6 The Auto-Installation Process #

This is the default interface while auto-installing. No special variables are required to activate it.

Start installing a system using the serial console by adding the keyword console (for example console=ttyS0) to the command line of the kernel. This starts linuxrc in console mode and later YaST in serial console mode.

This option can also be activated on the command line. To start YaST in text mode, add textmode=1 on the command line.

Starting YaST in the text mode is recommended when installing a client with less than 64 MB or when X11 should not be configured, especially on headless machines.

For testing/rescue purposes or because the NIC does not have a PROM or PXE you can build a bootable flash disk to use with AutoYaST. Flash disks can also store the control file.

Tip: Creating a Bootable Flash Disk

To create a bootable flash disk, copy either the SUSE Linux Enterprise Server ISO image of DVD1 or the Mini CD ISO image to the disk using the dd command (the flash disk must not be mounted, all data on the device will be erased):

dd if=PATH_TO_ISO_IMAGE of=USB_STORAGE_DEVICE bs=4M

You can use the original SUSE Linux Enterprise Server DVD-ROM number one in combination with other media. For example, the control file can be provided via a flash disk or a specified location on the network. Alternatively, create a customized DVD-ROM that includes the control file.

Booting via PXE requires a DHCP and a TFTP server in your network. The computer will then boot without a physical medium. For instructions on setting up the required infrastructure, see Chapitre 10, Installation à distance.

If you do installation via PXE, the installation will run into an endless loop. This happens because after the first reboot, the machine performs the PXE boot again and restarts the installation instead of booting from the hard disk for the second stage of the installation.

There are several ways to solve this problem. You can use an HTTP server to provide the AutoYaST control file. Alternatively, instead of a static control file, run a CGI script on the Web server that provides the control file and changes the TFTP server configuration for your target host. This way, the next PXE boot of the machine will be from the hard disk by default.

Another way is to use AutoYaST to upload a new PXE boot configuration for the target host via the control file:

<pxe>
  <pxe_localboot config:type="boolean">true</pxe_localboot>
  <pxelinux-config>
    DEFAULT linux
    LABEL linux
    localboot 0
  </pxelinux-config>
  <tftp-server>192.168.1.115</tftp-server>
  <pxelinux-dir>/pxelinux.cfg</pxelinux-dir>
  <filename>__MAC__</filename>
</pxe>

This entry will upload a new configuration for the target host to the TFTP server shortly before the first reboot happens. In most installations the TFTP daemon runs as user nobody. You need to make sure this user has write permissions to the pxelinux.cfg directory. You can also configure the file name that will be uploaded. If you use the “magic” __MAC__ file name, the file name will be the MAC address of your machine like, for example 01-08-00-27-79-49-ee. If the file name setting is missing, the IP address will be used for the file name.

To do another auto-installation on the same machine, you need to remove the file from the TFTP server.

Adding the command line variable autoyast causes linuxrc to start in automated mode. linuxrc searches for a configuration file, which should be distinguished from the main control file in the following places:

The configuration file used by linuxrc can have the following keywords (for a detailed description of how linuxrc works and other keywords, see Appendix C, Advanced Linuxrc Options):

These variables and keywords will bring the system up to the point where YaST can take over with the main control file. Currently, the source medium is automatically discovered, which in some cases makes it possible to initiate the auto-install process without giving any instructions to linuxrc.

The traditional linuxrc configuration file (info) has the function of giving the client enough information about the installation server and the location of the sources. Usually, this file is not required, but it is needed in special network environments where DHCP and BOOTP are not used or when special kernel modules need to be loaded.

All linuxrc keywords can be passed to linuxrc using the kernel command line. The command line can also be set when creating network bootable images or it can be passed to the kernel using a specially configured DHCP server in combination with Etherboot or PXE.

The command line variable autoyast can be used in the format described in the following list.

Several scenarios for auto-installation are possible using different types of infrastructure and source media. The simplest way is to use the source media (DVD number one) of SUSE Linux Enterprise Server. But to initiate the auto-installation process, the auto-installation command-line variable should be entered at system boot-up and the control file should be accessible for YaST.

In a scripting context, you can use a serial console for your virtual machine, that allows you to work in text mode. Then you can pass the needed parameters from an expect script or equivalent.

The following list of scenarios explains how the control file can be supplied:

Note: Disabling Network and DHCP

To disable the network during installations where it is not needed or unavailable, for example when auto-installing from DVD-ROMs, use the linuxrc option netsetup=0 to disable the network setup.

Note: Difference between the autoyast and autoyast2 Options

The options autoyast and autoyast2 are very similar but differ in one important point:

• When you use autoyast=http://..., you need to provide linuxrc with the network configuration.

• When you use autoyast2=http://..., linuxrc tries to configure the network for you.

If autoyast=default is defined, YaST will look for a file named autoinst.xml in the following three places:

With all AutoYaST invocation options, excluding default, it is possible to specify the location of the control file in the following ways:

If only the path prefix variable is defined, YaST will fetch the control file from the specified location in the following way:

As an example, for 192.0.2.91, the HTTP client will try:

C000025B
C000025
C00002
C0000
C000
C00
C0
C
0080C8F6484C
default

in that order.

To determine the hex representation of the IP address of the client, use the utility called /usr/bin/gethostip available with the syslinux package.

# /usr/bin/gethostip 10.10.0.1
10.10.0.1 10.10.0.1 0A0A0001

The easiest way to auto-install a system without any network connection is to use the original SUSE Linux Enterprise Server DVD-ROMs and a flash disk. You do not need to set up an installation server nor the network environment.

Create the control file and name it autoinst.xml. Copy the file autoinst.xml to the flash disk.

If you choose to pass information to linuxrc using the info file, it is possible to integrate the keywords in the XML control file. In this case the file needs to be accessible to linuxrc and needs to be named info.

Linuxrc will look for a string (start_linuxrc_conf in the control file which represents the beginning of the file. If it is found, it will parse the content starting from that string and will finish when the string end_linuxrc_conf is found. The options are stored in the control file in the following way:

....
  <install>
....
    <init>
      <info_file>
<![CDATA[
#
# Do not remove the following line:
# start_linuxrc_conf
#
install: nfs://192.168.1.1/CDs/full-i386
textmode: 1
autoyast: file:///info

# end_linuxrc_conf
# Do not remove the above comment
#
]]>

      </info_file>
    </init>
......
  </install>
....

Note that the autoyast keyword must point to the same file. If it is on a flash disk, then the option usb:/// needs to be used. If the info file is stored in the initial RAM disk, the file:// option needs to be used.

The post-installation and system configuration are initiated directly after the last package is installed on the target system and continue after the system has booted for the first time.

Before the system is booted for the first time, AutoYaST writes all data collected during installation and writes the boot loader in the specified location. In addition to these regular tasks, AutoYaST executes the chroot-scripts as specified in the control file. Note that these scripts are executed while the system is not yet mounted.

If a different kernel than the default is installed, a hard reboot will be required. A hard reboot can also be forced during auto-installation, independent of the installed kernel. Use the reboot property of the general resource (see Section 4.1, “General Options”).

Most of the system customization is done in the second stage of the installation. If you require customization that cannot be done using AutoYaST resources, use post-install scripts for further modifications.

You can define an unlimited number of custom scripts in the control file, either by editing the control file or by using the configuration system.