Grub2 + mdadm RAID1 = cycle reset before grub menu :(

[Logicos]

Hello!
I have the same problem in debian (sorry for offtopic), i installed raid1 from debian installer, and when I take off first hdd i can't boot from second disk (sdb) in array, i can't see grub and have reset and so again, but when i use only first one disk (sda) all work fine .

Same problem with Debian/Squeeze...

[rodrigomartinho]

Hello guys.

I am having the same problem. Any new information?

I have a Ubuntu Server 10.10 x64 with 4 HDs. A raid1 for /boot and a raid10 for /. The instalation ran ok, in the end it asks to install grub2, and it installs on all drives. The system works as expected, and if i fail any drive, except the first one, system boot degraded. If i fail the first drive, i got this cycle booting. For testing, I made grub play a sound, and when system goes at cycling boot i get no sound. Does this mean that grub aint starting at all?

If that was the case, i expected it to be bios related. But i tried all boot sequences in bios and no sucess.

I am starting to think I should try raid1 for the whole system.

I really wanted this to work, so I'll try and help with anything i am able.

Regards,

Rodrigo.

[rodrigomartinho]

Some more information:

I couldn't boot neither with super grub disk, nor with super grub2 disk.

I tried fooling grub too, installing it again on all drives, all being hd0 (tried that editing device.map by hand).

And tried to boot a live CD to reinstall grub, but could't mount my root (raid10), getting an unknown file system error.

[rodrigomartinho]

I found a way of booting the system when 1st disk fails. I have a 5th disk, which I will use as a spare. It wasn't installed yet. I installed it in place of sdd. Then I cloned 1st disk to it (dd if=/dev/sda of=/dev/sdd). Installed sdd back, waited for resynch and changed sda for the spare. The system booted. I tested changing the system to see if it would boot with an out of date clone, and booted too. I don't know if the system would boot this way after a kernel update. I didn't test it.

[psusi]

That is a REALLY bad idea and will lead to a corrupt raid array. If you want the system to boot from any drive, then don't bother using a separate /boot partition. If you leave /boot in the raid5 array, you can install grub2 to all of the disks so any of them can boot.

[YesWeCan]

Hello guys.

I am having the same problem. Any new information?

I have a Ubuntu Server 10.10 x64 with 4 HDs. A raid1 for /boot and a raid10 for /. The instalation ran ok, in the end it asks to install grub2, and it installs on all drives. The system works as expected, and if i fail any drive, except the first one, system boot degraded. If i fail the first drive, i got this cycle booting. For testing, I made grub play a sound, and when system goes at cycling boot i get no sound. Does this mean that grub aint starting at all?

If that was the case, i expected it to be bios related. But i tried all boot sequences in bios and no sucess.

I am starting to think I should try raid1 for the whole system.

I really wanted this to work, so I'll try and help with anything i am able.

Regards,

Rodrigo.

Hi. I have a 4 disk RAID10 with RAID1 for /boot on 2 disks.
So this thread is worrying me.
Out of interest, with your array working perfectly and all disks present and correct, have you tried booting off different disks using the bios?

[rodrigomartinho]

Hi psusi, thanks for your reply.

But your sugestion was already tried.

My system has 4 hard drives, with a Raid1 /boot. All other partitions are Raid10.

In the system install, grub2 is installed in all hard drives, and if any fail, except for the first one, the system boots degraded, as expected. But, if the first drive fails, the system doesn't start anymore. It goes on an infinite cycling boot. Just after the bios screen, it boots again. I don't know even if grub is started. For testing, I made grub to play a beep when starting. When pc goes to this cycling boot, the beep doesn't play.

I have a feeling (just a feeling) that this is Raid10 related.

I've tried a lot of suggestions I found on many forums, but none worked. That's why I started trying odd solutions, as this disk clone.

Why am I going to have a corrupt array? In my previous tests it worked. But I made just a fast test. After cloning the first drive, I made changes in all my partitions, and after changing the first drive for the clone, the system booted. I just had to manually re-add one of the partitions.

As I said, I didn't test it after a kernel update though.

Any information would be much apreciated.

Regards,

Rodrigo.

[rodrigomartinho]

Hello YesWeCan.

Yes, I did try it and it worked.

My guess is that the grub in /boot of every disk points to the first disk. When first disk is working, you can boot from any of the other disks. When first disk fails, you can't boot anymore.

[rodrigomartinho]

Hi. I have a 4 disk RAID10 with RAID1 for /boot on 2 disks.
So this thread is worrying me.
Out of interest, with your array working perfectly and all disks present and correct, have you tried booting off different disks using the bios?

YesWeCan, make your own test. Disconnect the first drive and try to boot the system.

It would be usefull if more people with a similar system (/boot on Raid1, / on Raid10) made this test.

If you don't know which is your first drive, try all, or you can discover that with smartmontools, which can tell you the serial numbers of of your drives.

[YesWeCan]

Sure I'll try that. It's important we figure this out before someone loses their array and/or their hair. I think I will do a backup first just in case. I may not get this done today.

I am thinking about how Grub on the other disks would end up referencing grub on sda, if this is what is wrong. Presumably the MBR had to point to the PBR on the same disk. Does the PBR have to point to the bootloader on the same disk? I would have thought that if the bootloader on a disk is executed you would at least get a grub menu or prompt, rather than the PC resetting and rebooting.

I'm no expert but resetting sounds like something the bios would do if the MBR was corrupted in some way or points to non-executable code. But if this were true you wouldn't have been able to boot from any drive other than sda.