* bootable RAID, number of member disks limitation
@ 2013-01-03 20:28 Chris Murphy
2013-01-03 22:00 ` Vladimir 'φ-coder/phcoder' Serbinenko
0 siblings, 1 reply; 5+ messages in thread
From: Chris Murphy @ 2013-01-03 20:28 UTC (permalink / raw)
To: The development of GNU GRUB
I'm finding GRUB2 will boot from md RAID 0, 1, 4, 5, 6, with /boot in the array, up to six member devices. At device 7, GRUB2 fails to boot the array, drops to a grub rescue prompt saying the mduuid couldn't be found.
With Btrfs single and RAID 0 profiles, /boot on Btrfs is bootable for up to 4 devices. At the 5th device in a volume, it's no longer bootable by GRUB2, I get a grub rescue prompt. I get variable results with RAID 10 for some reason where often it works with 5 and 6 disks.
The question is if this is expected, as-designed, and if it's stable? I feel like some subjective (certainly objective limits are OK) limit for number of member devices is acceptable, but needs to be known, so that installers can inhibit or warn users from creating unbootable arrays. And then how to communicate this downstream?
Chris Murphy
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: bootable RAID, number of member disks limitation
2013-01-03 20:28 bootable RAID, number of member disks limitation Chris Murphy
@ 2013-01-03 22:00 ` Vladimir 'φ-coder/phcoder' Serbinenko
2013-01-04 7:42 ` Chris Murphy
0 siblings, 1 reply; 5+ messages in thread
From: Vladimir 'φ-coder/phcoder' Serbinenko @ 2013-01-03 22:00 UTC (permalink / raw)
To: grub-devel
[-- Attachment #1: Type: text/plain, Size: 1684 bytes --]
On 03.01.2013 21:28, Chris Murphy wrote:
> I'm finding GRUB2 will boot from md RAID 0, 1, 4, 5, 6, with /boot in the array, up to six member devices. At device 7, GRUB2 fails to boot the array, drops to a grub rescue prompt saying the mduuid couldn't be found.
>
> With Btrfs single and RAID 0 profiles, /boot on Btrfs is bootable for up to 4 devices. At the 5th device in a volume, it's no longer bootable by GRUB2, I get a grub rescue prompt. I get variable results with RAID 10 for some reason where often it works with 5 and 6 disks.
>
> The question is if this is expected, as-designed, and if it's stable? I feel like some subjective (certainly objective limits are OK) limit for number of member devices is acceptable, but needs to be known, so that installers can inhibit or warn users from creating unbootable arrays. And then how to communicate this downstream?
GRUB has no limitation on number on devices other than free memory to
hold structures and 2^32 for IDs. If there is any other limit it's a bug
and please file a bug report with the images in question (no need to
install, just use grub-fstest) However BIOS limits the number of disks
that can be accessed. Theoretically there could be up to 128 (or 256 if
you use floppy numbers as well) accessible through BIOS. The convention
is to use only 16 possible IDs for HDD (0x80-0x8f). In practice many
BIOSes are limited to less (8 is usual limit).
>
>
> Chris Murphy
> _______________________________________________
> Grub-devel mailing list
> Grub-devel@gnu.org
> https://lists.gnu.org/mailman/listinfo/grub-devel
>
--
Regards
Vladimir 'φ-coder/phcoder' Serbinenko
[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 294 bytes --]
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: bootable RAID, number of member disks limitation
2013-01-03 22:00 ` Vladimir 'φ-coder/phcoder' Serbinenko
@ 2013-01-04 7:42 ` Chris Murphy
2013-01-04 15:18 ` Chris Murphy
2013-01-08 23:37 ` Jordan Uggla
0 siblings, 2 replies; 5+ messages in thread
From: Chris Murphy @ 2013-01-04 7:42 UTC (permalink / raw)
To: The development of GNU GRUB
On Jan 3, 2013, at 3:00 PM, Vladimir 'φ-coder/phcoder' Serbinenko <phcoder@gmail.com> wrote:
> GRUB has no limitation on number on devices other than free memory to
> hold structures and 2^32 for IDs. If there is any other limit it's a bug
> and please file a bug report with the images in question (no need to
> install, just use grub-fstest) However BIOS limits the number of disks
> that can be accessed. Theoretically there could be up to 128 (or 256 if
> you use floppy numbers as well) accessible through BIOS. The convention
> is to use only 16 possible IDs for HDD (0x80-0x8f). In practice many
> BIOSes are limited to less (8 is usual limit).
I'm totally unclear on how to use grub-fstest.
Should ls at grub rescue report all attached devices?
With single disk btrfs only (no md RAID), at a grub command prompt, ls reports
(hd0) (hd0,msdos1) (hd1) (hd2) (hd3)
Yet there are 11 devices attached. This is VirtualBox, so this very well may be a vbox limitation. I haven't tried it with KVM's SEABIOS yet.
With md RAID only (no Btrfs), ls reports:
(md/root) (hd0) (hd0,msdos1) (hd1) (hd1,msdos1) (hd2) (hd1,msdos2)
Chris Murphy
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: bootable RAID, number of member disks limitation
2013-01-04 7:42 ` Chris Murphy
@ 2013-01-04 15:18 ` Chris Murphy
2013-01-08 23:37 ` Jordan Uggla
1 sibling, 0 replies; 5+ messages in thread
From: Chris Murphy @ 2013-01-04 15:18 UTC (permalink / raw)
To: The development of GNU GRUB
On Jan 4, 2013, at 12:42 AM, Chris Murphy <chris@colorremedies.com> wrote:
> With md RAID only (no Btrfs), ls reports:
> (md/root) (hd0) (hd0,msdos1) (hd1) (hd1,msdos1) (hd2) (hd1,msdos2)
Also with 11 devices attached.
Chris Murphy
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: bootable RAID, number of member disks limitation
2013-01-04 7:42 ` Chris Murphy
2013-01-04 15:18 ` Chris Murphy
@ 2013-01-08 23:37 ` Jordan Uggla
1 sibling, 0 replies; 5+ messages in thread
From: Jordan Uggla @ 2013-01-08 23:37 UTC (permalink / raw)
To: The development of GNU GRUB
On Thu, Jan 3, 2013 at 11:42 PM, Chris Murphy <chris@colorremedies.com> wrote:
>
> On Jan 3, 2013, at 3:00 PM, Vladimir 'φ-coder/phcoder' Serbinenko <phcoder@gmail.com> wrote:
>
>> GRUB has no limitation on number on devices other than free memory to
>> hold structures and 2^32 for IDs. If there is any other limit it's a bug
>> and please file a bug report with the images in question (no need to
>> install, just use grub-fstest) However BIOS limits the number of disks
>> that can be accessed. Theoretically there could be up to 128 (or 256 if
>> you use floppy numbers as well) accessible through BIOS. The convention
>> is to use only 16 possible IDs for HDD (0x80-0x8f). In practice many
>> BIOSes are limited to less (8 is usual limit).
>
> I'm totally unclear on how to use grub-fstest.
I don't have a GNU/Linux system (other than my phone) to test this
with at the moment, but I think "grub-fstest --diskcount=11
/dev/sd{a..k} ls -- -l" should list all 11 disks, and their
partitions, with information about the filesystem each contains. If
(loop0,msdos1) is one of the devices that makes up this btrfs volume
then you should be able to list files on the filesystem with:
grub-fstest --diskcount=11 /dev/sd{a..k} ls '(loop0,msdos1)/'
>
> Should ls at grub rescue report all attached devices?
Yes, if it doesn't then that's a pretty clear indication that your
boot firmware can't properly handle many drives. You can also try
"grub-install --disk-module=native /dev/sdX" to test if grub can
access all of the drives using its own native disk drivers rather than
relying on the boot firmware.
>
> With single disk btrfs only (no md RAID), at a grub command prompt, ls reports
> (hd0) (hd0,msdos1) (hd1) (hd2) (hd3)
>
> Yet there are 11 devices attached. This is VirtualBox, so this very well may be a vbox limitation. I haven't tried it with KVM's SEABIOS yet.
I am surprised to hear that this problem is appearing with VirtualBox
rather than real hardware. If this is a bug in VirtualBox's BIOS then
a bug report should definitely be filed with them.
--
Jordan Uggla (Jordan_U on irc.freenode.net)
^ permalink raw reply [flat|nested] 5+ messages in thread
end of thread, other threads:[~2013-01-08 23:38 UTC | newest]
Thread overview: 5+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2013-01-03 20:28 bootable RAID, number of member disks limitation Chris Murphy
2013-01-03 22:00 ` Vladimir 'φ-coder/phcoder' Serbinenko
2013-01-04 7:42 ` Chris Murphy
2013-01-04 15:18 ` Chris Murphy
2013-01-08 23:37 ` Jordan Uggla
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.