Preparing to install from the network using PXE

This section describes how to configure TFTP and DHCP on a PXE server to enable PXE boot and network installation.

Network install overview

A network installation allows you to install CentOS to a system that has access to an installation server. At a minimum, two systems are required for a network installation:

PXE Server: A system running a DHCP server, a TFTP server, and an HTTP, HTTPS, FTP, or NFS server. While each server can run on a different physical system, the procedures in this section assume a single system is running all servers.

Client: The system to which you are installing CentOS. Once installation starts, the client queries the DHCP server, receives the boot files from the TFTP server, and downloads the installation image from the HTTP, HTTPS, FTP or NFS server. Unlike other installation methods, the client does not require any physical boot media for the installation to start.

To boot a client from the network, configure it in BIOS/UEFI or a quick boot menu. On some hardware, the option to boot from a network might be disabled, or not available.

The workflow steps to prepare to install CentOS from a network using PXE are as follows:

Steps

Export the installation ISO image (or the installation tree) to an NFS, HTTPS, HTTP, or FTP server.
Configure the TFTP server and DHCP server, and start the TFTP service on the PXE server.
Boot the client, and start the installation.

The GRUB2 boot loader supports a network boot from HTTP in addition to a TFTP server. Sending the boot files (the kernel and initial RAM disk - vmlinuz and initrd) over this protocol might be slow and result in timeout failures. An HTTP server does not carry this risk, but it is recommended that you use a TFTP server when sending the boot files.

Additional resources

To export the installation ISO image to a network location, see Creating installation sources for Kickstart installations for information.
To configure the TFTP server and DHCP server, and start the TFTP service, see Configuring a TFTP server for BIOS-based clients, Configuring a TFTP server for UEFI-based clients, and Configuring a network server for IBM Power systemsfor information.
Spacewalk can automate the setup of a PXE server. For more information, see the project website.

Configuring a TFTP server for BIOS-based clients

This procedure describes how to configure a TFTP server and DHCP server, and start the TFTP service on the PXE server for BIOS-based AMD and Intel 64-bit systems.

Procedure

As root, install the tftp-server package:
```
# yum install tftp-server
```
Allow incoming connections to the tftp service in the firewall:
```
# firewall-cmd --add-service=tftp
```
This command enables temporary access until the next server reboot. To enable permanent access, add the --permanent option to the command.

Configure your DHCP server to use the boot images packaged with SYSLINUX. A sample configuration in the /etc/dhcp/dhcpd.conf file might look like:

option space pxelinux;
option pxelinux.magic code 208 = string;
option pxelinux.configfile code 209 = text;
option pxelinux.pathprefix code 210 = text;
option pxelinux.reboottime code 211 = unsigned integer 32;
option architecture-type code 93 = unsigned integer 16;

subnet 10.0.0.0 netmask 255.255.255.0 {
	option routers 10.0.0.254;
	range 10.0.0.2 10.0.0.253;

	class "pxeclients" {
	  match if substring (option vendor-class-identifier, 0, 9) = "PXEClient";
	  next-server 10.0.0.1;

	  if option architecture-type = 00:07 {
	    filename "uefi/shim.efi";
	    } else {
	    filename "pxelinux/pxelinux.0";
	  }
	}
}

Access the pxelinux.0 file from the SYSLINUX package in the DVD1 ISO image file:

# mount -t iso9660 /path_to_image/name_of_image.iso /mount_point -o loop,ro

# cp -pr /mount_point/BaseOS/Packages/syslinux-tftpboot-version-architecture.rpm /publicly_available_directory

# umount /mount_point

Extract the package:

# rpm2cpio syslinux-tftpboot-version-architecture.rpm | cpio -dimv

Create a pxelinux/ directory within tftpboot/ and copy the required files, for example: pxelinux.0 libcom.c32, ldlinux.c32, vesamenu.c32 into it:
```
# mkdir /var/lib/tftpboot/pxelinux
```
```
# cp publicly_available_directory/tftpboot/pxelinux.0 /var/lib/tftpboot/pxelinux
```
Create the directory pxelinux.cfg/ in the pxelinux/ directory:
```
# mkdir /var/lib/tftpboot/pxelinux/pxelinux.cfg
```

Add a default configuration file to the pxelinux.cfg/ directory. A sample configuration file at /var/lib/tftpboot/pxelinux/pxelinux.cfg/default might look like:

default vesamenu.c32
prompt 1
timeout 600

display boot.msg

label linux
  menu label ^Install system
  menu default
  kernel images/CentOS-8/vmlinuz
  append initrd=images/CentOS-8/initrd.img ip=dhcp inst.repo=http://10.32.5.1/CentOS-8/x86_64/iso-contents-root/
label vesa
  menu label Install system with ^basic video driver
  kernel images/CentOS-8/vmlinuz
  append initrd=images/CentOS-8/initrd.img ip=dhcp inst.xdriver=vesa nomodeset inst.repo=http://10.32.5.1/CentOS-8/x86_64/iso-contents-root/
label rescue
  menu label ^Rescue installed system
  kernel images/CentOS-8/vmlinuz
  append initrd=images/CentOS-8/initrd.img rescue
label local
  menu label Boot from ^local drive
  localboot 0xffff

The installation program cannot boot without its runtime image. Use the inst.stage2 boot option to specify location of the image. Alternatively, you can use the inst.repo= option to specify the image as well as the installation source.
The installation source location used with inst.repo must contain a valid .treeinfo file.
When you select the CentOS 8 installation DVD as the installation source, the .treeinfo file points to the BaseOS and the AppStream repositories. You can use a single inst.repo option to load both repositories.

Create a subdirectory to store the boot image files in the /var/lib/tftpboot/ directory, and copy the boot image files to the directory. In this example, we use the directory /var/lib/tftpboot/pxelinux/images/CentOS-8/:
```
# mkdir -p /var/lib/tftpboot/pxelinux/images/CentOS-8/
# cp /path_to_x86_64_images/pxeboot/{vmlinuz,initrd.img} /var/lib/tftpboot/pxelinux/images/CentOS-8/
```

Start and enable the dhcpd service:

# systemctl start dhcpd
# systemctl enable dhcpd

Start and enable the xinetd service that manages the tftp service:
```
# systemctl start xinetd
# systemctl enable xinetd
```
The PXE boot server is now ready to serve PXE clients. You can start the client (the system to which you are installing CentOS), select PXE Boot when prompted to specify a boot source, and start the network installation.

Configuring a TFTP server for UEFI-based clients

This procedure describes how to configure a TFTP server and DHCP server, and start the TFTP service on the PXE server for UEFI-based AMD64, Intel 64, and 64-bit ARM systems.

Procedure

As root, install the tftp-server package:
```
# yum install tftp-server
```
Allow incoming connections to the tftp service in the firewall:
```
# firewall-cmd --add-service=tftp
```
This command enables temporary access until the next server reboot. To enable permanent access, add the --permanent option to the command.

Configure your DHCP server to use the boot images packaged with shim. A sample configuration in the /etc/dhcp/dhcpd.conf file might look like:

option space pxelinux;
option pxelinux.magic code 208 = string;
option pxelinux.configfile code 209 = text;
option pxelinux.pathprefix code 210 = text;
option pxelinux.reboottime code 211 = unsigned integer 32;
option architecture-type code 93 = unsigned integer 16;

subnet 10.0.0.0 netmask 255.255.255.0 {
	option routers 10.0.0.254;
	range 10.0.0.2 10.0.0.253;

	class "pxeclients" {
	  match if substring (option vendor-class-identifier, 0, 9) = "PXEClient";
	  next-server 10.0.0.1;

	  if option architecture-type = 00:07 {
	    filename "shim.efi";
	  } else {
	    filename "pxelinux/pxelinux.0";
		}
  }
}

Access the shim.efi file from the shim package, and the grubx64.efi file from the grub2-efi package in the DVD1 ISO image file:

# mount -t iso9660 /path_to_image/name_of_image.iso /mount_point -o loop,ro

# cp -pr /mount_point/BaseOS/Packages/shim-version-architecture.rpm /publicly_available_directory

# cp -pr /mount_point/BaseOS/Packages/grub2-efi-version-architecture.rpm /publicly_available_directory

# umount /mount_point

Extract the packages:

# rpm2cpio shim-version-architecture.rpm | cpio -dimv

# rpm2cpio grub2-efi-version-architecture.rpm | cpio -dimv

Copy the EFI boot images from your boot directory.

# cp publicly_available_directory/boot/efi/EFI/redhat/shimx64.efi /var/lib/tftpboot/uefi/

# cp publicly_available_directory/boot/efi/EFI/redhat/grubx64.efi /var/lib/tftpboot/uefi

You must rename the shimx64.efi file to shim.efi after it is copied.

Add a configuration file named grub.cfg to the tftpboot/ directory. A sample configuration file at /var/lib/tftpboot/uefi/grub.cfg may look like:

set timeout=60
menuentry 'CentOS 8' {
  linuxefi images/CentOS-8/vmlinuz ip=dhcp inst.repo=http://10.32.5.1/CentOS-8/x86_64/iso-contents-root/
  initrdefi images/CentOS-8/initrd.img
}

The installation program cannot boot without its runtime image. Use the inst.stage2 boot option to specify location of the image. Alternatively, you can use the inst.repo= option to specify the image as well as the installation source.
The installation source location used with inst.repo must contain a valid .treeinfo file.
When you select the CentOS 8 installation DVD as the installation source, the .treeinfo file points to the BaseOS and the AppStream repositories. You can use a single inst.repo option to load both repositories.

Create a subdirectory to store the boot image files in the /var/lib/tftpboot/ directory, and copy the boot image files to the directory. In this example, we use the directory /var/lib/tftpboot/images/CentOS-8/:
```
# mkdir -p /var/lib/tftpboot/images/CentOS-8/
# cp /path_to_x86_64_images/pxeboot/{vmlinuz,initrd.img} /var/lib/tftpboot/images/CentOS-8/
```

Start and enable the dhcpd service:

# systemctl start dhcpd
# systemctl enable dhcpd

Start and enable the xinetd service that manages the tftp service:
```
# systemctl start xinetd
# systemctl enable xinetd
```
The PXE boot server is now ready to serve PXE clients. You can start the client (the system to which you are installing CentOS), select PXE Boot when prompted to specify a boot source, and start the network installation.

Additional resources

For more information about shim, see the upstream documentation: Using the Shim Program.

Configuring a network server for IBM Power systems

This procedure describes how to configure a network boot server for IBM Power systems using GRUB2.

Procedure

As root, install the tftp-server package:
```
# yum install tftp-server
```
Allow incoming connections to the tftp service in the firewall:
```
# firewall-cmd --add-service=tftp
```
This command enables temporary access until the next server reboot. To enable permanent access, add the --permanent option to the command.
Create a GRUB2 network boot directory inside the tftp root.
```
# grub2-mknetdir --net-directory=/var/lib/tftpboot
Netboot directory for powerpc-ieee1275 created. Configure your DHCP server to point to /boot/grub2/powerpc-ieee1275/core.elf
```
The command’s output informs you of the file name that needs to be configured in your DHCP configuration, described in this procedure.
1. If the PXE server runs on an x86 machine, the grub2-ppc64-modules must be installed before creating a GRUB2 network boot directory inside the tftp root:
  # yum install grub2-ppc64-modules

Create a GRUB2 configuration file: /var/lib/tftpboot/boot/grub2/grub.cfg. Below is an example configuration file:

set default=0
set timeout=5

echo -e "\nWelcome to the CentOS 8 installer!\n\n"

menuentry 'CentOS 8' {
  linux grub2-ppc64/vmlinuz ro ip=dhcp inst.repo=http://10.32.5.1/CentOS-8/x86_64/iso-contents-root/
  initrd grub2-ppc64/initrd.img
}

The installation program cannot boot without its runtime image. Use the inst.stage2 boot option to specify location of the image. Alternatively, you can use the inst.repo= option to specify the image as well as the installation source.
The installation source location used with inst.repo must contain a valid .treeinfo file.
When you select the CentOS 8 installation DVD as the installation source, the .treeinfo file points to the BaseOS and the AppStream repositories. You can use a single inst.repo option to load both repositories.

Mount the DVD1 ISO image using the command:

# mount -t iso9660 /path_to_image/name_of_iso/ /mount_point -o loop,ro

Create a directory and copy the initrd.img and vmlinuz files from DVD1 ISO image into it, for example:
```
# cp /mount_point/ppc/ppc64/{initrd.img,vmlinuz} /var/lib/tftpboot/grub2-ppc64/
```

Configure your DHCP server to use the boot images packaged with GRUB2. A sample configuration in the /etc/dhcp/dhcpd.conf file might look like:

subnet 192.168.0.1 netmask 255.255.255.0 {
  allow bootp;
  option routers 192.168.0.5;
  group { #BOOTP POWER clients
    filename "boot/grub2/powerpc-ieee1275/core.elf";
    host client1 {
    hardware ethernet 01:23:45:67:89:ab;
    fixed-address 192.168.0.112;
    }
  }
}

Adjust the sample parameters (subnet, netmask, routers, fixed-address and hardware ethernet) to fit your network configuration. Note the file name parameter; this is the file name that was outputted by the grub2-mknetdir command earlier in this procedure.

Start and enable the dhcpd service:

# systemctl start dhcpd
# systemctl enable dhcpd

Start and enable xinetd service that manages the tftp service:
```
# systemctl start xinetd
# systemctl enable xinetd
```
The PXE boot server is now ready to serve PXE clients. You can start the client (the system to which you are installing CentOS), select PXE Boot when prompted to specify a boot source, and start the network installation.