[CentOS] mkinitrd trouble

Tue Feb 7 21:38:46 UTC 2006
Aleksandar Milivojevic <alex at milivojevic.org>

I've just updated one machine from 4.0 to 4.2.  The trouble is, the new 
kernel doesn't want to boot (2.6.9-22.EL, also tested 
  Usually, manually recreating initrd file solved this problem in the 
past for me.  But not this time.

Playing around, I also noticed that if I rebuild initrd for the old 
kernel, then it also fails to boot.  The only way to boot it up is 
using the original initrd image created when the system was initally 

The machine has two volume groups, sys and backup.  The sys volume 
group contains logical volumes with operating system, the backup volume 
group just holds some temporary application data.

Looking at what goes on the console during boot, the only difference 
between what happens is initialization of LVM (nash script in initrd 
image).  With the old one, it reports it has found two volume groups 
(backup and sys, in that order), and then that it actived all volumes 
in both of them.  With new initrd images, it reports it has found two 
volume groups, but activated only volumes from volume group sys.  The 
root file system is part of sys volume group (so its logical volume 
should be active).  The backup volume group contains file system with 
some data, and it's presence (or non-presence) theoretically shouldn't 
affect bootability of machine.  However, booting still fails with root 
file system not mounted and of course kernel panic since it can't find 

I've attempted doing several things (vgchange -a y, vgcfgbackup, 
checking the files it created, vgcfgrestore, then recreating initrd 
images).  But nothing helped.  I guess all needed device driver are 
correctly loaded, since it was able to detect volume group (and both 
volume groups are on the same RAID controller).  Running mkinitrd in 
verbose mode looks fine too.

Any help would be greatly appriciated.

This message was sent using IMP, the Internet Messaging Program.