Alan, I filed such a bug report awhile back... https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=187745 You can easily reproduce this problem. Start with a machine with Promise motherboard raid and configure a RAID-1 with two SATA drives in the BIOS. Now destroy the RAID-1 mirror in the BIOS by setting the two drives to Ultra IDE mode instead of RAID mode. From this point on dmraid will be confused by the residual tattooing from the original Promise raid. One has to resort to nodmraid to allow these two drives to be used for software raid. Specifically once you have a software raid installed on them, anaconda won't find these partitions unless nodmraid is used with the installer. Jack On Thu, Jun 14, 2007 at 04:52:50PM +0100, Alan Cox wrote: > > for motherboard RAID support. If your machine is ever configured > > that way, dmraid is too dumb to realize when you have stopped > > using motherboard RAID support and have switched to software > > RAID. For example, if you had two drives as a Promise RAID > > mirror and then switched them to simple IDE mode instead, the > > anaconda installer will always misread them after they are > > configured as software RAID unless you have disabled dmraid > > Sounds like that is worth filing as an RFE in bugzilla. dmraid in theory > knows enough about the raid formats to work out how to blank them > > (We can't just ignore them often or the BIOS does hideous things like 'Oh > look a raid mirror thats out of sync, let me just kick off the > destruction of your filesystem automatically') > > Alan
