Installation Methods and Overview

SUSE has long been known for providing quality documentation, and SLES 9 is no exception. The manual is broken into installation and administration sections. The installation section offers clear and useful advice on topics such as remote installation (SLES offers a VNC-based installation option to allow for installs on "headless" or blade servers) and simple topics, such as hard drive partitioning.

The installation itself can be handled in an automated fashion reminiscent of the default Windows approach, as well as in a hands-on fashion with numerous options to fine-tune the installation in terms of hard drive partitioning and package selection. We performed some simple installation tests of each method and found that while the more automated installation did a fine job of covering almost every contingency, it took a "kitchen sink" approach to package selection. Table 1.1 illustrates the amount of disk space required for the four installation types, ranging from a bare-bones base install with no graphics at all nor any exotic packages such as Apache to a full install in which relatively everything gets installed.

Table 1.1. SLES 9 Disk Space Requirement
INSTALLATION TYPE	DISK SPACE REQUIRED
Base system without graphics	400500MB
Minimal graphical system without KDE	600700MB
Standard (default)	1.5GB
Full	2.5GB

The less automated installation path gave us near-total flexibility, and that's the approach we recommend when the machine has a predetermined purpose: Select either the base system or the minimal graphical system and then manually select only the desired packages for installation. This results in a more streamlined, economic installation with fewer moving parts and less potential for extraneously running services and the security liabilities they incur.

TIP

Linux veterans will find almost every choice in the installation process familiar (regardless of previous distribution they may have used), and less-experienced administrators will probably receive all the help they need from hints provided by the installer. However, Linux novices will find the installation manual a useful tool while navigating an installation the first few times. If you don't have a printed manual, you can download the Adobe PDF document files located at www.novell.com/documentation/sles9/index.html.

Another installation option is to use AutoYaST, which allows for a one-time configuration to be recorded in an XML file and reused across multiple installations. The following sections provide a brief outline on performing SLES 9 installation using various methods.

NOTE

In the manual installation option, you have total control over the choice of drivers and settings. However, we have found that the automatic hardware detection performed by YaST is usually reliable. Therefore, the manual installation process using the Linuxrc program is not covered here. To learn more about Linuxrc, refer to Chapter 3 of the SLES 9 installation and administration guide (www.novell.com/documentation/sles9/pdfdoc/sles_9_admin_guide/sles_9_admin_guide.pdf).

CD-ROMBased Installation

The traditional installation method is to use the software CD-ROM set. SLES 9 consists of six CD-ROM discs. CD1 is bootable, so you simply insert it into the CD-ROM drive and boot the server from it. The rest of the discs contain software packages, documentation, and source code.

TIP

CD2 uses a 2.88MB diskette image instead of the bootable ISO image used on CD1. If you have trouble booting from CD1, give CD2 a try.

If your machine is not capable of booting from a CD, you can create a set of three boot diskettes. The necessary diskette images and instructions are in the boot directory of CD1. Refer to the README file if you are already on a Linux system or to README.DOS if you are on a DOS or Windows system. To create these diskettes on a DOS or Windows system, you will need the rawrite utility found in the rawrite directory under dosutils of CD1; a Windows version is located in rawwritewin under dosutils.

TIP

Because of limited space on the boot diskettes, additional device drivers are placed in five separate module diskette images. Therefore, you may want to create these five module diskettes at the same time you create the boot diskettes. The install script will prompt for the appropriate module disk if the driver is not found on the boot diskette.

During the CD-ROMbased installation process, YaST prompts you to change CDs when necessary (see Figure 1.1).

Figure 1.1. YaST prompts for the next CD required in the installation process.

VNC-Based Installation

The VNC-based installation method is designed mainly for the IBM's POWER processor-based systems, zSeries, and S/390 machines. These IBM systems can be configured to run a VNC host, so you can connect remotely using a VNC client to perform the installation.

NOTE

Virtual Network Computing (VNC) is a remote control software (see www.realvnc.com) program that allows you to view and interact with one computer (the "server") using a simple client program (the "viewer") on another computer anywhere on the network or even across the Internet. The beauty of VNC is that the two computers don't even have to be the same type or be running the same operating system. For example, you can use VNC to view an office Linux machine from your Mac at home (see sourceforge.net/projects/cotvnc). VNC is freely and publicly available and is in widespread active use by millions throughout industry and academia; it also is used extensively in private situations. You also can find a VNC server for NetWare at forge.novell.com/modules/xfmod/project/?vncnw.

To perform a VNC installation, you need to pass certain parameters to the kernel. This must be done before the kernel is launched. To do this, you need to include the following information in the Initial Program Load (IPL) configuration of the IBM server prior to starting the Linuxrc program:

 vnc=1 vncpassword=password install=source

vnc=1 indicates that the VNC server should be launched on the installation system. vncpassword= specifies the password that a client needs to supply when connecting to the host. The password must be at least six characters in length; otherwise, an error is generated when the VNC host starts up, and you will not be able to authenticate. The installation source (install=) can either be specified manually (by specifying the protocol and URL for the directory concerned, such as ftp://sles9.universalexport.ca/install_files or nfs://10.6.7.9/install_files), or it can be install=slp. In the latter case, the installation source is automatically determined by SLP query.

CAUTION

If install= is not specified, the server's local CD-ROM drive is used, which may not be what you want when doing a remote install.

The specific steps and syntax for setting up the VNC host on the IBM servers vary depending on the model, so you should refer to the system-specific documentation for details.

WARNING

The parameter keyword examples shown in section 3.1.7, "Passing Parameters to linuxrc," of the SLES 9 installation and administration guide do clearly reflect that the equal sign (=) is used to separate the keyword and its value when entered at the command line, and not a colon (:), as shown in the examples.

TIP

By default, a VNC-based install uses DHCP to obtain the IP information for the host. However, you can specify a static IP address by including hostip=x.x.x.x netmask=x.x.x.x as part of the Linuxrc parameter. You must specify netmask along with hostip; otherwise, DHCP will be used.

SLES 9 includes a Windows 32-bit VNC client (TightVNC) in the dosutils/tightvnc directory on CD1. You can also obtain the latest TightVNC client from www.tightvnc.com or use any one of the other VNC clients, such as the ones from www.realvnc.com.

The VNC server also has a small built-in HTTP server. With this, you can connect to the VNC server using a Java-enabled web browser. The VNC server listens for HTTP connections on port 5801; for example, the URL to use is http://server_name_or_ip_address:5801. The Java applet prompts you for your password (see Figure 1.2) and then displays the desktop.

Figure 1.2. Accessing the VNC server via HTTP.

A word of caution: VNC is unencrypted by default, so anyone who has access to the network path between you and the server and a packet sniffer can capture your sessions and extract passwords, server settings, or other entered data. Therefore, if you are performing a remote install, it is highly recommended that you perform only unsecured VNC installations over a trusted network or in a VMWare-like environment. If you must perform a VNC install across a nonsecured network (such as the Internet), we suggest you change the root password and any other password you entered during the unsecured session as soon as you finish the installation via an SSH or similar secure connection method.

Network-Based Installation

Often, it is much more convenient to place all the necessary installation files on a central server and perform the install across the network. This is generally faster than using the "disc jockey" method. Furthermore, files on a server can be easily updated. A side benefit of having such an installation server is the ease with which you can install packages at a later time; you can simply point YaST at this server instead of having to find the CDs and hope that the server's CD-ROM drive is not filled with dust bunnies. There are two ways in which you can set up the directory structure necessary for a network-based installation. One method is to copy the contents of each CD into a separate directory. This has the advantage of having a cleaner directory structure and makes updating specific CDs easier. The drawback is that when YaST needs to access contents on a different CD, you will be prompted to enter the name of that corresponding directory, thus slowing down the installation process a little. The other method is to copy the contents of all the CDs into a single directory. Using this setup, YaST will not prompt you to "switch" CDs.

Instead of setting up an ad hoc central installation server (such as an FTP server running on NetWare), SUSE LINUX provides an Installation Server module for YaST that supports installation across FTP, NFS, and HTTP connections. Consult Chapter 4, "Central Software Installation and Update," of the SLES 9 installation and administration guide for more details.

The Linuxrc installation process supports FTP, NFS, Samba, HTTP, TFTP, and SLP protocols for network-based installation.

AutoYaST-Based Installation

If you are rolling out servers using similar hardware that are similarly configured with the same set of basic services, you can make use of AutoYaST. AutoYaST relies on the hardware detection mechanism of YaST and normally uses default settings, but it can also be customized to suit your particular needs. Therefore, installation hosts need not be strictly identical. It is sufficient for them to have a similar hardware setup. You still need to take into account the limitations of the hardware itself, which cannot be circumvented by AutoYaST.

YaST includes an AutoYaST module, which can be used to create the necessary configuration. This configuration is then written to an XML file, so it can also be edited or even created manually. When you have the contents of the XML file to your satisfaction, copy it to a diskette so it can be used to install new systems. On the new server, pass the following in the boot options field

 autoyast=floppy:///filename.xml

where filename.xml is your configuration file.