Walt Reed wrote:
On Thu, Nov 30, 2006 at 10:49:37AM +1300, MrKiwi said:
Walt Reed wrote:
When building the "replacement" server, it can help to sync / add
accounts before all the third-party crap goes on. We do it as part of
the kickstart %post scripts. Kickstart from pxe-boot is awesome -
especially on HP servers... :-) Once a machine is installed in the rack
and powered up for the first time, it's online and usable with all the
packages we need, preconfigured, in about 15 minutes.
Walt - Can you show us your scripts?
I am working on something similar - a way to deploy a server
using kickstart and then a handwritten script to configure
things like postfix, ip, iptables, mysql, apache etc
I imagine however that all you clever people have already
got this in your toolbox of tricks,.
Well, obviously they are customized for our situation, and contain a lot
of info about our infrastructure that I am not at liberty to divulge,
but I can give you some pointers.
Critical to the whole process is setting up DHCP correctly, a tftp
server, customizing the files in tftp so that the correct options are
set for booting, setting up a local YUM repository, apache to serve the
repository and kickstart files, etc. The kickstart script is almost
trivial compared to all that. We also use an NFS server for many things.
Kickstart, in %post, runs:
/sbin/chkconfig to enable and disable all the services we need or don't
need
Runs ntpdate and hwclock
Creates a bunch or directories in /usr/local and other areas, and
mountpoints for various NFS / SAN filesystems
Adds the mount info to fstab
Starts portmapper
Mounts a bunch of filesystems
Copies a slew of "standard" config files to /etc and elsewhere from the
NFS mounted archive
Creates a bunch of "stock" users via useradd / groupadd (we also use
LDAP.)
Runs Yum to install a whole slew of vendor and local packages
When a server installs itself, we find it's IP by looking in the dhcp
leases file on the dhcp server, then ssh into the machine and run
another script that changes the hostname, sets the IP to a
static address, and uses the HP client tools to reset the ILO Admin
password. (ILO is HP's remote console system - we run all our servers
headless.) To remove the need for the graphical ILO license, we nuke the
GRUB splash image, and add the option to the install / boot to force
text mode.
When I setup the kickstart process, I ran it over and over on the same
machine until it completed flawlessly - making notes of anything I had
to do manually, and then adding back in to the scripts. This can take
quite a while, and again is very customized to our particular hardware /
network / software set / environment.
The redhat manual on setting up kickstart is actually quite good.
http://www.redhat.com/docs/manuals/enterprise/RHEL-4-Manual/sysadmin-guide/
_______________________________________________
CentOS mailing list
CentOS@xxxxxxxxxx
http://lists.centos.org/mailman/listinfo/centos
Thanks Walt,
I like your idea about killing the gui boot part - i should
do that too.
For testing my scripts i use vmware, do an install as below,
then set a snapshot. I can restore to a snapshot and have
the (virtual) machine booted and ready to run the script in
about 45 seconds. I find vmware is fantastic for burn-in and
development of deployment scripts so long as you aren't
testing the hardware specific parts of your script.
For those interested here is what i do;
I also use HP servers (DL380 and DL360) with iLO.
I boot from CentOS disk 1, then manually type "linux
ks=http://10.1.1.8/installs/HowTo/anaconda-ks.cfg";
anaconda-ks.cfg looks like this;
======
# Kickstart file automatically generated by anaconda.
install
cdrom
lang en_US.UTF-8
langsupport --default=en_NZ.UTF-8 en_NZ.UTF-8
keyboard us
xconfig --card "ATI Mach64 3D RAGE II" --videoram 1024
--hsync 31-80 --vsync 56-75 --resolution 800x600 --depth 16
network --device eth0 --bootproto dhcp
rootpw --iscrypted blahblahblahblah
firewall --enabled --port=22:tcp
selinux --disabled
authconfig --enableshadow --enablemd5
timezone Pacific/Auckland
bootloader --location=mbr
# The following is the partition information you requested
# Note that any partitions you deleted are not expressed
# here so unless you clear all partitions first, this is
# not guaranteed to work
clearpart --linux --drives=sda,sdb
part /boot --fstype ext3 --size=100 --ondisk=sda
part pv.9 --size=0 --grow --ondisk=sdb
part pv.8 --size=0 --grow --ondisk=sda
volgroup VolGroup00 --pesize=32768 pv.9 pv.8
logvol / --fstype ext3 --name=LogVol00 --vgname=VolGroup00
--size=1024 --grow
logvol swap --fstype swap --name=LogVol01
--vgname=VolGroup00 --size=768 --grow --maxsize=1536
%packages
grub
kernel-smp
e2fsprogs
lvm2
%post
======
(10.1.1.8 is my utility server with apache and a centos repo)
Then i do the same as you - cat /var/lib/dhcp/dhcpd.leases
for the IP, then putty in with SSH and paste the following
line into the putty shell;
cd /tmp;thisfile=install_jslmail.v3.sh;wget -nv -N
http://10.1.1.8/installs/HowTo/$thisfile;dos2unix
$thisfile;chmod 700 $thisfile;./$thisfile
This grabs my install script and runs it.
In the install script i have scripted the following ;
1. hack out the windozy linebreaks in the script (in case i
have tweaked it from windoze) into a temp script, then run
that script (ie restart yourself after a pass of dos2unix)
2. Set the root password
3. Turn off IP6
4. set the IP address
5. fetch my own repo file for yum (which means packages come
from my local repo @ 100mbps)
6. grab and install dags rpmforge rpm
7. echo "includepkgs=clamav clamav-devel clamav-db unrar
smbldap-tools perl-Crypt-SmbHash perl-IO-Socket-SSL
perl-ldap perl-Unicode-String perl-Unicode-Map8
perl-Unicode-MapUTF8 perl-Convert-BER perl-Jcode
perl-Unicode-Map
perl-Net-SSLeay">>/etc/yum.repos.d/rpmforge.repo
8. Do a yum -y update (this gets the latest kernel)
9. Reboot
At this point we have the new IP address, IP6 turned off and
the latest kernel loaded. Reconnect to the IP address
specified in the script and re-run to last command (ie Up
arrow, Enter - the script will skip steps 1-9)
10. grab the latest iptables config from my server, restart
iptables
11. Turn off some services with chkconfig
12. Turn on daily updates for "locate"
13. yum the following packages ; yum-plugin-protectbase,
joe, ntp, gcc, glibc-devel, autoconf, automake,
kernel-smp-devel, bzip2-devel curl curl-devel rpm-build
rpm-devel sendmail-devel zlib-devel
14 run ntpdate to my local timeserver
15 Set the hostname using "hostname $mynewhostname" - maybe
this is not optimal
16. Install mysql, set root pwd
17. Install openldap, openldap-clients, config and start
18. Add a bunch of ldap entries for the domain
19. Install php5 using "yum -y --enablerepo=centosplus
install php php-gd php-pear php-ldap php-mysql php-mbstring"
20. Install and start apache
21. Install phpldapadmin and config
22. Add a bunch of ldap users
23. Set some aliases and run newalias
24. Install samba, config and start
25. Install postfix, config and start
26. Install amavisd-new, clamd and spamassassin
27. Install phpMyAdmin - not really needed, remove later
28. Install webmin - also not really needed, remove later.
_______________________________________________
CentOS mailing list
CentOS@xxxxxxxxxx
http://lists.centos.org/mailman/listinfo/centos
