Before starting make sure that Libvirt and KVM are correctly working. The working directory for the following instructions, tested on Fedora 18, is /mnt/veewee/. There I mounted an ext4 filesystem with 30 GByte capacity.

First stop the IPv4 firewall and restart libvirtd afterwards.

service iptables stop
service libvirtd restart

Next create a storage pool. Write the following definition to the file /tmp/pool.xml and create the storage pool with virsh pool-create /tmp/pool.xml. The directory /mnt/veewee/images should exist.

<pool type="dir">
  <name>images</name>
  <target>
     <path>/mnt/veewee/images/</path>
    <format type='qcow2'/>
  </target>
</pool>

Then install necessary packages and continue with the installation of RVM and VeeWee.

yum install libxslt1-devel git patch make \
     bzip2 autoconf automake gcc-c++ readline-devel \
     libyaml-devel libffi-devel openssl-devel bison \
     libtool libvirt-devel
curl -o rvm-installer https://raw.github.com/wayneeseguin/rvm/master/binscripts/rvm-installer
sh rvm-installer
source /etc/profile.d/rvm.sh
rvm install 1.9.2
git clone https://github.com/jedi4ever/veewee.git
cd veewee
gem install bundler
bundle install --without restrictions

Now it is possible to define, customize and build images.

veewee kvm define Fedora-18-x86_64 Fedora-18-x86_64
veewee kvm build Fedora-18-x86_64 --workdir=/mnt/veewee/veewee
veewee kvm halt Fedora-18-x86_64

For customization modify the files in definitions/Fedora-18-x86_64/. After the build completed it is also possible to login with SSH.

On my system to get it running I had to change sda to vda in the file ks.cfg. Also I changed the used mirror in definitions.rb.

When the build finished first stop the instance and then publish the image with python-glanceclient on the image service.

glance image-create --container-format bare \
                    --disk-format qcow2 \
                    --is-public True \
                    --name Fedora-18-x86_64 \
< /mnt/veewee/images/Fedora-18-x86_64.raw

For further details have a look at the following pages.

• https://github.com/jedi4ever/veewee/blob/master/doc/requirements.md
• https://github.com/jedi4ever/veewee/blob/master/doc/installation.md
• https://github.com/jedi4ever/veewee/blob/master/doc/kvm.md
• https://github.com/jedi4ever/veewee/blob/master/doc/customize.md

While using libvirt_cpu_mode=host-model I saw the required CPU flag svm inside the instances. But loading the processor specific KVM module failed with the following messages in the kernel ring buffer.

[  664.051540] has_svm: can't execute cpuid_8000000a
[  664.051553] kvm: no hardware support

After changing the parameter libvirt_cpu_mode to host-passthrough it was possible to load the processor specific KVM module and to use KVM inside the instances.

[    1.585709] systemd[1]: Detected virtualization 'kvm'.
[    5.856891] kvm: Nested Virtualization enabled
[    5.856891] kvm: Nested Paging enabled

Zusammen mit Kurt Garloff von der Telekom zeigen wir heute die gemeinsame Entwicklung und Umsetzung der OpenStack-Umgebung bei der Telekom AG.

Der Vortrag beginnt um 11:30 Uhr im Open Source Forum (Halle 6, Stand F02). Wir freuen uns auf Ihr/euer Erscheinen!

Danach stehen wir auch an unserem Stand (F16/340 in Halle 6) mit mehreren OpenStack-Demos zur Verfügung.

Note: Since this is a talk in German this entry was posted in German.

Die Aufzeichnung des Vortrags steht auf techcast.com zur Verfuegung.

[VLANS]
tenant_network_type = vlan
network_vlan_ranges = physnet0:100:200,physnet0:500:550

[LINUX_BRIDGE]
physical_interface_mappings = physnet0:eth0,physnet1:eth1