IBM POWER5™ architecture is a high-performance 64-bit RISC (Reduced Instruction Set Computer) architecture with advanced features, such as:
- Server consolidation
- Large memory addresses
- World-class vertical scaling
- I/O performance
Linux® for POWER is a powerful platform that allows you to leverage advantages of both POWER5 hardware and open source tools. Some Linux distribution vendors, such as Red Hat®, Novell, and TurboLinux, have released Enterprise Linux distributions for eServer™ pSeries® and iSeries™ servers.
A single eServer p5 server can support multiple logical partitions (LPARs) simultaneously from the POWER5 architecture. Each LPAR is a subset of physical and logical resources and can be treated as a separate server that is capable of running an operating system, such as Linux. You can install many Linux instances on a single eServer p5 server's LPARs at the same time for various uses.
However, manual installation of many Linux instances from CD-ROM is time consuming and the situation can worsen if there is only one CD-ROM device installed on the eServer p5 server. In this article, I illustrate a cost-effective method to boot an eServer p5 server and install Red Hat Enterprise Linux on it from network. During the boot and installation process, there is no need to access the CD-ROM devices.
Install Linux from network overview
To boot the eServer p5 server and install Enterprise Linux from network, the eServer p5 server must have an Ethernet card and support booting from this Ethernet card. It must also be connected to a boot server and an installation server using a local area network (LAN), as shown in Figure 1 below. In the first stage (boot stage), the eServer p5 server gets the IP and boot image from the boot server. In the second stage (installation stage), the server gets the installation media from the installation server.
Figure 1. Network boot and install overview

The boot server is a machine that provides Enterprise Linux boot images to the eServer p5 server so that the Linux installer can be started. To do this, it usually hosts a DHCP (Dynamic Host Configuration Protocol) service so that an IP address can be assigned to the eServer p5 server with the boot image name and a TFTP (Trial File Transfer Protocol) service in order to transfer the boot image. The eServer p5 server acts as DHCP and TFTP clients of the boot server.
The installation server is a machine where Enterprise Linux installation media is stored. It hosts a network service, such as NFS (Network File System), FTP (File Transfer Protocol), and HTTP (Hypertext Transfer Protocol), so that the eServer p5 server can access the installation media.
Theoretically, you can use a machine of any OS platform and architecture, as long as it implements the required protocols, as a boot server or installation server. The boot server and installation server can be on a single machine or on different machines. In either case, the boot stage and installation stage are independent of one another. The network configuration of the eServer p5 server might be different in these two phases. If there are two or more Ethernet devices on the eServer p5 server, you can use one for boot and another for installation.
You also must have Enterprise Linux installation media, usually CDs or CD ISO image files. This article assumes that you have ISO image files. If not, you can easily create them from CD. To create an ISO image file from the Linux CD that has been inserted into the CD-ROM device, just execute the cp /dev/cdrom mycd.iso command.
To boot the eServer p5 server from network and install Enterprise Linux on it, you need to perform the following steps:
- Set up the boot server.
- Set up the installation server.
- Boot from network.
- Install from network.
This article guides you through these steps, one by one. To make things simple, I show examples of how to set up the boot server and installation server on a single PC machine with Red Hat Enterprise Linux AS 3 Update 5 installed, to boot an IBM p5 570 server from network, and to install Red Hat Enterprise Linux AS 4 Update 2 on this eServer p5 server. Special steps to install SuSE Linux Enterprise Server 9 SP 2 are also listed. The steps presented here are similar to the examples needed to install other Red Hat or SuSE distribution versions on other eServer p5 server models.
You need to install and configure DHCP and TFTP services on the boot server so that the boot image transfers to the eServer p5 server. Here the boot image is the installer binary of the Enterprise Linux distribution to be installed. Once the boot image loads, the Enterprise Linux installer starts.
Install and configure the DHCP service
The purpose of the DHCP service is to assign an IP address to the eServer p5 server and tell it which boot image to use. There is another protocol named BOOTP (Bootstrap Protocol) that you can also use. DHCP is a more flexible, backwards-compatible extension of BOOTP, and some of the latest systems can only boot using DHCP. In this article, only DHCP service is covered.
- Install the DHCP server package. Check whether the DHCP server package is installed. If not, install it now, as shown in Listing 1 below:
Listing 1. DHCP server package install[root@bootserver] # rpm -ivh dhcp-3.0.1-10_EL3.i386.rpm
- Edit the DHCP server configuration file --
/etc/dhcp.conf. You need to create this file if it does not exist.You usually need to specify the following configuration entries in this file:
- Subnet definitions: A subnet definition entry defines a subnet that the DHCP server dynamically assigns. A subnet is identified by a subnet IP and mask. IP ranges can also be specified for a subnet.
- Host definitions: A host definition entry defines which IP address to assign and which boot image a DHCP client (in this particular case, the eServer p5 server) uses by specifying its Ethernet card's MAC address.
An example of the DHCP server configuration file is shown in Listing 2 below:
Listing 2. DHCP server configuration fileoption domain-name "mydomain"; ddns-update-style none; default-lease-time 600; max-lease-time 7200; server-name "bootserver"; subnet 192.168.123.0 netmask 255.255.255.0 { range 192.168.123.200 192.168.123.201; deny unknown-clients; } host MyP5 { filename "boot.img"; server-name "bootserver"; hardware ethernet ae:32:20:00:b0:02; fixed-address 192.168.123.90; }In this example, the defined subnet is
192.168.123.0/255.255.255.0. The host entry is defined, which indicates that IP address192.168.123.90will be assigned to Ethernet cardae:32:20:00:b0:02and fileboot.img(in the TFTP server's root directory) will be used as the boot image. - Configure the boot server's IP address.
You must configure the boot server machine with an IP address, which is in one of the subnets defined in the DHCP server configuration file. Please refer to the RHEL 3 Reference Guide in the Resources section on how to configure the IP address.
- Listing 3 below illustrates how to start the DHCP server:
Listing 3. Starting the DHCP server[root@bootserver] # /etc/init.d/dhcpd stop [root@bootserver] # /etc/init.d/dhcpd start
Install and configure the TFTP service
The purpose of the TFTP service is to transfer a boot image to the eServer p5 server so that it will boot with this image. Many boot images can be stored in the TFTP server's root directory. The boot image file name specified in the DHCP service configuration file determines which boot image to transfer to the eServer p5 server.
- Install the TFTP server package. Check whether the TFTP server package is installed. If not, install it now, as shown in Listing 4 below:
Listing 4. TFTP server package install[root@bootserver] # rpm -ivh tftp-server-0.39-0.EL3.1.i386.rpm
You should also install the TFTP client package so that you can test whether the TFTP server works or not. See Listing 5 below.
Listing 5. TFTP client package install[root@bootserver] # rpm -ivh tftp-0.39-0.EL3.1.i386.rpm
The xinetd package must also be installed to start the TFTP server. By default, it should already be installed.
- Edit the TFTP server configuration file --
/etc/xinetd.d/tftp. You must turn on the switch (disable = no) so that the xinetd service starts the TFTP service.An example of the TFTP server configuration file is shown in Listing 6 below:
Listing 6. TFTP server configuration file example# default: off # description: The tftp server serves files using the trivial file transfer \ # protocol. The tftp protocol is often used to boot diskless \ # workstations, download configuration files to network-aware printers, \ # and to start the installation process for some operating systems. service tftp { disable = no socket_type = dgram protocol = udp wait = yes user = root server = /usr/sbin/in.tftpd server_args = -s /tftpboot per_source = 11 cps = 100 2 flags = IPv4 }Here the root directory of the TFTP server is
/tftpboot. - Copy the Enterprise Linux network boot images to the TFTP server root directory.
The first installation CD contains the network boot image for the Linux distribution.
- For RHEL AS 3 or 4, the network boot image file is
images/pseries/netboot.img. - For SLES 9, the network boot image file is
install.
For example, see Listing 7 to copy the network boot image to the TFTP server root directory for RHEL4U2:
Listing 7. Copying network boot image to TFTP server root directory[root@bootserver] # mount -o loop RHEL4-U2-ppc-AS-disc1.iso /media/cdrom [root@bootserver] # cp /media/cdrom/images/pseries/netboot.img /tftpboot [root@bootserver] # umount /media/cdrom
You can rename the network boot image so that the network boot images for many Linux distributions can coexist, as long as they are the same as what is specified in the DHCP server configuration file.
- For RHEL AS 3 or 4, the network boot image file is
- See Listing 8 to start the TFTP server.
Listing 8. Starting the TFTP server[root@bootserver] # /etc/init.d/xinetd restart
Set up the installation server
The installation server provides access to Enterprise Linux installation media during installation. It is typically stored in a specific directory. You can access it using various network protocols, such as NFS, FTP, and HTTP.
Prepare installation repository
The installation media can be organized in various structures (raw install tree, ISO image files, raw install disks, and YaST (Yet Another System Tool) install tree) in the installation media directory. Here I call this directory Linux installation repository. Different installation media directory structures are supported (Table 1) for each Linux distribution and each network access protocol.
Table 1. Supported installation media structures
| Linux distribution | Network protocol | Raw install tree | ISO image files | Raw install disks | YaST install tree |
|---|---|---|---|---|---|
| RHEL 3,4 | NFS | Yes | Yes | No | No |
| RHEL 3,4 | FTP, HTTP | Yes | No | Yes | No |
| SLES 9 | NFS, FTP, HTTP | Yes | No | No | Yes |
I also show examples of how to construct various kinds of installation repositories for RHEL4 U2 in the /install directory using CD ISO image files, such as:
- Raw install tree
All installation media is put into the installation repository. You can construct the installation repository by executing the following commands for each ISO image file, as shown in Listing 9 below:
Listing 9. Constructing raw install tree from ISO image files[root@bootserver] # mount -o loop,ro RHEL4-U2-ppc-AS-disc1.iso /media/cdrom [root@bootserver] # cp -arv /media/cdrom/* /install [root@bootserver] # umount /media/cdrom
- ISO image files
ISO image files are stored in the installation repository. Just copy all the ISO image files (or create symbolic links for them) into the installation repository, as shown in Listing 10 below. Do not change the file names. RHEL 3 and 4 installers support this kind of installation repository using NFS.
Listing 10. Copying ISO image files into the installation repository[root@bootserver] # cp RHEL4-U2-ppc-AS-disc?.iso /install
- Raw install disks
Separate directories are created for each disk under the installation repository, such as disk1, disk2, and so forth. Just mount the ISO image files or copy the installation media to these disk directories, as shown in Listing 11 below. RHEL 3 and 4 installers support this kind of installation repository using FTP or HTTP.
Listing 11. Constructing raw install disks from ISO image files[root@bootserver] # mkdir /install/disk1 [root@bootserver] # mount -o loop,ro RHEL4-U2-ppc-AS-disc1.iso /install/disk1
- YaST install tree
This type of install repository is constructed by the SuSE Linux YaST tool. The directory structure is different than above. You can construct it by selecting Miscellaneous | Installation Server in the YaST tool or by executing the
/sbin/yast2 instservercommand. The SLES 9 installer supports this kind of installation repository.
Install and configure the network service
You need to install and configure one of the following network services on the installation server so that the eServer p5 server can access the installation repository.
- NFS
For RHEL3, the NFS server package is installed by default. You need to add the following entry into the NFS export configuration file (/etc/exports) so that the installation repository (
/install) can be accessed using NFS, as shown in Listing 12 below:
Listing 12. Exporting installation media in NFS/install *(ro,no_root_squash)
You can specify an IP address instead of * in the export entry so that only the specified IP address can access the installation repository. This makes your NFS server more secure.
You also need to restart the NFS service to make the change effective, as shown in Listing 13 below:
Listing 13. Restarting NFS[root@bootserver] # service nfs restart
- FTP
Many FTP servers work the installation server. Very secure FTP daemon (vsftpd) is the recommended one. It is a GPL-licensed FTP server for UNIX® and Linux systems. It is included in Red Hat and SuSE Linux distributions as the default FTP server. Please refer to the RHEL 3 Reference Guide in the Resources section on how to set up a vsftpd server on RHEL 3.
- HTTP
Apache HTTP server is the most popular HTTP server on the Internet and it is widely supported by Linux distributions. Please refer to the RHEL 3 System Administration Guide in the Resources section on how to set up an Apache HTTP server on RHEL 3.
In the following sections, assume the Enterprise Linux installation media is located in the /install directory -- this directory is exported using NFS.
After setting up the boot server and installation server, connect them and the eServer p5 server to the LAN. Now you can boot the eServer p5 machine from network.
- Activate the eServer p5 server LPAR -- this is where Enterprise Linux will be installed. Open a HMC terminal to monitor the boot process.
- When you see the power-up screen (see Figure 2), press 1 to enter the SMS menu.
Figure 2. Power-up screen
- Choose 5. Select Boot Options > 1. Select Install/Boot Device > 6. Network. All network devices on your eServer p5 server will be listed (Figure 3). Choose the one from which you want to boot the machine.
Figure 3. Select network device to boot from
Here you also have a chance to view the Ethernet's device information. Make sure it is bootable and no IP address has been configured.
- Choose 2. Normal Mode Boot to boot from the Ethernet device. It contacts the boot server and transfers the boot image (Linux installer binary) from the boot server (see Figure 4).
Figure 4. Boot image in transfer
- After the boot image transfer is complete, the eServer p5 server LPAR will be booted from the image. The Enterprise Linux installer will be started (Figure 5 is an example for RHEL4).
Figure 5. Linux installer started
Once the installer starts, you can install Enterprise Linux from network. The detailed installer instructions depend on the Linux distribution. Here I mainly illustrate how to command the installer to use the installation media from the installation server.
- Choose English as the language during installation.
- For SLES9, choose Kernel Modules (Hardware Drivers). Then select Load ppc_pseries64 Modules to load the module for the Ethernet device, as shown in Figure 6 below. After that, choose Start Installation or System to start a new installation. This step is necessary to install SLES9 from network.
Figure 6. Load network module (SLES only)
- When prompting the installation method, choose to install from NFS, as shown in Figure 7 below.
Figure 7. Install from NFS
- Configure the eServer p5 server's IP address (Figure 8). You might specify to use dynamic or static IP configuration. Dynamic configuration will be OK if the DHCP server works at this point. Here I give an example using static IP configuration.
Figure 8. Specify local network address
- Specify the installation server's IP address and the directory where the Linux installation media is located (Figure 9).
Figure 9. Specify installation server
- After that, the installer gets the installation media from the installation server. The subsequent installation steps are the same as the ordinary installation from CD-ROM. Please refer to the installation guides provided by Linux distribution vendors to complete the remaining installation work.
By setting up a boot server and an installation server, you can boot your eServer p5 server from network and install Enterprise Linux on it. The required boot image is provided by the boot server and the required installation media is supplied by the installation server. Access to CD-ROM devices is not needed during the boot and installation procedure. Compared with the usual installation method from CD-ROM, installation from network provides the following benefits:
- You do not need to manually monitor the installation process. There is no need to frequently go in and out of the server room to change CDs during installation. You might finish the installation task remotely, such as at home.
- You do not need to use CD-ROM devices and Linux installation CDs during installation. This means you do not need to buy related hardware, software, and supplies.
- Since there is no bottleneck on CD-ROM devices, many Linux instances can be installed in parallel to save time.
- For Enterprise Linux, besides system installation, machines can be booted from the network for system rescue purposes. This rescue feature is provided by the Enterprise Linux installer.
Learn
-
Refer to the RHEL 4 Installation Guide for the IBM POWER Architecture for detailed steps to install RHEL4 on your p5 machine. The guide to install RHEL 3 is similar.
-
Refer to the SLES 9 Administration Guide for detailed steps to install SLES9 and how to administrate it.
-
The Installing Debian GNU/Linux 3.1 for PowerPC guide describes how to boot and install Debian from network in detail. It is also a good guide to install RHEL or SLES from network.
- The RHEL 3 Reference Guide describes how to configure an IP address and vsftpd on RHEL 3.
-
The RHEL 3 System Administration Guide describes how to configure Apache HTTP server on RHEL 3. Visit Apache HTTP Server Documentation for more documents on this open source project.
-
Read the following IBM Redbooks and Redpapers for information on IBM eServer p5 570 servers and how to operate LPARs using HMC:
- IBM eServer p5 570 Technical Overview and Introduction
- Effective System Management Using the IBM Hardware Management Console for pSeries
- Partitioning Implementations for IBM eServer p5 Servers
-
"Linux for pSeries installation and administration (SLES 9)"
(developerWorks, Nov 2004) gives a quick guide to install SLES9 on IBM pSeries servers. The installation on IBM p5 servers is somewhat similar.
- Want more? The developerWorks eServer™ zone hosts hundreds of informative articles and introductory, intermediate, and advanced tutorials on the eServer brand.
- The IBM developerWorks team hosts hundreds of technical briefings around the world which you can attend at no charge.
Get products and technologies
-
Download a free trial version of
Red Hat Enterprise Linux distribution.
-
Download a free trial version of
SuSE Linux Enterprise Server distribution.
Discuss
- Participate in the eServer forums.
- developerWorks blogs: Get involved in
the AIX community.
- developerWorks blogs: Get involved in
the developerWorks community.
Ya Liu is a Software Engineer at the IBM Software Development Lab in China. He currently works for the Linux Competency Center on the IBM WebSphere Information Integrator product. He has more than three years of Linux system administration and C/C++ development experience. You can contact him at liuya@cn.ibm.com.




