build kernel for server farm

I am pretty familiar with the build process and kernel install for a single Linux box, but I wanted to confirm I'm doing things in a sane way for a large deployment. All the machines are the same hardware and running standard setups.

First, I plan on compiling the kernel on a development box. From there my plan is basically tar /usr/src/linux, copy to each box, untar, copy bzImage and System.map to /boot, run make modules_install, edit lilo.conf, run lilo.

Tips? Comments?

Thanks
Matt

Network file system?

function.linuxpower.ca

For /usr/src and the kernel distribution? Or for the whole boot process?

I use NFS already for several shared filesystems. NFS isn't quite specific enough that I understand what you mean.

Could you give me a little more detail on exactly what you would use it for? If I've overlooked a howto or obvious document please RTFM me- I'm trying to learn from the pros here and would ideally like something that will scale well to over 100 machines.

Matt

Actually if they are exactly the same, why dont you just copy the relevant files over? ie. /lib/modules/uname -r/*, and as mentioned your bzImage, System.map, and lilo.conf (and of course remember to run lilo)?

Any particular reasons why you would want to go through the process of compiling again on each machine?

-dj

--- Matt Simonsen <[email protected]> wrote: > I am pretty
familiar with the build process and kernel install for a

http://careers.yahoo.com.au - Yahoo! Careers
- 1,000's of jobs waiting online for you!

If all of your systems are *identical*, try something like rsync or rdist (both can be told to use ssh too).

1) build your kernel and install it on your development box
2) do a test boot. Ensure everything is there for you as needed
3) build a Distfile like so:
##################
HOSTS = (
host1
host2
host3
)

KERNEL = (
/boot
/etc/lilo.conf
/etc/grub.conf
)

kernel:
${KERNEL} -> ${HOSTS}

	install ;
# could use a cmdspecial after this to run lilo if needed

modules:
/lib/modules -> ${HOSTS}

	install ;

################

Now, run rdist -P/usr/bin/ssh -overify -f Distfile

You should see it verify what will be updated on hosts[1234...]

Once that's done run a shell loop to reboot the systems as required

for i in host1 host2 host3 host4 ; do
ssh $i "shutdown -r now"
done

If they are not identical, you can do the same thing, only different /boot and /lib/module directories to each host would be required as well as seperate Distfile rules for each host or group of hosts.

Hope that helps.

Regards
James Bourne

James Bourne, Supervisor Data Centre Operations
Mount Royal College, Calgary, AB, CA
www.mtroyal.ab.ca

******************************************************************************
This communication is intended for the use of the recipient to which it is
addressed, and may contain confidential, personal, and or privileged
information. Please contact the sender immediately if you are not the
intended recipient of this communication, and do not copy, distribute, or
take action relying on it. Any communication received in error, or
subsequent reply, should be deleted or destroyed.
******************************************************************************


"There are only 10 types of people in this world: those who
understand binary and those who don't."

make rpm [1]

% rpm --install <your rpm>

[1] Either the genuine thing or the process of "go and build an rpm of your kernel". YMMV.

Regards
Henning

Dipl.-Inf. (Univ.) Henning P. Schmiedehausen       -- Geschaeftsfuehrer
INTERMETA - Gesellschaft fuer Mehrwertdienste mbH     [email protected]

Am Schwabachgrund 22  Fon.: 09131 / 50654-0   [email protected]
D-91054 Buckenhof     Fax.: 09131 / 50654-20   

You can avoid kernel version/config madness if you will arrange your farm to boot from single TFTP server and keep small readonly root fs with /boot/*, /lib/modules/* etc
on the NFS server (TFTP and NFS servers can be the same box).

You might want to have spare boot server, just in case ;) --
vda