Hi Jesus,
comments below.
Jesus Sanchez wrote:
> I've got some good news. At least now the cloned system boots. There
> was something related with the modules and therefore with the initrd file.
>
> I was using UYOK and not BOEL, sorry, but I tried to add some extra
> modules in the file /etc/sysconfig/kernel, line starting with
> INITRD_MODULES, run mkinitrd and repeat the whole process again
> (si_prepareclient, si_getclient, PXE-booting + autoinstall, reboot, ...)
>
> Now I can clone the system and when rebooting it starts using Grub,
> what I didn't force because finally I didn't add the PREFERED param
> under the BOOT section of systemconfig.conf file.
>
> I check the autoinstall logging and I saw several things:
>
> 1. The system tries to run lilo but it can't because kernel is too
> big (at least is what it says):
> Error: Cannot execute /sbin/lilo.
> Fatal: Kernel /boot/vmlinux-2.6.22.5-31-default.gz is too big
>
> at /usr/lib/systemconfig/Boot.pm line 112
This is a problem with the kernel shipped with your distribution and
lilo, not a systemimager / systemconfigurator issue. To reduce the
kernel size you could try to strip(1) it and re-compress it again, but
this would mean "changing" the kernel provided by the distribution, that
is never a good idea... better to use grub at this point.
>
> 2. After that it installs grub
>
> 3. At the end systemconfigurator says it has modified the following
> files:
> /etc/lilo.conf
> /etc/modprobe.conf
> /etc/modules.conf
> /etc/sysconfig/network/ifcfg-eth0
>
>
> There are some strange things:
>
> 1. The lilo.conf file created by systemconfigurator seems to be
> wrong, it has an orphan parameter which has to do with the root disk
> partition named by-id ('tor_6L080M0_L26FVVDG-part2'). Here you are both
> /etc/lilo.conf files, the original of the golden node and the one that
> generates systemconfigurator
>
> ----- /etc/lilo.conf of the golden node and the image
>
> # Modified by YaST2. Last modification on Wed Feb 6 13:51:17 CET 2008
> menu-scheme = Wb:kw:Wb:Wb
> timeout = 80
> lba32
> change-rules
> reset
> read-only
> prompt
> message = /boot/message
> boot = /dev/disk/by-id/scsi-SATA_Maxtor_6L080M0_L26FVVDG-part2
> default = oS_10.3
>
> image = /boot/vmlinuz-2.6.22.5-31-default
> ###Don't change this comment - YaST2 identifier: Original name: linux###
> label = oS_10.3
> append = "resume=/dev/sda1 splash=silent showopts"
> vga = 0x314
> initrd = /boot/initrd-2.6.22.5-31-default
> root = /dev/disk/by-id/scsi-SATA_Maxtor_6L080M0_L26FVVDG-part2
>
> image = /boot/vmlinuz-2.6.22.5-31-default
> ###Don't change this comment - YaST2 identifier: Original name: failsafe###
> label = Failsafe
> append = "showopts ide=nodma apm=off acpi=off noresume edd=off 3"
> vga = normal
> initrd = /boot/initrd-2.6.22.5-31-default
> root = /dev/disk/by-id/scsi-SATA_Maxtor_6L080M0_L26FVVDG-part2
>
> image = /boot/memtest.bin
> ###Don't change this comment - YaST2 identifier: Original name: memtest86###
> label = Memory_Test
>
> ----- end
>
> ----- /etc/lilo.conf on the cloned machine (modified by systemconfigurator)
>
> ##################################################
> # This file is generated by System Configurator. #
> ##################################################
>
> # Do all the normal things lilo does
> lba32
> map=/boot/map
> install=/boot/boot.b
> vga=normal
>
> # The number of deciseconds (0.1 seconds) to wait before booting
> prompt
> timeout=50
>
> # The boot device where lilo installs the boot block
> boot=/dev/sda
>
> # the default label to boot
> default=suse-10.3
>
> # Device to be mounted as the root ('/')
> root=/dev/sda2
> #----- Options for KERNEL0 -----#
> image=/boot/vmlinux-2.6.22.5-31-default.gz
> label=suse-10.3
> read-only
> append="auto BOOT_IMAGE=oS_10.3 ro root=/dev/sda2
> tor_6L080M0_L26FVVDG-part2 resume=/dev/sda1 splash=silent showopts"
> initrd=/boot/initrd-2.6.22.5-31-default
>
> ----- end
This is a bug of si_prepareclient that doesn't resolve non-standard
symlinks created by udev. In terms of "portability" it would be better
to resolve this name in the standard name. If you have potentially
heterogeneous clients, using a device naming policy based on hardware
informations, like vendor ID, etc. could be a problem because different
clients that have a different hardware could use the same image.
For this I think the following patch could fix the problem. Could you
test it?
---
Index: sbin/si_prepareclient
===================================================================
--- sbin/si_prepareclient (revision 4415)
+++ sbin/si_prepareclient (working copy)
@@ -75,6 +75,7 @@
use lib "USR_PREFIX/lib/systemimager/perl";
use strict;
use Carp;
+use Cwd 'abs_path';
use POSIX;
use File::Copy;
use File::Path;
@@ -746,6 +747,14 @@
$boot_dev =~ s/p[0-9]+$//;
}
+ # Resolve non-standard symlink names created by udev.
+ if (-l $boot_dev) {
+ $boot_dev = abs_path(readlink($boot_dev));
+ }
+ if (-l $root_dev) {
+ $root_dev = abs_path(readlink($root_dev));
+ }
+
# OK, we've all the info needed by systemconfigurator! Write the
# configuration file.
open(OUT, ">$sc_config_file") or
---
>
> 2. The system boots using grub but the file menu.lst has not been
> modified. At least the labels are the ones from the previous
> installation of openSUSE 10.1 of the golden node. Obviously it works
> because of the links of the /boot directory to the right versionof the
> kernel and initrd.
systemconfigurator should modify menu.lst, but since lilo has been
detect before, grub installation is not performed (even if lilo fails).
Are you using a post-install script to install grub?
Anyway adding the PREFERED option to systemconfig.conf could be a way to
force grub installation via systemconfigurator.
And looking at the full output of systemconfigurator would help to
confirm if this hypothesis is correct.
>
> ----- /boot/grub/menu.lst on the golden node, the image and the clone system
>
> # Modified by YaST2. Last modification on Wed Oct 18 16:28:13 CEST 2006
> color white/blue black/light-gray
> default 0
> timeout 8
> gfxmenu (hd0,1)/boot/message
>
> ###Don't change this comment - YaST2 identifier: Original name: linux###
> title SUSE Linux 10.1
> root (hd0,1)
> kernel /boot/vmlinuz root=/dev/sda2 vga=0x314 resume=/dev/sda1
> splash=silent showopts
> initrd /boot/initrd
>
> ###Don't change this comment - YaST2 identifier: Original name: failsafe###
> title Failsafe -- SUSE Linux 10.1
> root (hd0,1)
> kernel /boot/vmlinuz root=/dev/sda2 vga=normal showopts ide=nodma
> apm=off acpi=off noresume edd=off 3
> initrd /boot/initrd
>
> ###Don't change this comment - YaST2 identifier: Original name: memtest86###
> title Memory Test
> kernel (hd0,1)/boot/memtest.bin
>
> ----- end
>
>
> ----- /boot directory initrd and vmlinuz files
>
> lrwxrwxrwx 1 root root 26 Feb 26 22:12 initrd ->
> initrd-2.6.22.5-31-default
> -rw-r--r-- 1 root root 4544722 Feb 26 17:45 initrd-2.6.22.5-31-default
> -rwxr-xr-x 1 root root 1960389 Sep 24 21:09 vmlinux-2.6.22.5-31-default.gz
> lrwxrwxrwx 1 root root 27 Feb 26 22:12 vmlinuz ->
> vmlinuz-2.6.22.5-31-default
> -rw-r--r-- 1 root root 1594088 Sep 24 21:02 vmlinuz-2.6.22.5-31-default
>
> ----- end
>
> What should I do to be able to run /sbin/lilo? Trying using UYOK to
> reduce the size of the kernel? Modifying manually the existing
> /etc/lilo.conf on the image?
No. See above.
>
> And regarding Grub, what is being used accidentally, does
> systemconfigurator modify the menu.lst config file? Can I change the
> labels of the file directly on the image?
See above.
-Andrea
-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
_______________________________________________
sisuite-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/sisuite-users
