generate a kvm environment on vm

• login on vm
• edit /srv/kvm/config/imagename.conf and add this or similar settings
  • NIC1=int
  • NIC2=ext
  • NIC3=aei
• if you edit nothing NIC1 is automatically int and NIC2 and NIC3 are not used
• start /usr/local/sbin/kvmrun imagename
  • If no screen is running and if no image is present or if the image is too small for an installation it creates an image, a tunnel interface and starts a kvm with network boot option to make a FAI installation. After the FAI installation the KVM goes down and restarts with the harddrive boot option.
  • If no screen is running but an image with a proper size exists the kvm starts booting this image in a screen.
  • If a screen is running, kvmrun attaches to this screen.

FAI installation

A newly created hard drive image will be filled with a FAI installation. The first KVM start uses the network boot option. There a mac-address appears which was determined by kvmrun. Put the mac-address into the DHCP server and give the node a name, e.g. imagename. Write a proper PXE configuration for the FAI installation. Tell the FAI server that this node belongs to the class KVM.

what kvmrun does

• it creates two tunnel intrefaces, if not present: tap_int_imagename and tap_ext_imagename and computes mac addresses using a hash algorithm
• Creates, if necessary, an qcow2 image which grows by demand. The name is _/srv/kvm/images/imagename.img.
• Creates or attaches to a screen with a session name imagename.

-- HenningFehrmann - 12 Jan 2011