Kickstart Infrastructure SOP

From FedoraProject

Revision as of 20:51, 17 February 2009 by Toshio (Talk | contribs)

Jump to: navigation, search
Infrastructure InfrastructureTeamN1.png

kickstart scripts provide our install infrastructure. We only have a handful of different kickstart scripts, those in a xen guest and the xen box itself.


Contact Information

Owner: Fedora Infrastructure Team

Contact: #fedora-admin, sysadmin-main

Location: Phoenix

Servers: puppet1 (stores kickstarts and install media)

Purpose: Provides our install infrastructure


Our kickstart infrastructure lives on the proxy servers and puppet1. All install media and kickstart scripts are located on puppet1. Because the RHEL binaries are not public we have these bits blocked. You can add your bits to (from puppet1):


Physical Machine (dom0)

Xen Dom0 installs are far riskier then the DomU installs below. This is because if an install goes bad, your options to rebuild it are somewhat limited.

PXE Booting

If PXE booting just follow the prompt after doing the pxe boot (most hosts will pxeboot via console hitting f12.


This only works on an already booted box, many boxes at our colocations may have to be rebuilt by the people in those locations first. Also make sure the IP you are about to boot to install from is allowed to our IP restricted as noted above (in Introduction).

Download the vmlinuz and initrd images.

wget \
    -O /boot/vmlinuz-install
wget \
    -O /boot/initrd-install.img

grubby --add-kernel=/boot/vmlinuz-install \
       --args="ks= \
       method= \
       ksdevice=link ip=$IP gateway=$GATEWAY netmask=$NETMASK dns=$DNS" \
       --title="install el5" --initrd=/boot/initrd-install.img
The default here is to use the xen-host-nohd config which requires you to connect via VNC to the box and configure its drives. If this is a new machine or you are fine with blowing everything away, you can instead use as your kickstart

Double and triple check your configuration settings (cat /boot/grub/menu.lst), especially your IP information. In places like ServerBeach not all hosts have the same netmask or gateway. Once everything is ready run:

echo "savedefault --default=0 --once" | grub --batch
shutdown -r now


Once the box logs you out, start pinging the IP address. It will disappear and come back. Once you can ping it again, try to open up a VNC session. It can take a couple of minutes after the box is back up for it to actually allow vnc sessions. The VNC password is in the kickstart script on puppet1: "grep vnc /mnt/fedora/app/fi-repo/rhel/ks/xen-host":

vncviewer $IP:1

If using the standard kickstart script, one can watch as the install completes itself, there should be no need to do anything. If using the xen-host-nohd script, one will need to configure the drives.

Post Install

Do not forget to change the root password away from the default!

If all goes well, the vnc session will close, the box will reboot and come back up as the new host. The default root password is also listed in the kickstart script, from puppet1: grep rootpw /mnt/fedora/app/fi-repo/rhel/ks/xen-host Most physical machines are to be used as xen hosts, If that is the case with this host, just install puppet, update the box and follow the normal puppet instructions

Virtual Machine (domU)


Almost all of our virtual machines run off of LVM. Step 1 is to create the LVM partition you want.

lvcreate -n $NEWHOST -L 15G VolGroup00
Machines in PHX that are running on the netapp over iscsi are stored on volgroup xenGuests
If building on Telia1 make sure to use the second xen bridge "-b xenbr1"


Once the size of the new machine is set, we need to run the virtinstall. As before ensure that the ip listed below has access to the site. This can be tricky, normally the ip, route and netmask can be templated from the dom0. This is not the case at server beach (see below for clarification)

virt-install -p -n $NEWHOST -r 1024 -f /dev/VolGroup00/$NEWHOST --nographics \
  -l \
  -x "ks= ip=$IP \
   netmask= gateway=$GATEWAY dns=$DNS1,$DNS2"

Or in PHX:

virt-install -p -n $NEWHOST -r 1024 -f /dev/xenGuests/$NEWHOST --nographics \
   -l \
   -x "ks= \
   ip=$IP netmask= gateway= dns=,"

This install should be completely non-interactive.

To install a different OS change the install source (like Fedora 8's install source)

Post Installation

Do not forget to change the root password away from the default!

The installation process is pretty simple, the post configuration may not be depending on if the box you've installed has a reverse DNS lookup. Here's the checklist:

1. Ensure the hostname is set properly in /etc/sysconfig/network and /etc/hosts

1. Ensure the system is up to date and can contact its yum mirror yum -y update .

1. For an external box make sure /etc/resolv.conf contains search while internal hosts (in PHX) should contain search (this should be scripted by the kickstart file! -matt)

PPC Machine

PPC boxes are just used for builders/composers and are all in PHX.


These instructions only apply in PHX, and they presume that dhcp is already set up for the host. Also make sure the IP you are about to boot to install from is allowed to our IP restricted as noted above (in Introduction). Then, you'll need to grab the installer kernel and initrd

wget \
    -O /boot/vmlinuz-install

wget \
    -O /boot/initrd-install.img

grubby --add-kernel=/boot/vmlinuz-install --initrd /boot/initrd-install.img \
   --args="ks= ip=dhcp" --title "rekick"

Note that these instructions rely on dhcp. And if you put in the full ip information in yaboot.conf, yaboot gets very unhappy and is unable to boot.

Now, you'll need to reboot and watch the console carefully and select booting the 'rekick' option when the yaboot prompt comes up. That or change the default if you're brave.


Unfortunately, yaboot < 1.3.14 doesn't support a boot once, so you'll have to either watch the console and select the 'rekick' boot option or change the default if you're brave. This can take a couple of minutes as the ppc boxes spend a while in OF.

Post Install

To install a different OS change the install source (like Fedora 8's install source)

After the install, you'll want to change the network configuration to be static instead of dhcp. Just edit /etc/sysconfig/network and /etc/sysconfig/network-scripts/ifcfg-eth0

Network Boot/Rescue

Note that these instructions will only work in PHX and depend on the fact that dhcp is set up for the host.

You can also boot the machine from the network and start an install that way. To do this, you need to ensure that the machine has an entry in /etc/dhcpd.conf on lockbox like those for ppc1-4. Then, watch for the machine to boot and enter the SMS menu by hitting 1 when prompted. From the SMS menu, you can choose boot options (5) and then navigate to network boot. This will load yaboot over the network. Due to spanning tree, this will take a while as it has to wait 60 seconds before even trying to get the address and then each file.

Once you have a yaboot prompt, you can either choose the default which kicks off a builder install or select 'rescue' to boot into rescue mode on the machine


Make sure the correct hostname is set (edit /etc/hosts and /etc/sysconfig/network if necessary). Edit /etc/resolv.conf to have the correct search path. This should contain for all PHX machines, for all VPN machines, and for all machines (in that order).


Once the box is booted (virtual or not) follow the steps in the Puppet SOP


After puppet has done it's magic, set up the VPN. OpenVPN


Sign the cert on the certmaster. On puppet1:

sudo certmaster-ca -l # To list cert requests
sudo certmaster-ca -s [hostname]

FAS accounts

Run fasClient -i to get all the home directories populated.


Get the SSH public key from /etc/ssh/ and add it to the master known_hosts file in puppet (modules/ssh/files/ssh_known_hosts).

Server Beach

Server beach has some interesting network infrastructure as it relates to our ability to do virtualization. Basically the dom0 is given an ip on one network as normal but the virtual hosts (when we request IP's) are given an address on a different network and one without a gateway. The best bet is to make sure that you request at least one IP for the host to be a gateway. This is a terrible waste of an IP but until a better method is found this will work. Once you have your IP addresses all that is required is to create an aliased interface on the host with that IP. /etc/sysconfig/network-scripts/ifcfg-eth0:1 A reboot later and you can treat this xen host as a normal xen host (with bridged networking and such)