACADDR -net tap,ifname=tap0,script=no -nographic

	Note
	In this example, tap0 is the name of our Tun/Tap interface. Adapt it with the Tun/Tap name on witch you want to attach your guest OS.

Or, you can start the guest OS using libvirt

• Edit the domain file

eg : domain.xml

 <domain type='kvm'>
  <name>squeeze</name>
  <memory>524288</memory>
  <vcpu>1</vcpu>
  <os>
    <type arch="x86_64">hvm</type>
  </os>
  <clock sync="localtime"/>
  <on_poweroff>destroy</on_poweroff>
  <on_reboot>restart</on_reboot>
  <on_crash>destroy</on_crash>
  <devices>
    <emulator>/usr/bin/kvm</emulator>
    <disk type='file' device='disk'>
      <driver type='qcow2'/>
      <source file='/tmp/squeeze-x64-base.qcow2'/>
      <target dev='vda' bus='virtio'/>
     <shareable/>
    </disk>
    <interface type='ethernet'>
      <target dev='tap0'/>
      <script path='no'/>
      <mac address='00:16:3e:78:b8:04'/>
    </interface>
    <serial type='pty'>
      <source path='/dev/ttyS0'/>
      <target port='0'/>
    </serial>
    <console type='pty'>
      <source path='/dev/ttyS0'/>
      <target port='0'/>
    </console>
  </devices>
 </domain>

• Now, the guest OS can be started.

jdoe@griffon-37:~$ virsh create domain.xml

• You can use virsh to manage your guest OS.

jdoe@griffon-37:~$ virsh list
jdoe@griffon-37:~$ virsh console squeeze

	Note
	- You can use the script /usr/local/bin/random_mac to generate a mac address. - Use CTRL+AltGr+] to disconnect from virsh console

Misc notes

• Use KVM contextualization, if you want to perform some customization (like using static address) for your VMs.
• For the deployment of an important number of guest OS, you can use the subnet reservation. See also the virtual network interlink.
• Manage the console for all your guest OS with Screen
• Account: root:grid5000
• Use tunctl if you need to delete a Tun/Tap device.

jdoe@griffon-37:~$ /usr/sbin/tunctl -d tap0

• Some clusters don't have virtualization support in the BIOS. To check :

jdoe@griffon-37:~$ egrep '^flags.*(vmx|svm)' /proc/cpuinfo