Alan McKinnon writes: > Apparently, though unproven, at 18:57 on Saturday 04 June 2011, Alex > Schuster did opine thusly:
> > A mother board died, so I put the hard drives into an old spare PC, > > rebuilt the kernel via chroot in order to include some necessary > > drivers, and all was well. Until I plugged in the 2nd hard drive. > > Then, the thing reboots after these messages: > > > > [...] > > VFS: Mounted root (ext3 filesystem) readonly on device 8:6 > > Freeing unused kernel memory: 304k freed > > kjournald starting. Commit interval 5 seconds > > > > Instead, I think init should start. When I unplug the drive, it boots > > fine. > > Most obvious things first: > > What does the system present that BIOS as? Hmm, I cannot parse that sentence. The BIOS shows the identical SATA drives as 3rd and 4th master devices. These are the only hard drives > Is it now trying to boot off the second drive? No. sdb5 is a clone of sda5, but an older version, so I know that sda5's Grub is running. It loads the right kernel (there is no initramfs), then reboots. And even if the root=/dev/sda6 kernel parameter would find the wrong drive, it should just boot, as sdb6 also is a valid root partition (it's a rdiff-backup of sda6). For the moment, I unplugged the backup drive. I also had to add a --partial to the vgchange command in /lib/rcscripts/addons/lvm-start.sh, so the init script would not return an error because of missing physical volumes for a volume group that spans both drives. Which looks like a bug to me, I think this init script should be more tolerant. I may get another PC as replacement soon, so maybe this problem will go away then. It was working well in the old PC, the only change I made was to compile a new kernel, because the current PC has other IDE hardware than the old one. Wonko