I've just tried setting up a new machine (a Dell Precision 670) and hit this bug *hard*. The setup I was aiming for was a software RAID 1 for root (pair of SCSI discs, to be replaced by some small flash drives later), and at a later date, a software RAID 5 array for everything else (collection of SATA discs). I initially wanted to install Testing, but at the time I was looking for install images those available were having disk detection issues, so I plumped for a Lenny image instead, which installed just fine, and booted fine every time. Once Lenny was installed I upgraded to Testing, and once that was installed the reboot always failed, with the dreaded "no devices" preventing md0 from coming up. What really annoyed me at this point was the fact that I could never catch the last few log messages before the busybox shell popped up, as straight after it appeared some more boot messages about my disks would always scroll it off the top of the screen. This should have been a clue as to the issue, but I had no reason to even assume that Debian would have such a fundamental bug! Surely waiting until disk detection is complete (or at the very least, any dependencies of the root filesystem!) is pretty fundamental? Anyway. I've spent countless evenings after work trying to get this to work until I stumbled upon a mailing list post suggesting rootdelay as a workaround, and searching on that brought me here. Would it be possible to get a message to appear when the root device isn't available suggesting the use of rootdelay as a stopgap until this gets fixed? - Jamie
Attachment:
signature.asc
Description: OpenPGP digital signature