Jim Popovitch (jimpop@xxxxxxxxx) said: > Well, I think that comment #2 identifies the issue as the kernel install > process not accurately determining the components necessary to build > initrd. I see the shortcut of just throwing lvm into the mix to 'make > it work', I just think that is not the best solution. I would hope that > the kernel install process relies on more than just believing that the > lvm tools are available just because mkinitrd is installed. Perhaps > not. The initrd for the kernel is created in the postinstall script of the kernel; this requires that everything needed to create the initrd be present at that point. Due to RPM transaction ordering, unless there is a specific requirement for the LVM tools somewhere in the kernel or its dependencies (such as mkinitrd), the LVM tools may get installed after the kernel, which would lead to a broken initrd. You could (I suppose) hack up a new RPm package property that enforces ordering specifics (package A must always be installed before package B in the same transaction set) without any actual requirements. But that would be sort of a hack. Bill
