lilo vs grub

Chris Ricker kaboom at gatech.edu
Tue Oct 21 20:06:05 UTC 2003


On Tue, 21 Oct 2003, Alexandre Oliva wrote:

> On Oct 21, 2003, Alan Cox <alan at redhat.com> wrote:
> 
> > Lilo [...] can handle raid properly.
> 
> Can you define `properly' in the light of the long e-mail I posted
> about this the other day?
> 
> http://www.redhat.com/archives/fedora-test-list/2003-October/msg02399.html

There are at least two different issues w/ grub and raid in contrast to how 
lilo and raid work

0. grub-install fails with (most? all?) RAID1, whether trying to go to the
components (/dev/hda, for example) or going to the raid device (/dev/md0),
so you have to drop down in the grub shell and install it manually onto the
components. That's a problem given the baroque nature of the grub command
shell.... lilo(8), in contrast, works with RAID1 devices.
1. default grub installation done by anaconda is on only one of the 
components in the mirror, while lilo installation is on all

There may be others as well.

Your comments above only deal with the second problem. My general opinion is
that a solution which works some of the time (installing to all components)  
is better than one which doesn't ever work (installing to only one) in cases
of failure. Beyond that, though, there's the bigger problem that people 
who've used lilo w/ RAID have the expectation that the boot loader is 
getting put everywhere, so they don't know they need to follow up behind 
anaconda and fix things.

Your comments don't address the first issue at all.

Neither of these are showstoppers for grub + raid at least for me (hey, I
don't use lilo anywhere anymore, and haven't for a few years), but they're
differences from lilo that have to be known about by the people being forced
to migrate now, and they're differences people used to lilo's raid behavior
often don't like, which is why this thread has dragged on so long ;-)

later,
chris





More information about the fedora-test-list mailing list