Re: Wrong kernel selected during boot during RAID testing
On Tue, Nov 08, 2005 at 08:38:57AM -0600, Thomas F. O'Connell wrote:
> I used a sarge amd64 netinst image to configure a two-disk system
> with two RAID1s, one of which is a mount point for /boot and one of
> which is used as a physical volume for LVM.
> I had experimented with the sarge amd64 installer, the sid amd64
> installer, and the Ubuntu amd64 installer trying to get this setup
> working. Eventually, once I thought I had the partitioning process
> figured out, I returned to sarge.
> So I go through the installation process from scratch once again,
> trying to clear out everything and rebuild my partitions. sarge
> chooses grub as a bootloader, which is fine with me. I reboot,
> everything comes up clean, and I go through base-config. Then I shut
> down. At this point, the server is using the default 2.6.8 amd64-
> generic kernel that comes with the distro.
> Now I remove one of the two drives while the box is cold and start
> the machine. When it boots, there's an error about not finding the
> map for the kernel and somehow it finds a 2.6.12 kernel and boots
> with that, but none of the modules are there because the box was
> built and installed with a 2.6.8 kernel. So when it comes up, it's
> running the wrong kernel and has no networking.
> Where did this kernel come from, and how can I remove it? I feel like
> it must be a vestige of my installer experimentation, but I don't
> know how to clean the drives any more thoroughly during the
> partitioning stage of the sarge installer.
Make sure you install grub to the MBR of _both_ drives. or install the
generic MBR (using install-mbr) to both drives and install grub to the
boot sector of the raid device of the boot partition. I personally just
install grub to both MBRs. Right now you probably have an old grub
install on the other disk pointing to the wrong place. Or perhaps you
are not using raid on /boot and have hence no mirror on the other drive
so it contains old data.