<martinh@gnu.org>This HOWTO briefly describes how to use the RedHat Linux KickStart system to rapidly install large numbers of identical Linux boxes. For advanced users, we describe how to modify the KickStart installation procedure to do your own thing, and give a quick guide to building RPM packages of your own.
Copyright (c) 1998 Martin Hamilton, All rights reserved. This is free documentware; you can redistribute it and/or modify it under the terms of version 2 or later of the GNU General Public License.
If you got this document from a Linux HOWTO mirror site or a CD-ROM, you might want to check back to the KickStart HOWTO home page to see if there's a newer version around.
RedHat Linux version 5 comes with a a little-known (and until now, not hugely documented) feature called KickStart. This lets you automate most/all of a RedHat Linux installation, including:
Eagle eyed RedHat users will probably have realised by now that these are essentially the main steps involved in the manual installation of a RedHat Linux system. KickStart makes it possible for you to script the regular installation process, by putting the information you would normally type at the keyboard into a configuration file.
But wait - there's more :-)
Having finished the normal installation process, KickStart also lets
you specify a list of shell level commands which you would like to be
executed. This means that you can automatically install extra local
software not distributed as part of RedHat Linux (yes, there are even
more free software programs than the ones you get with the RedHat
distribution. Some can't be distributed by RedHat on legal grounds,
e.g. the ssh and PGP encryption systems) and carry
out any tidying up you may need to do in order to get a fully
operational system.
There are two approaches to a KickStart install - one is to simply copy your KickStart configuration file to a RedHat boot floppy. The other is to use a regular boot floppy and get your KickStart config file off the network.
In both cases, you'll need:
If you want to fetch your config file over the network, you'll need to export it via NFS - this is the only access method supported at the moment. The config file lets you specify a different NFS server to fetch the RedHat distribution itself from.
You can configure a static IP address for your machine - e.g. a special one reserved for KickStart installations. Alternatively, if you don't want to hard code an IP address in the config file you can tell KickStart to use a BOOTP/DHCP server to fetch this. Some servers will allocate new addresses in a given range automatically, e.g. the CMU BOOTP server with dynamic addressing extensions.
More information on NFS and BOOTP/DHCP is in Appendix A.
Essentially, all you have to do is copy your KickStart config file to /ks.cfg on the RedHat boot floppy, e.g.
mcopy ks.cfg a:
However - the RedHat boot floppy is pretty full, and you may find that
you have to delete some of the other files to make room for the
KickStart config file. I was able to make enough room for mine by
deleting the various message files normally displayed by the
SYSLINUX boot loader, e.g.
mdel a:\*.msg
Another approach would be to throw away the drivers for some of the hardware you don't have - see the section on modifying the boot floppy below.
You may also want to edit syslinux.cfg, the SYSLINUX
config file. This also lives in the top level directory of the RedHat
boot floppy. For instance, the following syslinux.cfg will
cause KickStart mode to be entered into automatically as the machine
boots up, without the normal delay:
default ks
prompt 0
label ks
kernel vmlinuz
append ks=floppy initrd=initrd.img
Note that you almost probably want to base your boot and supplementary floppies on the most recent disk images available in the updates/i386 on your local RedHat mirror site. Older images may be buggy or have driver support for less hardware.
There are three main sections to the config file:
There are some other possibilities which we won't talk about here, but might work. For more information check out the sample KickStart config in misc/src/install/ks.samp and doc/README.ks under the top level i386 RedHat distribution directory on your CD-ROM or local RedHat mirror site.
The available directives which I've been using are:
Language configuration, e.g. for English
lang en
Network configuration, e.g. to use BOOTP/DHCP
network --bootp
NFS server and directory to install from, e.g.
nfs --server chicken.swedish-chef.org /mnt/cdrom
Select keyboard type, e.g. for UK keyboards
keyboard uk
Clear the Master Boot Record - removes any existing operating system boot loader from your disk
Clear existing partitions - e.g. to remove all existing disk partitions prior to installation
clearpart --all
Partition the disk, e.g. to make a root filesystem of 500MB
part / --size 500
Make a fresh installation of RedHat Linux.
Set the mouse being used, e.g. for a PS/2 or compatible "bus mouse"
mouse ps/2
Set the timezone, e.g. for local time in the UK
timezone --utc Europe/London
Set the initial root password, based on a previously derived encrypted password
rootpw --iscrypted XaacoeGPmf/A.
Install the LILO boot loader, e.g. in the Master Boot Record
lilo --location mbr
Packages to install - see below.
Post-installation shell commands - see below.
Note that the directory where KickStart is looking for the RedHat distribution should have a subdirectory RedHat, which contains the RedHat distribution tree for the platform in question. In the above example, we should see something like the following files and directories:
/mnt/cdrom/RedHat
/mnt/cdrom/RedHat/base
/mnt/cdrom/RedHat/contents
/mnt/cdrom/RedHat/i386
/mnt/cdrom/RedHat/instimage
/mnt/cdrom/RedHat/RPMS
/mnt/cdrom/RPM-PGP-KEY
If you're installing off a CD-ROM rather than off the network, the contents should look something like this:
RedHat
RedHat/base
RedHat/contents
RedHat/i386
RedHat/instimage
RedHat/RPMS
RPM-PGP-KEY
If you have the RedHat distribution for multiple architectures (e.g. on an NFS server - they're too big to fit more than one architecture's version onto a single CD-ROM), you'll notice that each distribution has the same files and directories under a subdirectory, e.g.
alpha/RPM-PGP-KEY
i386/RPM-PGP-KEY
sparc/RPM-PGP-KEY
There should be a file architecture/Redhat/architecture,
e.g. i386/Redhat/i386.
If you want to create your own encrypted passwords, it's very easy using Perl, e.g.
% perl -e 'print crypt("schmurrdegurr", "Xa") . "\n";'p
Other options (or mooted options), which I've not tried:
Install off CD-ROM rather than network.
Explicitly declare device details, e.g.
device ethernet 3c509 --opts "io=0x330, irq=7"
device include scsi for SCSI
controllers and cdrom for proprietary CD-ROM drives.
Upgrade an existing installation rather than make a fresh installation.
Configure X Window server, graphics card and monitor. e.g.
xconfig --server "Mach64" --monitor "tatung cm14uhe"
I've not delved too deeply into this last one, because I'm not ever
planning to run X on the console of any of my KickStarted machines.
I'm told that running xconfig within KickStart itself is a
bit flaky, but the same functionality is also available from the
command line via Xconfigurator - so you might be best off
leaving this to the post-installation script.
Here's how this first part of a KickStart config file looks when we put all the bits together:
lang en
network --static --ip 198.168.254.253 --netmask 255.255.255.0
--gateway 198.168.254.1 --nameserver 198.168.254.2
nfs --server chicken.swedish-chef.org /mnt/cdrom
keyboard uk
zerombr yes
clearpart --all
part / --size 500
part swap --size 120
install
mouse ps/2
timezone --utc Europe/London
rootpw --iscrypted XaacoeGPmf/A.
lilo --location mbr
Note that some of the RedHat documentation refers to an invocation
of the network directive which doesn't actually work in
practice: network --option. The correct invocation is to
put network followed by --static, --bootp
or --dhcp. Be aware that the BOOTP and DHCP options are
different - to the extent that they even use different code.
You can add the --grow parameter to a part
directive to indicate that it's OK to grow the partition beyond
the size you specify. It probably only makes sense to have one
partition tagged with --grow.
The start of the packages section of the KickStart config file is
indicated by the presence of a %packages directive on a line
of its own. This is followed by one or both of two types of package
specifier - individual packages may be installed by giving the name of
their RPM (excluding the version and platform information), and groups
of packages may be installed by giving their group name.
Here's a sample packages section for a KickStart config file:
%packages
@ Base
netkit-base
bind-utils
ncftp
rdate
tcp_wrappers
traceroute
cmu-snmp
So, what are these groups ? Well, there are a number of groups defined by default in a file called base/comps under the RedHat distribution's top level directory. Here are the ones which were current at the time of writing:
You'll notice that they correspond to the various configurations which you're prompted for during a manual installation. Note that some of the packages in a given package group are duplicated in other groups, and that you can install multiple groups of packages without this causing problems. Each group's entry in the comps listing looks similar to this:
0 Extra Documentation
sag
lpg
howto
faq
man-pages
end
It seems that groups with a 1 next to their name (the first line above) are selected for installation by default. You can customise the Linux installation process even further by creating your own groups or redefine existing ones by editing this file.
This is probably the best feature of all, and something which there is no direct equivalent to in the manual installation process. What we can do here is specify a sequence of shell level commands which should be executed after the main installation (disk partitioning, package installation, and so on) is complete.
The beginning of this section is signified by the %post
directive in the KickStart config file. In what follows you can take
advantage of all of the utilities which have been installed on your
newly built Linux system, e.g.
%post
ln -s /etc/rc.d/init.d /etc/init.d
ln -s /etc/rc.d/rc.local /etc/rc.local
ln -s /usr/bin/md5sum /usr/bin/md5
ln -s /usr/bin/perl /usr/local/bin/perl
chmod ug-s /bin/linuxconf
mkdir /var/tmp/tmp
perl -spi -e 's!image=/boot/vmlinuz-.*!image=/boot/vmlinuz!' /etc/lilo.conf
rm /etc/rc.d/rc*.d/*sendmail
You can also use I/O redirection and here documents:
cat <<EOF >>/etc/passwd
squid:*:102:3500:Squid Proxy:/usr/squid:/bin/bash
EOF
cat <<EOF >>/etc/group
cache:x:3500:
EOF
Modify the run-time startup scripts:
cat <<EOF >>/etc/rc.local
echo 8192 > /proc/sys/kernel/file-max
echo 32768 > /proc/sys/kernel/inode-max
[ -x /usr/sbin/sshd ] && /usr/sbin/sshd
[ -x /usr/sbin/cfd ] && /usr/sbin/cfd
EOF
Set up crontab entries:
cat <<EOF >/tmp/crontab.root
# Keep the time up to date
0,15,30,45 * * * * /usr/sbin/ntpdate -s eggtimer 2>&1 >/dev/null
# Recycle Exim log files
1 0 * * * /usr/exim/bin/exicyclog
# Flush the Exim queue
0,15,30,45 * * * * /usr/exim/bin/exim -q
EOF
crontab /tmp/crontab.root
rm /tmp/crontab.root
And even install other RPMs which you made yourself:
rpm -i ftp://chicken.swedish-chef.org/rpms/squid.rpm
rpm -i ftp://chicken.swedish-chef.org/rpms/ssh.rpm
rpm -i ftp://chicken.swedish-chef.org/rpms/exim.rpm
rpm -i ftp://chicken.swedish-chef.org/rpms/cfengine.rpm
rpm -i ftp://chicken.swedish-chef.org/rpms/linux.rpm
ssh-keygen -b 1024 -f /etc/ssh_host_key -N ""
depmod -a
Note that you can achieve the same effect by making your own RPMs containing the commands you want executed - see below for more information. Give them a carefully chosen name and you can force them to be installed first (e.g. name starts with 'aaa') or last (e.g. name starts with 'zzz').
Be aware that a less painful way of doing root crontab entries is to create them as files in one or more of the directories /etc/cron.hourly, /etc/cron.daily, /etc/cron.weekly and /etc/cron.monthly.
More information about making your own RPMs is available in Appendix B.
Boot the to-be-installed machine off your RedHat boot floppy as usual,
but instead of pressing RETURN at the SYSLINUX
prompt, type linux ks.
If you're lucky, this will be all you have to type!
If you customised your RedHat boot floppy as outlined above, you won't even need to do this bit :-)
Since we're really just automating the normal steps involved in a RedHat installation, the normal dialogs may appear if/when KickStart gets confused about what to do next. The most likely case is that your network interface won't be detected automatically, and you'll be prompted for its IRQ and I/O address space. KickStart tends to need help for ISA bus cards, but detects PCI bus cards automatically.
You can keep an eye on what KickStart is doing by by switching virtual consoles as usual:
The RedHat boot disk boot.img is in MS-DOS format, using the
SYSLINUX program to boot up. The supplementary disk
supp.img is a Linux ext2 filesystem. If you have
support for the loopback filesystem in your Linux kernel, you can mount
both of these files in your filesystem and hack at them:
# mkdir -p /mnt/boot /mnt/supp
# mount -o loop -t msdos boot.img /mnt/boot
# mount -o loop supp.img /mnt/supp
Now you should be able to see and manipulate the files on the boot and
supplementary disk under /mnt/boot and /mnt/supp
respectively. Phew! Note that older versions of mount may
not be able to handle the -o loop option. In these cases
you'll need to explicitly use losetup to configure the
loopback device for each file, e.g.
# losetup /dev/loop0 boot.img
# mount -t msdos /dev/loop0 /mnt/boot
You might also need to explicitly use the -t ext2 option when
mounting an ext2 filesystem like the one on the supplementary disk.
But, it looks like people with modern Linux distributions shouldn't have to
worry about this.
Of course, if you don't want to mess around too much, you can cut a corner and manipulate actual floppy disks rather than these floppy disk images. If time is important, you'll probably prefer to use the loopback devices, since you can hack around with the disk images without incurring the latency associated with a genuine floppy disk read/write.
If you want to mess around with the installation procedure itself, the source code can be found on the RedHat CD-ROM or your local RedHat mirror site. It's in misc/src/install under the i386 distribution top level directory.
If you examine the RedHat boot disk you'll see that, in addition to the Linux kernel vmlinuz, there's a large file initrd.img:
- -rwxr-xr-x 1 root root 559 May 11 15:48 boot.msg
- -rwxr-xr-x 1 root root 668 May 11 15:48 expert.msg
- -rwxr-xr-x 1 root root 986 May 11 15:48 general.msg
- -rwxr-xr-x 1 root root 968842 May 11 15:48 initrd.img
- -rwxr-xr-x 1 root root 1120 May 11 15:48 kickit.msg
- -r-xr-xr-x 1 root root 5352 May 11 15:48 ldlinux.sys
- -rwxr-xr-x 1 root root 875 May 11 15:48 param.msg
- -rwxr-xr-x 1 root root 1239 May 11 15:48 rescue.msg
- -rwxr-xr-x 1 root root 402 May 11 15:48 syslinux.cfg
- -rwxr-xr-x 1 root root 444602 May 11 15:48 vmlinuz
You guessed it, this is another ext2 filesystem saved as a file
- - but with a twist. It's actually compressed as well. You can uncompress
it and then mount the result, e.g.
# gzip -dc /mnt/boot/initrd.img >/tmp/initrd.ext2
# mkdir /mnt/initrd
# mount -o loop /tmp/initrd.ext2 /mnt/initrd
Probably the most important part of this filesystem is the collection of loadable kernel modules which are included with the boot disk. If you need to merge in a new version of a driver, you'll need to either replace vmlinuz with a new kernel which has this statically linked, or replace it in the modules collection. What's more, you may need to throw other modules away to make room.
The modules collection is the file modules/modules.cgz.
Wondering what that might be ? It's actually a compressed
cpio archive, believe it or not! And you thought nobody used
cpio any more... Actually RPM itself uses cpio
internally, too. Here's how to hack around with it:
# gzip -dc /mnt/initrd/modules/modules.cgz >/tmp/modules.cpio
# cpio -itv <modules.cpio >modules.listing
# mkdir modules
# cpio -idumv <../modules.cpio
I don't believe that there is currently a way under Linux (at least in mainstream distributions) to transparently access compressed filesystems. Let me know if you know better!
If you change anything, remember to:
cpio to recreate the archive. How to do this is
left as an exercise for the reader...gzip to compress the resulting archive.gzip
again.
Finally, you can now create new boot floppies using this modified boot disk setup, e.g.
# cat boot.img >/dev/fd0
Q: After KickStart installation, my machine won't boot up.
The BIOS complains with a message like Missing operating system.
A: Sounds like the partition with the root filesystem on isn't
bootable. Use fdisk to toggle its bootable status.
Q: After the floppy boots, I get the message:
Error opening files for kickstart copy: File exists.
A: Use a more recent version of boot.img and supp.img - look in the updates directory of your local RedHat mirror site. There was a bug in some older versions of these for RedHat 5.1.
Q: Can you have all outstanding patches (update RPMs) applied automatically too ? How ?
A1: Copy the RPMs you want installing to the RPMS directory from which the installation is going to take place, get rid of the older RPMs, and update the file RedHat/base/hdlist with the new RPM details. See Appendix C for a script from Eric Doutreleau to do this for you. If you do this yourself, remember to run genhdlist afterwards!
A2: Try this Perl script: patchup. This compares the RPMs your system has installed with those in a nominated directory and reports on the ones it thinks need updating. It can even install them for you if you trust it to.
A3: rpm2hml has a much more powerful (12MB of C code vs. a page of Perl!) version of A2.
Q: A single config file on the install server for all of the clients, perhaps as a fallback after trying IPADDR-kickstart ?
A1: Use the BOOTP/DHCP 'boot file' parameter bf to set the filename.
A2: Add a a record bf=/kickstart/ks.cfg to the relevant
entry in /etc/bootptab.
Q: More flexibility when things go wrong - e.g. prompt for alternate locations if distribution not found on CD-ROM.
A: ?
Q: Explicit exclusion of packages - e.g. everything apart from sendmail.
A: Rebuild the BASE package without sendmail.
Q: Choose which services are started automatically on boot-up by the run-level scripts under /etc/rc.d/.
A: The chkconfig utility lets you configure which services are run automatically on boot-up. You can run this in your post-installation script section, e.g. to run ypbind in run levels 3, 4 and 5:
chkconfig --level 345 ypbind on
and it will start the ypbind level on the 345 level.
Q: When executing the shell commands in the %post section,
bring any output up in another virtual console rather than overwriting the
main screen. Could be done in the shell commands section using
open?.
A: No problem - do something like this:
exec >/dev/tty5
Q: Does the filesystem creation code check for bad blocks ?
A: If you switch to the virtual console where the filesystem creation output is being displayed, you won't see any mention of the 'read-only' test being performed. Looks like the answer is no.
Q: Can I arrange things so some of my machines are configured differently from others ?
A: You could move the host dependent stuff into the scripted section of the KickStart config - e.g. only install a given RPM if on a given machine. It would be useful to have a conditional installation feature in the packages section of the config file, e.g. switching on architecture, or hostname/domain name/IP address.
Q: Are there any changes between RedHat 5.1 and 5.2 ?
A1: Lots of changes in the installer, but mostly bug fixes or cosmetic improvements. No impact on KickStart as far as I can tell - from a diff -rcs of the two misc/src/install directories.
A2: RH5.2 now apparently includes the automatic IP allocation/DHCP
patches to bootpd, but they have left out the documentation
which tells you how to use it.
Q: (How) can you clear a specific partition or partitions ? e.g. to leave /home but zap /.
A: You can't - yet!
Q: Can you arrange to have your partitions created across
multiple drives ? e.g. / on sda and /home
on sdb.
A: Don't think so - looks like you only get access to the first drive from the partitioning tool.
Q: Is it possible to specify existing partitions to be included in the mount table, or is it only possible to specify the creation of new partitions that will then be included?
A: ?
Q: After running mkkickstart, where is the file
it creates?
A: It doesn't create a file - it dumps the KickStart
config to stdout.
Q: In virtual console 4 (Alt-F4) I get Unable to load NLS
charset cp437(nls_cp437). What does this mean ? Should I be
worried ?
A: Sounds like you're trying to mount a CD-ROM burned with the Joliet (Unicode extensions to ISO 9660. In theory the filenames on the CD-ROM might get munched and not make it through to Linux correctly. In practice it doesn't seem to cause any problems - could be a spurious dependency ?
Q: Why am i getting the X Window System installed ? I didn't put it in my list of packages.
A: The XFree86-VGA16 RPM is a 'base' component, and
as such always gets installed - unless you change the definition of
the base class.
Q: In my post-installation script, can I use the packages which have been installed by now to do funky things not possible with the limited tools on the floppies ?
A: Yep - e.g. if you chose to install Perl when you put your KickStart config together, almost anything is possible in about five lines :-)
Thanks to Eric Doutreleau for the info about chkconfig, the
SYSLINUX config file hack, and the Perl script for updating
your distribution server's RPMs. Thanks to Robert Kaminsky for
extensive investigations. Thanks to Piete Brooks, Flavia Regina
Munhoz, Tom Toffoli, Bob Robbins, Charlie Brady and Ragen Herrington,
for their comments and questions.
If you're wondering what on earth this BOOTP and DHCP stuff is, more information is available at the DHCP WWW site. NFS is documented separately in detail in the NFS HOWTO, and there's a DHCP mini-HOWTO too. I've tried to provide enough details here to help you get started, whilst not treating the topics in depth - let me know if you think this is overkill.
In the BOOTP/DHCP + NFS configuration we're discussing, the KickStart config file should be NFS mountable by the machine being installed from /kickstart/IPADDR-kickstart on the BOOTP/DHCP server, where IPADDR is the IP address of the new machine, e.g. /kickstart/198.168.254.254-kickstart for the machine 198.168.254.254.
You should be able to override this location by returning
the bf parameter (boot file) in your BOOTP/DHCP response. It
may even be possible to have this NFS mounted off another machine
entirely.
To NFS export some directories from an existing Linux box, create the file /etc/exports with contents something like:
/kickstart *.swedish-chef.org(ro,no_root_squash)
/mnt/cdrom *.swedish-chef.org(ro,no_root_squash)
Note that if you didn't register the IP addresses you're going to be using in the DNS, you may be told to get lost by the NFS server and/or the RPC portmapper. In this you can probably get away with putting IP address/netmask pairs in the config files, e.g.
/kickstart 198.168.254.0/255.255.255.0(ro,no_root_squash)
and in /etc/hosts.allow:
ALL: 194.82.103.0/255.255.255.0: ALLOW
This is because most Linux distributions use TCP wrappers to do access control for some or all of the NFS related daemons. Be aware that the /etc/exports syntax tends to be different on other Unix variants - the NFS servers bundled with Linux distributions tend to offer a much wider range of options than the ones shipped with other versions of Unix.
Be aware that if you include a root password in your KickStart config file, or NFS export directories containing sensitive information, you should take care to expose this information to as few people as possible. This can be done by making the NFS export permissions as fine grained as you can, e.g. by specifying a particular host or subnet to export to rather than a whole domain. If you keep a special IP address free for KickStart installations, everything's nice and simple, but you'll have to change it later - or reconfigure the machine to get its IP address via BOOTP/DHCP.
Most NFS servers require you to tell mountd and nfsd
(on some versions of Unix they're prefixed with a rpc.) that
the /etc/exports file has changed - usually by sending a
SIGHUP. There's often a program or script called
exportfs, which will do this for you, e.g.
# exportfs -a
If you didn't have NFS up and running when this machine was booted, the directories may not be exported automatically. Try rebooting, or running the following programs as root:
# portmap
# rpc.nfsd
# rpc.mountd
As noted, on some systems the rpc. prefix isn't used. In
most modern Unix distributions, these programs can be found in the
/usr/sbin or /usr/libexec directories. These might
not be in your path already, e.g. if you used su to become
root. The portmap program is also sometimes called
rpcbind, e.g. on Solaris, some versions of nfsd
require a command line argument specifying the number of instances of
the server to run, and you may find you also need to run another
daemon called biod. The above should suffice on most (all?)
Linux systems.
If you're using the CMU BOOTP server with DHCP and dynamic addressing extensions referred to earlier, a sample /etc/bootptab entry (/etc/bootptab is the normal location of the BOOTP/DHCP configuration file) would look something like this:
.dynamic-1:ip=198.168.254.128:T254=0x30:T250="ds=198.168.254.2:
dn=swedish-chef.org:sm=255.255.255.0:gw=198.168.254.1:
dl=0xFFFFFFFF":
(wrapped for clarity)
This says to allocate IP addresses dynamically on encountering new machines, starting at 198.168.254.128 and continuing for the next 48 (the hexadecimal value 30) addresses. Each client will be passed back the value of T250. In this case that sets:
ds to 198.168.254.2dn to swedish-chef.orgsm to 255.255.255.0gw to 198.168.254.1dl (how long the address is
valid for) to "forever"
There seem to be a number of other versions of this server kicking around which do not support dynamic addressing. For these, you would have to list the hardware (typically Ethernet MAC) address of each to-be-installed machine in /etc/bootptab, and the entries would look something like this:
bork.swedish-chef.org:ip=198.168.254.128:ha=0000E8188E56:
ds=198.168.254.2:dn=swedish-chef.org:sm=255.255.255.0:
gw=198.168.254.1:dl=0xFFFFFFFF":
(wrapped for clarity)
Note that the parameter ha corresponds to the hardware
address of the machine being installed.
The RPM package format is already very well documented, particularly in the book Maximum RPM by Ed Bailey, which you can download from the RPM WWW site - also available from all good book stores! This is just a couple of quick hints for people in a hurry.
RPM packages are built from a spec file. This consists (in a similar fashion to the KickStart config file) of a recipe of steps that need to be taken in order to build the package - it's expected that you'll have to build it from source, potentially for multiple platforms, and may need to apply patches before compiling. Once built and installed, a binary RPM will be created from the files and directories you specify as being associated with the package. It's important to note that RPM has no idea of which files and directories are related to a given package - you have to tell it.
Here's a sample specification for a custom RPM of the Squid WWW cache server:
Summary: Squid Web Cache server
Name: squid
Version: 1.NOVM.22
Release: 1
Copyright: GPL/Harvest
Group: Networking/Daemons
Source: squid-1.NOVM.22-src.tar.gz
Patch: retry-1.NOVM.20.patch
%description
This is just a first attempt to package up the Squid Web Cache for easy
installation on our RedHat Linux servers
%prep
%setup
%build
configure --prefix=/usr/squid
perl -spi -e 's!#( -DALLOW_HOSTNAME_UNDERSCORES)!$1!' src/Makefile
make
%install
make install
%files
/usr/squid
Here's how to build this RPM:
% mkdir -p SOURCES BUILD SRPMS RPMS/i386
% cp ~/squid-1.NOVM.22-src.tar.gz SOURCES
% cp ~/retry-1.NOVM.20.patch SOURCES
% rpm -ba squid-1.NOVM.22+retry-1.spec
This will automatically create a subdirectory under the BUILD
directory, into which it'll unpack the source code and then apply the
patch (there are a number of options available for patching - check
the book for details). Now, RPM will automatically build the package
by running configure and then make, install it using
make install, and take a snapshot of the files under
/usr/squid. It's the latter which will form the binary RPM
of the Squid software.
Note that we can insert arbitrary shell commands into the unpacking,
building and installing processes, e.g. the call to perl
which tweaks one of Squid's compile-time parameters.
The final binary RPM will be left under the RPMS directory in
the platform specific subdirectory i386. In this case it
will be called squid-1.NOVM.22-1.i386.rpm. Note that the
filename is created by concatenating the values of the following
parameters from the spec file: Name, Version and
Release - plus the hardware platform in question,
i386 in this case. Try to bear this in mind when creating
your own RPMs, to avoid giving them overly long or painful names!
It's also worth bearing in mind that you can build RPMs without having to rebuild the whole software package, e.g.
Summary: Linux 2.0.36 kernel + filehandle patch + serial console patch
Name: linux
Version: 2.0.36+filehandle+serial_console
Release: 1
Copyright: GPL
Group: Base/Kernel
Source: linux-2.0.36+filehandle+serial_console.tar.gz
%description
This is just a first attempt to package up the Linux kernel with patches
for installation on our RedHat Linux servers
%prep
echo
%setup
echo
%build
echo
%install
echo
%post
/sbin/lilo
%files
/lib/modules/2.0.36
/boot/vmlinuz
In this case we simply create an RPM based on the /boot/vmlinuz file and the contents of the directory /lib/modules/2.0.36, and execute /sbin/lilo after the package has been installed on a target machine. Let me know if you know much neater way of writing the spec file than this.
Here is Eric's script for munging updated RPMs into the RedHat distribution area:
#!/usr/bin/perl
#
$redhatdir="/cdrom/i386";
$rpmdir="/cdrom/i386/RedHat/RPMS/";
$updatedir="/cdrom/updates/";
@OTHERDIR=($updatedir);
foreach $dir (@OTHERDIR)
{
print "update for $dir\n";
system(" find $dir -name \"*.rpm\" -exec cp {} $rpmdir \\; ");
}
chdir($contribdir) || die "peux pas aller dans $contribdir $!\n";
system("chmod -R 755 $redhatdir");
chdir($rpmdir) || die "problem to go in $rpmdir $!\n";
#
# remove the old file
#
opendir(DIR,'.');
@package=grep(/\.rpm$/,readdir(DIR));
foreach $file (@package)
{
$file =~ /(.*)\-([\d+|\.]+\w*)\-(\d+)\.[i386|noarch].*/;
$nom=$1;
$version=$2;
$buildvers=$3;
if ($NOM{$nom})
{
$version2=$VERSION{$nom};
$buildver2=$BUILDVERS{$nom};
$file2=$FILE{$nom};
$nom2=$NOM{$nom};
if ( $version2 gt $version )
{
print "$file2 is newer than $file\n";
unlink($file);
}
else
{
if ( $version2 lt $version )
{
print "$file is newer than $file2\n";
unlink($file2);
$VERSION{$nom}=$version;
$BUILDVERS{$nom}=$buildvers;
$FILE{$nom}=$file;
$NOM{$nom}=$nom;
}
else
{
if ( $buildver2 > $buildvers )
{
print "$file2 : $buildver2 est mieux que $file : $buildvers\n";
unlink($file);
}
else
{
print "$file2 : $buildver2 is older than $file : $buildvers\n";
unlink($file2);
$VERSION{$nom}=$version;
$BUILDVERS{$nom}=$buildvers;
$FILE{$nom}=$file;
$NOM{$nom}=$nom;
}
}
}
}
else
{
$VERSION{$nom}=$version;
$BUILDVERS{$nom}=$buildvers;
$FILE{$nom}=$file;
$NOM{$nom}=$nom;
}
}
# we do the hard thing here
#
system("$redhatdir/misc/src/install/genhdlist $redhatdir");