Setting up KVM to PXE boot virtual machines from a local TFTP server

I had occasion to test some failures that a partner is seeing in the field when trying to pxeboot the Ubuntu Server installer.  I’d never set up KVM to do this before, as I almost exclusively use MAAS to do all server deployments, and in fact, this is required for my work.  The partner is doing this as a side project and it seemed like a nice way to waste an afternoon.  The following is somewhat specific to my desktop running Ubuntu 18.04 LTS.  16.04 systems require a few extra steps to get a qemu that supports PXE booting.

On Ubuntu 16.04.x qemu-kvm from the Pike Ubuntu Cloud Archive (UCA) need to be installed.
The Pike Cloud Archive can be enabled like this:

$ sudo add-apt-repository cloud-archive:pike
$ sudo apt update

If you have 16.04 LTS,  after adding the cloud-archive repo and updating, proceed as you would for 17.04 and later:

$ sudo apt install -qq -y libvirt-daemon-system qemu-kvm virt-manager

Next, we need to get tftp installed and verify it’s running.

$ sudo apt install tftpd-hpa
$ sudo service tftpd-hpa status
● tftpd-hpa.service - LSB: HPA's tftp server
Loaded: loaded (/etc/init.d/tftpd-hpa; generated)
Active: active (running) since Thu 2018-09-13 09:35:49 EDT; 5s ago
Docs: man:systemd-sysv-generator(8)
Process: 16715 ExecStop=/etc/init.d/tftpd-hpa stop (code=exited, status=0/SUCC
Process: 16720 ExecStart=/etc/init.d/tftpd-hpa start (code=exited, status=0/SU
Tasks: 1 (limit: 4915)
CGroup: /system.slice/tftpd-hpa.service
└─16745 /usr/sbin/in.tftpd --listen --user tftp --address :69 --secur

Sep 13 09:35:49 galactica systemd[1]: Starting LSB: HPA's tftp server...
Sep 13 09:35:49 galactica tftpd-hpa[16720]: * Starting HPA's tftpd in.tftpd
Sep 13 09:35:49 galactica tftpd-hpa[16720]: ...done.
Sep 13 09:35:49 galactica systemd[1]: Started LSB: HPA's tftp server.

Now we need to start setting up the tftpboot directories.  I’m using /srv/tftp to host the files:

$ sudo mkdir /srv/tftp
$ cd /srv/tftp
$ $ sudo wget -nH -r --cut-dirs=8 http://archive.ubuntu.com/ubuntu/dists/bionic-updates/main/installer-amd64/current/images/netboot/

This will also leave a lot of cruft in the directory, perhaps that’s fixable with other wget options but this gets the files and directories you need.  You can clean up the cruft

$ sudo rm -f *.gif index.html* MANIFEST* MD5SUMS* SHA*
$ ls
boot.img.gz ldlinux.c32 pxelinux.0 udeb.list
debian-cd_info.tar.gz mini.iso pxelinux.cfg vmlinuz
initrd.gz netboot.tar.gz ubuntu-installer xen

Now we need to set permissions and ownership of the files:

/srv/tftp$ cd ../
/srv$ sudo chmod -R 777 tftp/
/srv$ sudo chown -R nobody: tftp/

Next, make sure TFTPD can see the right directory by pointing

$ cat /etc/default/tftpd-hpa
# /etc/default/tftpd-hpa

TFTP_USERNAME="tftp"
TFTP_DIRECTORY="/srv/tftp"
TFTP_ADDRESS=":69"
TFTP_OPTIONS="--secure --create"

Finally restart tftpd and verify it’s listening on the correct IP address (the gateway address for KVM’s bridge)

$ nc -uvz 192.168.123.1 69
Connection to 192.168.123.1 69 port [udp/tftp] succeeded!

This means that tftpd is active and listening on port 69 on my libvirt network (my ip addressing may be different from yours).  Now there’s a couple things to set for libvirt that will allow the VMs to grab the right boot files.  Add the lines (in bold) to the config:

$ virsh net-edit default
<network>
  <name>default</name>
  <uuid>c1ea51e9-ac51-455b-a6ff-7a222b6f94eb</uuid>
  <forward mode='nat'/>
  <bridge name='virbr0' stp='on' delay='0'/>
  <mac address='52:54:00:42:de:cf'/>
  <ip address='192.168.123.1' netmask='255.255.255.0'>
    <tftp root='/srv/tftp'/>
    <dhcp>
      <range start='192.168.123.2' end='192.168.123.254'/>
      <bootp file='pxelinux.0' server='192.168.123.1'/>
    </dhcp>
  </ip>
</network>

Again, note that my IP addressing may be different.  Once the config has been modified to serve up PXE files, restart the virtual network:

$ virsh net-destroy default
Network default destroyed

$ virsh net-start default
Network default started

Now let’s create a VM to PXEBOOT the installer.  I’ll be using virt-manager for this, as when dealing with qemu, I prefer the simplicity of a GUI over remembering a large swath of command line arguments.

Image of the initial config screen

This is the first config screen. Be sure to select PXE boot.

Image of the VM OS template setting screen

The Generic settings are fine here.

Image of the VM Memory and vCPU config screen

I upped the RAM to 2Gb just to make it a little sturdier.

Image of the VM Config Disk Setup Screen

20Gb should be enough for this experiment.​

Image of the VM Config Summary

Summary Screen: Be sure that the “Default” network has been chosen

Image of the bootup screen showing the VM PXE Booting

The VM is now PXE booting

Image of the netboot installer menu

The VM has now successfully booted via PXE

Bookmark the permalink.

Leave a Reply

Your email address will not be published. Required fields are marked *