Linux Network installation server

For successfull network installation you will need

1. DHCP or BOOTP-Server
2. TFTP Share (Trivial File Transfer Protocol)
3. Network installation source (HTTP, FTP or NFS)

So in general: you will need to configure your DHCP or BOOTP server to answer a bootp request from the openfirmware, usually this is done by adding a section with the mac-address of the machine to be installed to your server configuration.
Next we copy our network boot image from CD to /tftpboot directory and create full distribution network installation source on NFS share (or FTP, HTTP).

*dhcpd.conf example

Using AIX or VIO as a network installation server

Contents of /etc/bootptab:

sles1:bf=/tftpboot/sles1:ip=192.168.10.11:ht=ethernet:sa=192.168.10.10:sm=255.255.255.0:
sles2:bf=/tftpboot/sles2:ip=192.168.10.12:ht=ethernet:sa=192.168.10.10:sm=255.255.255.0:
sles3:bf=/tftpboot/sles3:ip=192.168.10.13:ht=ethernet:sa=192.168.10.10:sm=255.255.255.0:

Contents of /tftpboot:

# ls -al /tftpboot
total 32168
drwxrwxrwx   3 root     system          256 Nov 03 09:30 .
drwxr-xr-x  17 root     system         4096 Nov 04 07:48 ..
-rwxrwxrwx   1 root     staff       5484730 Nov 03 11:08 sles1
-rwxrwxrwx   1 root     staff       5484730 Nov 03 11:08 sles2
-rwxrwxrwx   1 root     staff       5484730 Nov 03 11:08 sles3

• Edit /etc/inetd.conf and uncomment bootps and tftp lines
• refresh -s inetd
• Use IPL Settings, directed bootp requests from client LPARs (assign adress and server address)

Example: installing Red Hat over network with kickstart on Power

First we need to decide either to use directed bootp request or bootp broadcast request. If you are installing from different subnet, you should use directed bootp request (providing installation server IP in the IPL or on the openfirmware command line). If you add your client mac address to the /etc/bootptab line you will not be asked for IP again. If you ommit the mac address the installation will stop and anaconda will ask for IP configuration.

Step 1: prepare server

Kernel

Copy your boot kernel to servers /tftpboot directory. You will find the boot kernel on the first Red Hat CD in the directory /images/pseries. The file is called netboot.img. Here is an example from U4 CD1:

# ls -al images/pseries/netboot.img 
-rwxr-xr-x  1 root root 6355144 Aug  3 04:24 images/pseries/netboot.img

bootptab

Add the appropriate line to your /etc/bootptab

p650-lpar2:bf=/tftpboot/p650-lpar2:ip=192.168.1.111:ht=ethernet:sa=192.168.1.1:sm=255.255.255.0:

you can also use ha:XXXXXXXXXXXX to add your mac address. Note that ha: should be after ht=ethernet

Start bootp and tftp

• If not done before start your bootp and tftp server.
  • Edit /etc/inetd.conf and uncomment bootps and tftp lines
  • refresh -s inetd

Client installation

• Activate the client LPAR and in vterm press 8 to get the OF prompt
• Once there you need to issue ls command and determine the OpenFirmware path to the network device:

  0 > ls 
  000000c83378: /ibm,serial
  000000c84100: /chosen
  ...
  000000d191b8: /vdevice
  000000d1bc48:   /vty@30000000
  000000d1cb20:   /l-lan@30000002

• Set device alias pointing to your network card you want to use for installation:

  0 > devalias net /vdevice/l-lan@30000002   ok

• Issue your boot command
  Here is an example for broadcast request (only works in the same subnet with mac address included in the bootptab)

  0 > boot net:ks=nfs:192.168.100.133:/export/redhat/ks/u4.cfg ksdevice=eth0

  And here is an example for directed boot request:

  0> boot net:SERVERIP,,CLIENTIP,GW-IP ks=nfs:<server>:/<path> ksdevice=eth0

Debugging

You can check tftp is running and working properly by "tftp localhost" and "get netboot.img"
If you suspect any bootp problems, kill the daemon and start it in foreground in debug mode:

# /usr/sbin/bootpd -s -d -d -d -d -d