...making Linux just a little more fun!
By Mark Nielsen
White Box is very good for testing and developing since it doesn't require any licenses to use. It also should be considered for production use. Please consider contributing towards the White Box. It is a great project!
What is Kickstart? A means of automating Linux installations and upgrades. First, you create a single configuration file which you put on a floppy disk. Then, you boot your computer from the floppy disk, and it will be configured based on the Kickstart configuration file. If the configuration file is set up right, no questions or prompts are asked during the installation.
So what are the Kickstart tricks? In this article, I'm going to talk about the experience I've gained using Kickstart during the last 6 years and in the course of my work at Nuasis as a Systems Administrator and Programmer.
Please read the following manual before reading the rest of this article.
%ksappend http://10.10.10.10/include_this_file.ksinto your ks.cfg file. It will probably also work with another file on the floppy disk or cdrom (but why would you do that?).
append ip=dhcp
in the isolinux.cfg file of the boot floppy or cdrom, or by typing in
"ip=dhcp" at
the boot prompt, you can boot your computer from a DHCP server for
Kickstarting
purposes. The Kickstart process will also try to use DHCP if you try to get
something from the net and the network settings were not specified.
NOTE: The ip address during Kickstart doesn't have to be the same as
when
the computer reboots after installation.
Besides setting up the network settings, DHCP has the capability of telling your computer where to get the Kickstart file. Here is the problem: The installation program has no way of getting a file except through NFS. Trying to pass a url for a the location of the file won't work. I hate NFS. I have always had problems with it. I prefer to use a web server for Kickstarts.
There are two solutions around this:
Thus, the ks.cfg file located on the floppy or cdrom would look like this:
keyboard us lang en_US langsupport en_US network --bootproto=dhcp %ksappend http://10.10.10.10/include_this_file.ks
The isolinux.cfg file would look like
default my_default prompt 1 timeout 600 display boot.msg F1 boot.msg F2 options.msg F3 general.msg F4 param.msg F5 rescue.msg F7 snake.msg label my_default kernel vmlinuz append ip=dhcp initrd=initrd.img kssendmac ksdevice=eth0 lang=en_US label linux kernel vmlinuz append ks initrd=initrd.img label text kernel vmlinuz append initrd=initrd.img text label expert kernel vmlinuz append expert initrd=initrd.img label ks kernel vmlinuz append ks initrd=initrd.img label lowres kernel vmlinuz append initrd=initrd.img lowresThus, you can get around the fact that DHCP can't send a url as the location of your Kickstart file. Also, include_this_file.ks doesn't have to be a text file. It could be a script which prints out the correct Kickstart file based on the ip or MAC address.
Another trick -- let the DHCP server be a DNS server as well.
If your include command is something like this:
%ksappend http://Kickstart.foo/include_this_file.ksThen have DHCP point to itself to be the DNS server, and it can assign an ip address for "Kickstart.foo" (which could be itself). This enables you use an arbitrary server. Practically, you would probably have the dhcp, dns, and Kickstart services run on the same computer.
A sample isolinux.cfg file for the bootable cdrom would look like this:
default my_default prompt 1 timeout 600 display boot.msg F1 boot.msg F2 options.msg F3 general.msg F4 param.msg F5 rescue.msg F7 snake.msg label my_default kernel vmlinuz append ks=http://10.10.10.10/Kickstart/wb.ks ip=10.10.10.100 gateway=10.10.10.1 netmask=255.255.255.0 dns=10.10.10.50 initrd=initrd.img kssendmac \ ksdevice=eth0 lang=en_US label linux kernel vmlinuz append ks initrd=initrd.img label text kernel vmlinuz append initrd=initrd.img text label expert kernel vmlinuz append expert initrd=initrd.img label ks kernel vmlinuz append ks initrd=initrd.img label lowres kernel vmlinuz append initrd=initrd.img lowres
mkdir /mnt/hda mkdir /mnt/hdb mkdir /mnt/hdc mkdir /mnt/hdd mknod /dev/hda b 3 0 mknod /dev/hdb b 3 64 mknod /dev/hdc b 22 0 mknod /dev/hdd b 22 64 echo '/dev/hda /mnt/hda iso9660,ro 0 0 /dev/hdb /mnt/hdb iso9660,ro 0 0 /dev/hdc /mnt/hdc iso9660,ro 0 0 /dev/hdd /mnt/hdd iso9660,ro 0 0' >> /etc/fstab /mnt/sysimage/usr/bin/eject /dev/hda /mnt/sysimage/usr/bin/eject /dev/hdb /mnt/sysimage/usr/bin/eject /dev/hdc /mnt/sysimage/usr/bin/eject /dev/hdd
Here is a useful trick given the configuration below. The "wb.ks" file on the web server doesn't have to be a text file. If can actually be a script which detects the ip address or MAC address of the computer being Kickstarted and print out the appropriate Kickstart file. Thus, the url for the Kickstart file will stay the same, but the network settings might change. This is useful if you don't have a DHCP server. For Perl, look at the %ENV hash and for Python, look at the os.environ dictionary, to get the ip address or the MAC address. You can also pass arguments in the url if you know what you are doing (for example, pass the host name, and then the perl or python script can determine which Kickstart file to use if it doesn't recognize your ip address or MAC address).
A sample isolinux.cfg file for the bootable cdrom would look like this:
default my_default prompt 1 timeout 600 display boot.msg F1 boot.msg F2 options.msg F3 general.msg F4 param.msg F5 rescue.msg F7 snake.msg label my_default kernel vmlinuz append ks=http://10.10.10.10/Kickstart/wb.ks ip=10.10.10.100 gateway=10.10.10.1 netmask=255.255.255.0 dns=10.10.10.50 initrd=initrd.img kssendmac \ ksdevice=eth0 lang=en_US label linux kernel vmlinuz append ks initrd=initrd.img label text kernel vmlinuz append initrd=initrd.img text label expert kernel vmlinuz append expert initrd=initrd.img label ks kernel vmlinuz append ks initrd=initrd.img label lowres kernel vmlinuz append initrd=initrd.img lowres
An example in the ks.cfg file would be:
%post wget -q -O /mnt/sysimage/usr/src/local/sompackage.rpm http://myserver.local/sompackage.rpm
Below, I am going to show you how to install any binary file. You have to be careful about how much space you use up. First off, the installation program uses an image to get a lot of the binaries it uses during installation. An example image is ftp://mirror.physics.ncsu.edu/pub/whitebox/3.0/en/os/i386/RedHat/base/netstg2.img
Now, what you do is modify this image. Here is an example script:
### in case you run this script twice mkdir -p mnt umount mnt rm -rf new_image rm -f netstg2.img ## Get and mount the old image wget -q -O netstg2_orig.img ftp://mirror.physics.ncsu.edu/pub/whitebox/3.0/en/os/i386/RedHat/base/netstg2.img mkdir -p mnt mount -o loop netstg2_orig.img mnt ## Make the new image directory mkdir -p new_image ## rsync the old image to the new image rsync -a mnt/* new_image/ ## rsync a bunch of other stuff over rsync -a /usr/sbin/ssh* new_image/usr/sbin/ rsync -a /usr/bin/ssh* new_image/usr/bin/ rsync -a /usr/bin/rsync* new_image/usr/bin/ rsync -a /usr/bin/nmap* new_image/usr/bin/ rsync -a /usr/bin/scp* new_image/usr/bin/ rsync -a /bin/bash new_image/usr/bin/ ## Make a dependency list -- very crude ldd /usr/sbin/sshd | cut -d ' ' -f3 | cut -d '.' -f1 > List.dep ldd /usr/bin/ssh | cut -d ' ' -f3 | cut -d '.' -f1 >> List.dep ldd /usr/bin/nmap | cut -d ' ' -f3 | cut -d '.' -f1 >> List.dep ldd /bin/bash | cut -d ' ' -f3 | cut -d '.' -f1 >> List.dep ldd /usr/bin/rsync | cut -d ' ' -f3 | cut -d '.' -f1 >> List.dep ldd /usr/bin/scp | cut -d ' ' -f3 | cut -d '.' -f1 >> List.dep ldd /usr/bin/screen | cut -d ' ' -f3 | cut -d '.' -f1 >> List.dep ### execute the perl script to copy over the libraries chmod 755 Copy_Lib.pl ./Copy_Lib.pl ### Make your image. mkcramfs --verbose new_image netstg2.img ### replace netstg2.img on the Kickstart server or cdrom ### with the new one you just created.
Here is the Copy_Lib.pl script:
#!/usr/bin/perl open(FILE,"List.dep"); my @Lines = <FILE>; close FILE; (@Lines) = grep($_ =~ /[a-zA-Z0-9]/, @Lines); my $Home = "new_image"; foreach my $Line (@Lines) { chomp $Line; my $Reverse = reverse $Line; my (@Temp) = split(/\//, $Reverse, 2); my $Dir = reverse($Temp[1]); print `mkdir -vp $Home/$Dir`; $Command = "rsync -av --ignore-existing $Line* $Home/$Dir/" ; $Command2 = "rsync -av --copy-links --ignore-existing $Line* $Home/$Dir/"; print "$Command\n"; print `$Command`; print "$Command2\n"; print `$Command2`; }
I had trouble changing the default shell to bash during the Kickstart. So, in the %pre section of the ks.cfg file, I entered this command:
cp /usr/bin/bash /then when I clicked on Alt-F2 during installation time, I executed the command:
./bashand I got a good bash shell.
Here is another interesting tidbit, you can also install a service. For example, I tried to get sshd running during a Kickstart. I copied over all the libraries, created a shadow password file, modified /etc/passwd, and actually got sshd to run at Kickstart time automatically. However, when I tried to connect to the computer being Kickstarted, it would crash sshd running on that computer. Crashing sshd didn't affect the Kickstart process. It wasn't important enough for me to get it working, so I gave up.
If I actually could get
to work, I could watch the Kickstart process from my desktop.
I think running an Apache web server should be fine.
I could setup the web server
with a few python scripts. Those scripts would report on the health of
the Kickstart process. That would be cool. Python already exists in the
Kickstart program.
I pretty much have no future plans with Kickstart except to contribute to the Yum project by making a Yum server capable of making Kickstart configurations and managing Yum repositories. My dream is to replace RH satellite servers with Yum servers. I would of course use the Python programming language to do this, as it is my favorite programming language and Yum is already written in Python (yeah!).
Mark Nielsen works at Nuasis. In
his spare time, he writes articles relating to Free Software (GPL) or Free
Literature (FDL). His latest goal is to help Yum to replace commercial
satellite servers as much as possible. You can contact him at his old
website at http://tcu-inc.com.