Rich/Chance,
I had EXACTLY the same problem. This post prompted me to fix it.
Thanks a lot.
My setup is basic Redhat 6.0 but with clean 2.2.14 kernel sources,
mingo's 2.2.14-B1 patch & raidtools-19990824-0_90_tar.gz. After an
unsuccessful attempt to boot my kernel from a raid-1 array using
Michael Robinton's excellent root-raid howto and RH6.0's lilo, I
built
the RH6.2 lilo source RPM (LILO-0_21-15_SRC.RPM).
The RPM has the following patches:
Patch0: lilo-0.21-include.patch
Patch1: lilo-0.21-1.1.patch
Patch2: http://www.insync.net/~frantzc/lilo.patch
Patch3: lilo.raid1
Patch4: lilo-0.21-loopdev.patch
Patch5: lilo-0.21-second.patch
I (mindlessly) applied all the Redhat patches and now lilo works OK
- with
the exception of the behaviour that Chance describes below.
Does anyone have any ideas why lilo-0.21.15 messes up writing both
partitions when given "boot=/dev/mdN"??
I have not attempted to apply Richard Bollinger's 'lilo.raid1.also'
patch yet.
But I will now!
Martin Munt
[EMAIL PROTECTED] or [EMAIL PROTECTED]
> -----Original Message-----
> From: Richard Bollinger [SMTP:[EMAIL PROTECTED]]
> Sent: Monday, May 08, 2000 9:59 PM
> To: Chance Reschke; [EMAIL PROTECTED]
> Subject: Re: Lilo RAID-1 woes
>
> I'm not sure how to fix it, but I'm suspicious that some portion of the
> data
> which LILO should have written to both mirrors is only being written to
> the
> "first" drive in the mirror; which in your case is your second physical
> drive.
[MM] I'm sure you are right.
I have the same thing with a pair of IDE disks.
In /etc/raidtab, /dev/md5 is made up of /dev/hdc1and /dev/hda1 - in
that
order - and, sure enough, I have the same problem on boot.
If I physically remove hda, lilo boots up OK on (what is now) the
"first"
disk. This supports your idea that the second physical disk has the
boot
info. I did not test this by physically removing hda - I wish I had
now!
> Note that LILO says its working on /dev/sdb1 first, then /dev/sda1.
> On my test system, the boot disk is listed first in the array and it works
> fine as advertised. [MM] My lilo lists in hdc hda order too.
>
> Its a bit scary, but you can change the order of the drives in
> /etc/raidtab
> and rerun the mkraid command to change the order to what you'd like it to
> be; with /dev/sda1 as raid_disk 1 and /dev/sdb1 as raid_disk 2, without
> affecting any of the data actually stored on the drives. [MM] This fixed
> it
for me (Thanks, Rich!). I did: raidstop the array (in my case
/dev/md5);
change the order in raidtab to hda1, hdc1 and marked hdc1 as a
"failed-disk"; mkraid /dev/md5 - it runs degraded on /dev/hda1 only;
raidhotadd /dev/hdc1- it reconstructs; run lilo again; reboot. If
you need more
> explicit instructions we can take it off of the list. [MM] Please include
> me
in any off-list discussion. I would like to follow this one up.
> Rich B
> ----- Original Message -----
> From: "Chance Reschke" <[EMAIL PROTECTED]>
> To: <[EMAIL PROTECTED]>
> Cc: <[EMAIL PROTECTED]>
> Sent: Monday, May 08, 2000 4:25 PM
> Subject: Lilo RAID-1 woes
>
>
> > Hi,
> >
> > I've been trying to get Lilo to work with Linux software RAID-1. I'm
> > using the patches from Richard Bollinger's post from 5/4/2000, and
> > things almost work. I can use lilo to update the boot sector and
> > maps, but lilo will not boot unless I enter a valid kernel label at
> > the boot time lilo prompt.
> >
> > When I run lilo it appears to update the boot sector and maps;
> > however, after rebooting the lilo prompt appears, followed after the
> > timeout period by the message "Error 0x80". The lilo prompt then
> > reappears, and the pattern repeats. At any point while the lilo
> > prompt is being displayed I can enter a valid lilo label and the
> > corresponding kernel will boot. Pressing <TAB> at the lilo prompt
> > does not list the configured labels. Running 'lilo -R labelname'
> > before rebooting doesn't help.
> >
> > While I'm aware that the 0x80 error is supposed to refer to "Disk
> > timeout"
>
[MM] I assumed this meant "an error occurred reading disk 0x80"
(i.e. first bios disk) - silly me!
> I'm confident both disks in my array are up and running.
> >
> > My lilo.conf:
> >
> > boot=/dev/md0
> > root=/dev/md0
> > install=/boot/boot.b
> > prompt
> > timeout=50
> > image=/boot/vmlinuz-2.2.14smp
> > label=2.2.14
> > read-only
> >
> > I'm using the raid-2.2.14-B1 patches to a clean 2.2.14 source tree,
> > and lilo version 21 with patches from:
> >
> > http://www.elliott-turbo.com/lilo/lilo.raid1
> > http://www.elliott-turbo.com/lilo/lilo.raid1.also
> >
> > Running 'lilo -v -v' I get:
> >
> > LILO version 21, Copyright 1992-1998 Werner Almesberger
> >
> > boot = /dev/sdb, map = /boot/map.0811
> > Reading boot sector from /dev/sdb
> > Merging with /boot/boot.b
> > Secondary loader: 8 sectors.
> > Boot image: /boot/vmlinuz-2.2.14smp
> > Setup length is 4 sectors.
> > Mapped 1298 sectors.
> > Added 2.2.14 *
> > /boot/boot.0810 exists - no backup copy made.
> > Map file size: 10240 bytes.
> > Writing boot sector.
> > boot = /dev/sda, map = /boot/map.0801
> > Reading boot sector from /dev/sda
> > Merging with /boot/boot.b
> > Secondary loader: 8 sectors.
> > Boot image: /boot/vmlinuz-2.2.14smp
> > Setup length is 4 sectors.
> > Mapped 1298 sectors.
> > Added 2.2.14
> > /boot/boot.0800 exists - no backup copy made.
> > Map file size: 10240 bytes.
> > Writing boot sector.
> >
> > Any help fixing this would be great.
> >
> > Thanks,
> > Chance
> >
> > --
> > Chance Reschke
> > UW Astronomy
> >
> >
