Re: Problem with Grub

[Felix Miata]

David Wright composed on 2021-06-22 11:00 (UTC-0500):

> On Sat 19 Jun 2021 at 14:51:53 (-0400), Felix Miata wrote:

>> As already mentioned, you needn't have paid. Two unrelated things were 
>> likely to
>> have caused this to happen. Vista and or Win10 could have been installed in 
>> MBR
>> mode, which is compatible with having Grub on the MBR, but the failure could 
>> have
>> been that os-prober was either not installed or not configured to include 
>> any menu
>> option for Windows.

> Would I be right in assuming that you could still manually boot
> Windows at this point by using Grub's command line, along the
> following lines:

> GRUB> insmod fat
> GRUB> set root=(hd0,gpt1)
> GRUB> chainloader (${root})/efi/Microsoft/Boot/bootmgfw.efi
> GRUB> boot

> (That's the sequence I used when I had to do this myself, and
> I originally navigated my way there by means of Grub's TAB
> command-completion.)


Sure. Grub menus are scripts that call some of the same commands available while
at a Grub shell prompt.

>> The other likely failure mode could have been that either Windows could have 
>> been
>> installed in UEFI mode, while Debian was installed in legacy (MBR) mode. All
>> installed operating systems on UEFI PCs need to be installed in the same 
>> mode to
>> permit Grub to be able to boot all of them.

> Presumably if they had managed to install Debian/BIOS on a
> Windows/UEFI computer, then they ought to be able to switch between
> these modes at will. That's how I installed Debian on a machine where
> my wife still needed to be able to boot Windows. (No chance of being
> permitted to alter the booting scheme while that need remained.)


Technically it's /not/ needed for all to be in same mode, but /when/ not, and
presuming firmware competence, navigating the PC firmware to some extent is
required for each instance of need to boot an OS that has been installed to
another mode. The "need" has to do with using installed software, usually Grub 
on
x86 compatible hardware on which Gnu/Linux has been installed, to select which 
OS
to boot. Hence, the "need" applies in a Grub in maximum control context.
-- 
Evolution as taught in public schools is, like religion,
based on faith, not based on science.

 Team OS/2 ** Reg. Linux User #211409 ** a11y rocks!

Felix Miata

Re: Problem with Grub

[David Wright]

On Sat 19 Jun 2021 at 14:51:53 (-0400), Felix Miata wrote:
> William Lee Valentine composed on 2021-06-19 10:19 (UTC-0600):
> 
> > I had installed a back version of Debian in a partition on a
> > 500-megahertz computer that was otherwise running Windows 2000 and
> > MS-DOS.
> 
> > When I had finished installing Linux, on that machine, Grub wanted to
> > know whether I wanted it installed in the master boot record. It
> > reported seeing Windows 2000 and MS-DOS in other bootable partitions. I
> > agreed. Grub has allowed me to boot any of the three of these when the
> > BIOS has executed.
> 
> > I later installed Debian 10.2 in a partition on a 64-bit computer that
> > was otherwise running Windows 10.
> 
> > When I had finished installing Linux, Grub wanted to know whether I
> > wanted it installed on the master boot record. It reported seeing
> > "Windows Vista" in another bootable partition. I agreed. This time,
> > however, Grub modified the master boot record to allow only Linux to be
> > booted. I had to pay to have Windows 10 reinstalled.
>   
> As already mentioned, you needn't have paid. Two unrelated things were likely 
> to
> have caused this to happen. Vista and or Win10 could have been installed in 
> MBR
> mode, which is compatible with having Grub on the MBR, but the failure could 
> have
> been that os-prober was either not installed or not configured to include any 
> menu
> option for Windows.

Would I be right in assuming that you could still manually boot
Windows at this point by using Grub's command line, along the
following lines:

GRUB> insmod fat
GRUB> set root=(hd0,gpt1)
GRUB> chainloader (${root})/efi/Microsoft/Boot/bootmgfw.efi
GRUB> boot

(That's the sequence I used when I had to do this myself, and
I originally navigated my way there by means of Grub's TAB
command-completion.)

> The other likely failure mode could have been that either Windows could have 
> been
> installed in UEFI mode, while Debian was installed in legacy (MBR) mode. All
> installed operating systems on UEFI PCs need to be installed in the same mode 
> to
> permit Grub to be able to boot all of them.

Presumably if they had managed to install Debian/BIOS on a
Windows/UEFI computer, then they ought to be able to switch between
these modes at will. That's how I installed Debian on a machine where
my wife still needed to be able to boot Windows. (No chance of being
permitted to alter the booting scheme while that need remained.)

> > I tried1 again, this time avoiding the master boot record entirely. I
> > asked that Grub install itself on a 3.5" diskette (in a USB floppy
> > drive). It did not. It installed itself instead on the master boot
> > record, again allowing only Linux to be booted. Again I had to pay to
> > have Windows 10 reinstalled.
> 
> > I have now another 64-bit computer, running Windows 10, whose BIOS
> > provides the option of booting from a USB device. If I install Debian
> > 10.2 in a partition on this computer, would I tell Grub to make the
> > partition bootable? Would Grub instead install itself on the master boot
> > record anyway, allowing only Linux to be booted? I can not afford to
> > lose access to Windows 10 again.
>   
> Newer PCs that come with Windows already installed usually have it installed 
> in
> UEFI mode. That eliminates any relevance of the MBR to the boot process. If 
> you
> boot the Debian installation media in UEFI mode, then it will install in UEFI
> mode. If the installation fails to include os-prober and include a menu 
> option for
> Windows, it can be added at any later time. Windows itself need not be 
> touched,
> though it needs special configuration (disable fast boot) in order to permit
> Debian to be able to access files on Windows filesystems.
> 
> I recommend having Debian reuse (use it without reformatting it) the ESP 
> partition
> that UEFI Windows uses for booting, rather than creating a separate ESP 
> partition
> used only for Debian. The seemingly small size of the Windows ESP is a 
> non-issue
> in most cases (boot disk smaller than 16TB). Usually it is 100MB, which is 
> plenty
> for Debian times 10 or 20 separate installations.
> 
> It's usually necessary or at least desirable on a Windows 10 intended PC to 
> change
> a BIOS setting intended to inhibit booting from any operating system other 
> than
> Windows. Change it from Windows to Other.

Cheers,
David.

Re: Problem with Grub

[Joe]

On Sat, 19 Jun 2021 10:19:57 -0600
William Lee Valentine  wrote:


> 
> I later installed Debian 10.2 in a partition on a 64-bit computer that
> was otherwise running Windows 10.
> 
> When I had finished installing Linux, Grub wanted to know whether I
> wanted it installed on the master boot record. It reported seeing
> "Windows Vista" in another bootable partition. I agreed. This time,
> however, Grub modified the master boot record to allow only Linux to
> be booted. I had to pay to have Windows 10 reinstalled.
> 
Apart from other answers to this, anything with Win10 originally
installed (and you don't say this, it might be an upgrade, perhaps from
Vista) will have a recovery partition which can be triggered from a key
during boot. OK, it completely nukes the original installation,
returning the computer to factory-issued, but it costs nothing.

Another point: buster will install dual-boot to a UEFI Win10, but on my
netbook, will not dual-boot afterwards. It may still be possible to
dual-boot using the startup boot menu key, not GRUB, though it isn't on
my machine.

-- 
Joe

Re: Problem with Grub

[Felix Miata]

William Lee Valentine composed on 2021-06-19 10:19 (UTC-0600):

> I had installed a back version of Debian in a partition on a
> 500-megahertz computer that was otherwise running Windows 2000 and
> MS-DOS.

> When I had finished installing Linux, on that machine, Grub wanted to
> know whether I wanted it installed in the master boot record. It
> reported seeing Windows 2000 and MS-DOS in other bootable partitions. I
> agreed. Grub has allowed me to boot any of the three of these when the
> BIOS has executed.

> I later installed Debian 10.2 in a partition on a 64-bit computer that
> was otherwise running Windows 10.

> When I had finished installing Linux, Grub wanted to know whether I
> wanted it installed on the master boot record. It reported seeing
> "Windows Vista" in another bootable partition. I agreed. This time,
> however, Grub modified the master boot record to allow only Linux to be
> booted. I had to pay to have Windows 10 reinstalled.

As already mentioned, you needn't have paid. Two unrelated things were likely to
have caused this to happen. Vista and or Win10 could have been installed in MBR
mode, which is compatible with having Grub on the MBR, but the failure could 
have
been that os-prober was either not installed or not configured to include any 
menu
option for Windows.

The other likely failure mode could have been that either Windows could have 
been
installed in UEFI mode, while Debian was installed in legacy (MBR) mode. All
installed operating systems on UEFI PCs need to be installed in the same mode to
permit Grub to be able to boot all of them.

> I tried1 again, this time avoiding the master boot record entirely. I
> asked that Grub install itself on a 3.5" diskette (in a USB floppy
> drive). It did not. It installed itself instead on the master boot
> record, again allowing only Linux to be booted. Again I had to pay to
> have Windows 10 reinstalled.

> I have now another 64-bit computer, running Windows 10, whose BIOS
> provides the option of booting from a USB device. If I install Debian
> 10.2 in a partition on this computer, would I tell Grub to make the
> partition bootable? Would Grub instead install itself on the master boot
> record anyway, allowing only Linux to be booted? I can not afford to
> lose access to Windows 10 again.

Newer PCs that come with Windows already installed usually have it installed in
UEFI mode. That eliminates any relevance of the MBR to the boot process. If you
boot the Debian installation media in UEFI mode, then it will install in UEFI
mode. If the installation fails to include os-prober and include a menu option 
for
Windows, it can be added at any later time. Windows itself need not be touched,
though it needs special configuration (disable fast boot) in order to permit
Debian to be able to access files on Windows filesystems.

I recommend having Debian reuse (use it without reformatting it) the ESP 
partition
that UEFI Windows uses for booting, rather than creating a separate ESP 
partition
used only for Debian. The seemingly small size of the Windows ESP is a non-issue
in most cases (boot disk smaller than 16TB). Usually it is 100MB, which is 
plenty
for Debian times 10 or 20 separate installations.

It's usually necessary or at least desirable on a Windows 10 intended PC to 
change
a BIOS setting intended to inhibit booting from any operating system other than
Windows. Change it from Windows to Other.
-- 
Evolution as taught in public schools is, like religion,
based on faith, not based on science.

 Team OS/2 ** Reg. Linux User #211409 ** a11y rocks!

Felix Miata

Re: Problem with Grub

[Douglas McGarrett]

On 6/19/21 12:19 PM, William Lee Valentine wrote:

I had installed a back version of Debian in a partition on a
500-megahertz computer that was otherwise running Windows 2000 and
MS-DOS.

When I had finished installing Linux, on that machine, Grub wanted to
know whether I wanted it installed in the master boot record. It
reported seeing Windows 2000 and MS-DOS in other bootable partitions. I
agreed. Grub has allowed me to boot any of the three of these when the
BIOS has executed.

I later installed Debian 10.2 in a partition on a 64-bit computer that
was otherwise running Windows 10.

When I had finished installing Linux, Grub wanted to know whether I
wanted it installed on the master boot record. It reported seeing
"Windows Vista" in another bootable partition. I agreed. This time,
however, Grub modified the master boot record to allow only Linux to be
booted. I had to pay to have Windows 10 reinstalled.

I tried again, this time avoiding the master boot record entirely. I
asked that Grub install itself on a 3.5" diskette (in a USB floppy
drive). It did not. It installed itself instead on the master boot
record, again allowing only Linux to be booted. Again I had to pay to
have Windows 10 reinstalled.

I have now another 64-bit computer, running Windows 10, whose BIOS
provides the option of booting from a USB device. If I install Debian
10.2 in a partition on this computer, would I tell Grub to make the
partition bootable? Would Grub instead install itself on the master boot
record anyway, allowing only Linux to be booted? I can not afford to
lose access to Windows 10 again.

Thank you for your assistance.

-- William Lee Valentine


I suppose it's kinda late for this advice, but you did not have to pay 
twice for Windows.
You should have just ordered a disk or a USB stick with Windows on it. 
Install it--if
it's a newer computer, it probably has the windows code number in the 
BIOS, but
even if it doesn't, you should be able to install Windows any number of 
times from the

one source device, ON THE SAME COMPUTER. I don't know what you mean by
"paying to have it installed"--- you install it yourself from the 
software you have paid
for ONCE. It is not difficult, only time consuming, and you may have to 
try a couple of

times to get the earliest parts the way you want them. Then install Linux.
(I haven't installed Debian, but I've installed a number of Linux 
systems, both deb and KDE

over Windows, and usually it "just works.")
--doug

Re: Problem with Grub

[Eduardo M KALINOWSKI]

On 19/06/2021 13:19, William Lee Valentine wrote:

I have now another 64-bit computer, running Windows 10, whose BIOS
provides the option of booting from a USB device. If I install Debian
10.2 in a partition on this computer, would I tell Grub to make the
partition bootable? Would Grub instead install itself on the master boot
record anyway, allowing only Linux to be booted? I can not afford to
lose access to Windows 10 again.

You probably need the os-prober package installed so that other OS's are 
detected and included in the boot menu.



--
BOFH excuse #206:

Police are examining all internet packets in the search for a 
narco-net-trafficker

Eduardo M KALINOWSKI
edua...@kalinowski.com.br

Re: Problem with Grub

[IL Ka]

Hello.

You do not need to reinstall Windows to fix boot problems. Download Windows
iso from the Microsoft website along with a special tool to burn it to the
flash drive, boot from it and Windows will repair most booting issues.

Motherboard firmware may act in Legacy/BIOS mode or UEFI mode.

Master boot record is not used in UEFI mode.
Which mode do you use?

Problem with Grub

[William Lee Valentine]

I had installed a back version of Debian in a partition on a
500-megahertz computer that was otherwise running Windows 2000 and
MS-DOS.

When I had finished installing Linux, on that machine, Grub wanted to
know whether I wanted it installed in the master boot record. It
reported seeing Windows 2000 and MS-DOS in other bootable partitions. I
agreed. Grub has allowed me to boot any of the three of these when the
BIOS has executed.

I later installed Debian 10.2 in a partition on a 64-bit computer that
was otherwise running Windows 10.

When I had finished installing Linux, Grub wanted to know whether I
wanted it installed on the master boot record. It reported seeing
"Windows Vista" in another bootable partition. I agreed. This time,
however, Grub modified the master boot record to allow only Linux to be
booted. I had to pay to have Windows 10 reinstalled.

I tried again, this time avoiding the master boot record entirely. I
asked that Grub install itself on a 3.5" diskette (in a USB floppy
drive). It did not. It installed itself instead on the master boot
record, again allowing only Linux to be booted. Again I had to pay to
have Windows 10 reinstalled.

I have now another 64-bit computer, running Windows 10, whose BIOS
provides the option of booting from a USB device. If I install Debian
10.2 in a partition on this computer, would I tell Grub to make the
partition bootable? Would Grub instead install itself on the master boot
record anyway, allowing only Linux to be booted? I can not afford to
lose access to Windows 10 again.

Thank you for your assistance.

-- William Lee Valentine


--
This email has been checked for viruses by Avast antivirus software.
https://www.avast.com/antivirus

Solved: Problem with grub-install (d-i, soft RAID6, Jessie)

[W. Martin Borgert]

Many thanks to Christian, David, and Pascal!
With your hints I was able to solve the problems.

Lessons learned:

 1. d-i needs to be in "low" questions mode (expert) to allow
manual selection of partition type (e.g. DOS/MBR or GPT).

 2. d-i seems to call grub-install on /dev/md0, even if user
explicitely selects /dev/sdXN - one need to select the
manual option and type the right device names. Or go to
shell and grub-install --force manually.

 3. Changing the partition type of one disk in RAID6 is easy: If
all disks are OK, it is relatively safe to just remove one
disk (mdadm fail + remove), copy the partition table from
another disk (in case of MBR it can be done with sfdisk) and
add the removed disk again. It gets synced very quickly.

Cheers

Re: Problem with grub-install (d-i, soft RAID6, Jessie)

[Pascal Hambourg]

Christian Seiler a écrit :
> On 01/08/2016 02:51 AM, Pascal Hambourg wrote:
>> Current tools won't align for compatibility purpose any more. The
>> default 1-MiB alignment is only for performance purpose, not compatibility.
> 
> This is new to me - I always thought that this was entirely for
> compatibility. I'm curious: How would that help performance-wise?
> On SSDs it shouldn't matter at all - and if you have rotational
> media, you'd need to figure out the physical (not the logical)
> geometry of the drive to be effective - and if I'm not mistaken
> there's no standardized way of getting to that via software. What
> am I missing here?

The real physical geometry does not matter at all. What matters is only
the physical read/write/erase block size. You get a read-modify-write
penalty if you don't align write operations on proper block boundaries.
1 MiB was chosen as a multiple of all common block sizes for both SSD
and 512e "Advanced  Format" hard disks.

See for example

Re: Problem with grub-install (d-i, soft RAID6, Jessie)

[Pascal Hambourg]

W. Martin Borgert a écrit :
> 
> I try to install Jessie with a soft RAID6 of four disks, 1 TB each.
> The idea is to use the complete disk for RAID and then only on top
> separate swap and root.
> 
> The installation worked fine on one machine, but for some reason,
> grub does not install on the fourth disk. It turns out, that the
> fourth disk is GPT, while the others are DOS (= MBR, right?). Also,
> the fourth disks only partition is of type "Linux RAID", while the
> partitions of the other three disks are "Linux raid autodetect".

On a GPT disk, grub-install requires a small "BIOS boot" partition (less
than 100 kB is enough) or --force to use block lists.

> AFAIK, I don't need GPT for such small disks, right?

GPT is not required for 1 TB but has other advantages.

> Is there an easy and safe way to change the disk from GPT to DOS/MBR?

Maybe gdisk can do that.

Re: Problem with grub-install (d-i, soft RAID6, Jessie)

[Christian Seiler]

On 01/08/2016 02:03 AM, Pascal Hambourg wrote:
> Christian Seiler a écrit :
>> Problem is that GPT doesn't have such a gap.
> 
> GPT usually also has such a gap for partition alignment purpose.

Well sure, but MBR also has all the CHS baggage - and while it isn't
used anymore on modern systems, there still is the de-factor standard
of aligning the first partition with the first CHS-sector that is so
ingrained historically that the gap between MBR and the first
partition is something that may be relied on.

Whereas with GPT, which only uses LBA from the get-go, you could in
principle create a partition at the first block after the end of the
partition table, so on 512-byte/"LBA-sector" disks with the minimum
partition table size, that'd be LBA #34 - with no gap at all.

Or to put it differently: because of the CHS baggage of MBR, there
is software that wouldn't support it if the first partition wasn't
aligned; for compatibility reasons every tool that manipulates MBR
partitions will therefore want to leave a gap after the MBR. And
while many GPT tools will also leave a gap for alignment purposes,
this isn't required anymore, and any conforming GPT implementation
will cope with non-"aligned" GPT partitions.

So yes, my statement was an oversimplification - but it is not
unreasonable to have a gapless disk with GPT - hence the
explanation why grub doesn't use this gap anymore. (Because you
could also create a partition there afterwards.)

>> At least Jessie's d-i appears to only be able to create MBR
>> partition tables
> 
> That's wrong. I have created GPT partition tables with Wheezy's
> installer. IIRC, it creates a GPT partition table by default when booted
> in EFI mode.

I didn't try EFI mode recently. But in legacy BIOS mode, I've found
no option of making the installer create a GPT table - I actually
had to use a shell to do it manually. (If there is an option, it's
*too* well hidden.)

Regards,
Christian



signature.asc
Description: OpenPGP digital signature

Problem with grub-install (d-i, soft RAID6, Jessie)

[W. Martin Borgert]

Hi,

I try to install Jessie with a soft RAID6 of four disks, 1 TB each.
The idea is to use the complete disk for RAID and then only on top
separate swap and root.

The installation worked fine on one machine, but for some reason,
grub does not install on the fourth disk. It turns out, that the
fourth disk is GPT, while the others are DOS (= MBR, right?). Also,
the fourth disks only partition is of type "Linux RAID", while the
partitions of the other three disks are "Linux raid autodetect".

On a second machine, supposed to be identical, grub-install failed
on all disks, but I could not yet check, whether GPT is the culprit,
too.

Questions:

AFAIK, I don't need GPT for such small disks, right?

Is there an easy and safe way to change the disk from GPT to DOS/MBR?

Who did set up the disks in that heterogenous way? d-i? Why?

Thanks in advance!


PS: Last time I had a similar install was with etch or lenny, IIRC.
Despite the issue I face, d-i made so much of progress in that area.
Wonderful!

Re: Problem with grub-install (d-i, soft RAID6, Jessie)

[Christian Seiler]

On 01/08/2016 01:22 AM, W. Martin Borgert wrote:
> I try to install Jessie with a soft RAID6 of four disks, 1 TB each.
> The idea is to use the complete disk for RAID and then only on top
> separate swap and root.
> 
> The installation worked fine on one machine, but for some reason,
> grub does not install on the fourth disk. It turns out, that the
> fourth disk is GPT, while the others are DOS (= MBR, right?). Also,
> the fourth disks only partition is of type "Linux RAID", while the
> partitions of the other three disks are "Linux raid autodetect".

Ok, if you use MBR, only the first 512 bytes are used for the MBR
partition table, so that the gap between those bytes and the first
partition (and there is *always* a gap, the size depends on the
details of the logical geometry of the disk) has traditionally
been used under Linux (among others) to put the boot loader there.

Problem is that GPT doesn't have such a gap. If you don't use UEFI,
you can create a small GPT partition (1 - 2 Megabytes is
sufficient) of the type "BIOS boot partition" - the boot loader may
then be installed there (if it exists).

If you already have partitioned that disk, it's unclear whether you
can actually fit another partition somewhere. (Note that if the
disk is small enough, with a modern BIOS less than 2 TB, the BIOS
boot partition doesn't have to be at the beginning of the disk, it
could also be at the end.)

So if there is some small amount of space left, just create a new
partition of type "BIOS boot partition" on the GPT disk and then
grub-install should work. Would be the easiest solution.

> AFAIK, I don't need GPT for such small disks, right?

Not necessarily, no. ~2TB is the limit for MBR.

> Is there an easy and safe way to change the disk from GPT to DOS/MBR?

Well, if you can get the exact position of your GPT partition
(start and size), you could wipe the partition table and then
create a new MBR partition table with the same boundaries. I
wouldn't call that safe (because you have to be very careful and
be aware of the units in which you have to specify this) - and
then the MBR partitions may not be aligned properly, which is
probably not a huge issue if you have a Linux-only system
(because Linux + GRUB can cope with that and no other code reads
the partition table), but I'd still be a bit wary of it.

So while possible, I wouldn't classify it as "safe and easy".

OTOH, it's RAID6, so worst case scenario you have to do a bit of
resyncing. ;-)

(The other way round is easier, because you can just use gdisk
that can automatically convert MBR to GPT in most cases.)

> Who did set up the disks in that heterogenous way? d-i? Why?

At least Jessie's d-i appears to only be able to create MBR
partition tables - but if there already is a partition table on
the disk, then (if I remember correctly) it will also use a GPT
table for installation.

So what likely happened is that your disk with GPT already had
a GPT partition table on it, while the others either had a
traditional MBR partition table or none at all - and that the
installer then reused the GPT of the last disk, while reusing
and/or creating the MBRs of the other three disks.

Regards,
Christian



signature.asc
Description: OpenPGP digital signature

Re: Problem with grub-install (d-i, soft RAID6, Jessie)

[Pascal Hambourg]

Christian Seiler a écrit :
> 
> Ok, if you use MBR, only the first 512 bytes are used for the MBR
> partition table, so that the gap between those bytes and the first
> partition (and there is *always* a gap, the size depends on the
> details of the logical geometry of the disk) has traditionally
> been used under Linux (among others) to put the boot loader there.
> 
> Problem is that GPT doesn't have such a gap.

GPT usually also has such a gap for partition alignment purpose. But
GRUB won't use it.
The gap is currently about 1 MiB regardless of the partition table type.

> At least Jessie's d-i appears to only be able to create MBR
> partition tables

That's wrong. I have created GPT partition tables with Wheezy's
installer. IIRC, it creates a GPT partition table by default when booted
in EFI mode.

Re: Problem with grub-install (d-i, soft RAID6, Jessie)

[Christian Seiler]

On 01/08/2016 02:51 AM, Pascal Hambourg wrote:
> Christian Seiler a écrit :
>> for compatibility reasons every tool that manipulates MBR
>> partitions will therefore want to leave a gap after the MBR.
> 
> Current tools won't align for compatibility purpose any more. The
> default 1-MiB alignment is only for performance purpose, not compatibility.

This is new to me - I always thought that this was entirely for
compatibility. I'm curious: How would that help performance-wise?
On SSDs it shouldn't matter at all - and if you have rotational
media, you'd need to figure out the physical (not the logical)
geometry of the drive to be effective - and if I'm not mistaken
there's no standardized way of getting to that via software. What
am I missing here?

Regards,
Christian



signature.asc
Description: OpenPGP digital signature

Re: Problem with grub-install (d-i, soft RAID6, Jessie)

[David Christensen]

On 01/07/2016 04:22 PM, W. Martin Borgert wrote:

I try to install Jessie with a soft RAID6 of four disks, 1 TB each.
The idea is to use the complete disk for RAID and then only on top
separate swap and root.  ...
grub does not install on the fourth disk. It turns out, that the
fourth disk is GPT, while the others are DOS (= MBR, right?). Also,
the fourth disks only partition is of type "Linux RAID", while the
partitions of the other three disks are "Linux raid autodetect". ...


I prefer to put /boot, swap, and root on a single, small, fast SSD, and 
set up additional volumes as needed for /home, bulk data, or whatever. 
In addition to performance considerations, having small system drives 
allows me to take and restore images quickly, and archive lots of them 
without eating up too much space.




AFAIK, I don't need GPT for such small disks, right?


MBR can accommodate ~2 TB drives, but GPT has additional features that 
you might like.




Who did set up the disks in that heterogenous way? d-i?


Did you wipe all four drives, or at least the first and last 1 MB, 
before installing Debian?



David

Re: Problem with grub-install (d-i, soft RAID6, Jessie)

[Pascal Hambourg]

Christian Seiler a écrit :
> 
> Well sure, but MBR also has all the CHS baggage - and while it isn't
> used anymore on modern systems, there still is the de-factor standard
> of aligning the first partition with the first CHS-sector that is so
> ingrained historically that the gap between MBR and the first
> partition is something that may be relied on.

Even in the MBR/MSDOS partition table scheme, the alignment is not based
on CHS pseudo-geometry any more.

> Or to put it differently: because of the CHS baggage of MBR, there
> is software that wouldn't support it if the first partition wasn't
> aligned;

Only very old software.

> for compatibility reasons every tool that manipulates MBR
> partitions will therefore want to leave a gap after the MBR.

Current tools won't align for compatibility purpose any more. The
default 1-MiB alignment is only for performance purpose, not compatibility.

> So yes, my statement was an oversimplification - but it is not
> unreasonable to have a gapless disk with GPT - hence the
> explanation why grub doesn't use this gap anymore. (Because you
> could also create a partition there afterwards.)

Or grow the partition table and overwrite the gap.

>>> At least Jessie's d-i appears to only be able to create MBR
>>> partition tables
>> That's wrong. I have created GPT partition tables with Wheezy's
>> installer. IIRC, it creates a GPT partition table by default when booted
>> in EFI mode.
> 
> I didn't try EFI mode recently. But in legacy BIOS mode, I've found
> no option of making the installer create a GPT table - I actually
> had to use a shell to do it manually. (If there is an option, it's
> *too* well hidden.)

Maybe you have to select "expert install" at the boot menu.

Re: problem installing grub-legacy

[Freeman]

On Sun, Mar 28, 2010 at 05:25:34AM +, T o n g wrote:
 hi,
 
 Long story, I don't want Grub2/Grub-PC, etc, but get upgraded to it. Now 
 I can't go back, because grub-common is listed as a dependency of grub-
 legacy. But the grub-common package has been upgraded for use with 
 version 2 only.
 

I posted this past week about grub-common confusion that turned a simple
legacy install into a long session.  Grub-common rides in with legacy and
makes an /etc/grub.d directory and scatters files all over the /boot/grub
directory, including grub.cfg .  But then it just sits there and tries to
look like Grub-PC is installed.

My system still boots off /boot/grub/menu.lst, kernel updates and update-grub 
still
modify menu.lst, ^c still drops into a grub 0.97 terminal from the boot
menu.

I did another legacy install last night.  I do think I remember getting
Grub-PC in some dependency resolution, but I was able to remove it and
install legacy without forcing anything.

-- 
Kind Regards,
Freeman

http://bugs.debian.org/release-critical/


-- 
To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org 
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Archive: http://lists.debian.org/20100329075704.ga4...@europa.office

Conclusion: problem installing grub-legacy

[T o n g]

On Mon, 29 Mar 2010 00:57:04 -0700, Freeman wrote:

 I don't want Grub2/Grub-PC, etc, but get upgraded to it. Now
 I can't go back, because grub-common is listed as a dependency of grub-
 legacy. But the grub-common package has been upgraded for use with
 version 2 only.
 
 I posted this past week about grub-common confusion that turned a simple
 legacy install into a long session.  Grub-common rides in with legacy
 and makes an /etc/grub.d directory and scatters files all over the
 /boot/grub directory, including grub.cfg .  But then it just sits there
 and tries to look like Grub-PC is installed.

Yes, I saw that. In fact, you can see that my words are borrowed from 
yours, :-) Just that I wasn't sure of the solution.

 My system still boots off /boot/grub/menu.lst, kernel updates and
 update-grub still modify menu.lst, ^c still drops into a grub 0.97
 terminal from the boot menu.
 
 I did another legacy install last night.  I do think I remember getting
 Grub-PC in some dependency resolution, but I was able to remove it and
 install legacy without forcing anything.

OK, thanks for the explanation. 

FYI, I hate grub-common messing up with my /boot/grub and /etc/grub.d 
directory, so I'll use 'equivs' to create a dummy packages to circumvent 
such silly package dependencies. 

 equivs - Circumvent Debian package dependencies

This package provides a tool to create Debian packages that only
contain dependency information.

HTH

-- 
Tong (remove underscore(s) to reply)
  http://xpt.sourceforge.net/techdocs/
  http://xpt.sourceforge.net/tools/


-- 
To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org 
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Archive: http://lists.debian.org/hoqjia$os...@dough.gmane.org

Re: Conclusion: problem installing grub-legacy

[Freeman]

On Mon, Mar 29, 2010 at 04:12:58PM +, T o n g wrote:
 On Mon, 29 Mar 2010 00:57:04 -0700, Freeman wrote:
 
  I don't want Grub2/Grub-PC, etc, but get upgraded to it. Now
  I can't go back, because grub-common is listed as a dependency of grub-
  legacy. But the grub-common package has been upgraded for use with
  version 2 only.
  
  I posted this past week about grub-common confusion that turned a simple
  legacy install into a long session.  Grub-common rides in with legacy
  and makes an /etc/grub.d directory and scatters files all over the
  /boot/grub directory, including grub.cfg .  But then it just sits there
  and tries to look like Grub-PC is installed.
 
 Yes, I saw that. In fact, you can see that my words are borrowed from 
 yours, :-) Just that I wasn't sure of the solution.
 

Right, missed that. Good choice of words. :-)

...

 
 FYI, I hate grub-common messing up with my /boot/grub and /etc/grub.d 
 directory, so I'll use 'equivs' to create a dummy packages to circumvent 
 such silly package dependencies. 
 
  equivs - Circumvent Debian package dependencies
 
 This package provides a tool to create Debian packages that only
 contain dependency information.
 

Noted, and thanks. I learn more than just one thing daily on this list. 
(Which is easy given my starting point.)

-- 
Kind Regards,
Freeman

http://bugs.debian.org/release-critical/


-- 
To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org 
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Archive: http://lists.debian.org/20100329174726.ga27...@europa.office

Re: problem installing grub-legacy

[Joey Hess]

Stephen Powell wrote:
Aren't you glad you use Dial?  Don't you wish everybody did?
 
 s/Dial/lilo/

Given that I have in the past spent up to 2 days remotely walking family
members through fixing machines that booted to LI or LIL .. no.
I only wish lilo on enemies -- but still prefer enemies use soap.

 But seriously, it's obvious that somebody messed up with the package
 dependencies.  There's not much point in having a grub-legacy package
 if it can't be installed, now is there? 

Except that's is clearly not the case that grub-legacy cannot be
installed, if you actually look at the dependencies of grub-common. I
have grub-legacy installed here, works fine.

-- 
see shy jo


signature.asc
Description: Digital signature

Re: problem installing grub-legacy

[Stephen Powell]

On Mon, 29 Mar 2010 15:46:28 -0400 (EDT), Joey Hess wrote:
 Stephen Powell wrote:
Aren't you glad you use Dial?  Don't you wish everybody did?
 
 s/Dial/lilo/
 
 Given that I have in the past spent up to 2 days remotely walking family
 members through fixing machines that booted to LI or LIL .. no.
 I only wish lilo on enemies -- but still prefer enemies use soap.

Interesting.  I've been using Linux for 10 years, and I've never seen,
or even heard of, the error you describe.  I'm sorry your family members
had trouble with it.  For me, it has always worked perfectly.

 But seriously, it's obvious that somebody messed up with the package
 dependencies.  There's not much point in having a grub-legacy package
 if it can't be installed, now is there? 

 Except that's is clearly not the case that grub-legacy cannot be
 installed, if you actually look at the dependencies of grub-common. I
 have grub-legacy installed here, works fine.

The OP claimed that he could not install.  I took his word for it.
But he has subsequently found a way to circumvent his problems.
I'm glad he found a way to do what he wanted to do.  But I frankly
don't see why grub-legacy requires grub-common.  It brings in a lot
of unnecessary files, from what I can see.

-- 
  .''`. Stephen Powellzlinux...@wowway.com
 : :'  :
 `. `'`
   `-


-- 
To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org 
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Archive: 
http://lists.debian.org/1370107920.167581269895486703.javamail.r...@md01.wow.synacor.com

Re: problem installing grub-legacy

[James Brown]

T o n g wrote:
 hi,

 Long story, I don't want Grub2/Grub-PC, etc, but get upgraded to it. Now 
 I can't go back, because grub-common is listed as a dependency of grub-
 legacy. But the grub-common package has been upgraded for use with 
 version 2 only.

 Comment? 

 Thanks

   
I have such problem about two weeks with my Dedian testing working under
a virtual machine.


-- 
To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org 
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Archive: http://lists.debian.org/4baf32d6.7000...@gmail.com

Re: problem installing grub-legacy

[Stephen Powell]

On Sun, 28 Mar 2010 01:25:34 -0400 (EDT), Tong wrote:

 Long story, I don't want Grub2/Grub-PC, etc, but get upgraded to it. Now 
 I can't go back, because grub-common is listed as a dependency of grub-
 legacy. But the grub-common package has been upgraded for use with 
 version 2 only.

I don't mean to make light of your problem, but this reminds me of an
old bath soap commercial.  Many years ago, Dial soap had an advertising
slogan that went like this:

   Aren't you glad you use Dial?  Don't you wish everybody did?

s/Dial/lilo/

No doubt my friends and acquaintances on this list will accuse me of
having a lilo fixation, but this opportunity was too good to pass up.

But seriously, it's obvious that somebody messed up with the package
dependencies.  There's not much point in having a grub-legacy package
if it can't be installed, now is there?  As I see it, you have the
following options:

(1) Force grub-legacy to be installed with overrides to the package
management system, leaving you with a broken package management
system.

(2) Live with grub-pc until somebody gets their act together and fixes
this problem, or

(3) Forget the whole ugly grub mess and switch to lilo.

I know what I would do, but it's your call.

-- 
  .''`. Stephen Powellzlinux...@wowway.com
 : :'  :
 `. `'`
   `-


-- 
To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org 
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Archive: 
http://lists.debian.org/1004221449.22262471269833008563.javamail.r...@md01.wow.synacor.com

problem installing grub-legacy

[T o n g]

hi,

Long story, I don't want Grub2/Grub-PC, etc, but get upgraded to it. Now 
I can't go back, because grub-common is listed as a dependency of grub-
legacy. But the grub-common package has been upgraded for use with 
version 2 only.

Comment? 

Thanks

-- 
Tong (remove underscore(s) to reply)
  http://xpt.sourceforge.net/techdocs/
  http://xpt.sourceforge.net/tools/


-- 
To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org 
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Archive: http://lists.debian.org/homp8e$fa...@dough.gmane.org

Problem about grub automatically changes my root device

[Niu Kun]

Dear all,

I've just upgrade my sarge system to lenny. We now encountered a problem.
Our original kernel is 2.4.27 and our present kernel is 2.6.26.
Our main SATA disk is recognized as hdd at the time of installation.
But in lenny, it is recoginzed as sda.
I find that each time I install a kernel image package, it still 
automatically changes the sda, which is adjusted as my system's 
requirement by myself, to hdd, which is the original main device number.
I can fix this manually each time I install a new kernel package. But 
how can I make it automatically?

Any suggestions would be appreciated and thanks in advance.

Regards,
Kun


--
To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org 
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org

Re: Problem about grub automatically changes my root device

[Boyd Stephen Smith Jr.]

On Friday 02 October 2009 15:41:11 Niu Kun wrote:
 I've just upgrade my sarge system to lenny. We now encountered a problem.
 Our original kernel is 2.4.27 and our present kernel is 2.6.26.
 Our main SATA disk is recognized as hdd at the time of installation.
 But in lenny, it is recoginzed as sda.
 I find that each time I install a kernel image package, it still
 automatically changes the sda, which is adjusted as my system's
 requirement by myself, to hdd, which is the original main device number.
 I can fix this manually each time I install a new kernel package. But
 how can I make it automatically?

Read the documentation in /boot/grub/menu.lst
-- 
Boyd Stephen Smith Jr.   ,= ,-_-. =.
b...@iguanasuicide.net  ((_/)o o(\_))
ICQ: 514984 YM/AIM: DaTwinkDaddy `-'(. .)`-'
http://iguanasuicide.net/\_/



signature.asc
Description: This is a digitally signed message part.

Re: Problem about grub automatically changes my root device

[Niu Kun]

Boyd Stephen Smith Jr. 写道:

On Friday 02 October 2009 15:41:11 Niu Kun wrote:
  

I've just upgrade my sarge system to lenny. We now encountered a problem.
Our original kernel is 2.4.27 and our present kernel is 2.6.26.
Our main SATA disk is recognized as hdd at the time of installation.
But in lenny, it is recoginzed as sda.
I find that each time I install a kernel image package, it still
automatically changes the sda, which is adjusted as my system's
requirement by myself, to hdd, which is the original main device number.
I can fix this manually each time I install a new kernel package. But
how can I make it automatically?



Read the documentation in /boot/grub/menu.lst
  

That's of great help. I've figured it out.
Thanks.


--
To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org 
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org

problem with grub loader and screen resolution in gnome

[raman narasimhan]

during  debian installation from a new r4 cd that i downloaded from
debian.org, i got the following error

grub fatal error

during installation and then i installed the lilo loader instead of debian.

now i can access debian though i still dont know how to configure lilo to
access windows. this is the first time i got grub fatal error during debian
installation

A. How do i revert back to grub loader??
B. Or how can i configure lilo to get access to my Windows??

Another problem i'm having is that i'm not being able to change the screen
resolution.. all icons appear very large in size...  I tried changing the
resolution but i found only one resolution is supported 640*480..


C. how do i make my display normal??

Re: problem with grub loader and screen resolution in gnome

[Boyd Stephen Smith Jr.]

On Thursday 04 December 2008, raman narasimhan [EMAIL PROTECTED] wrote 
about 'problem with grub loader and screen resolution in gnome':
A. How do i revert back to grub loader??

From within the running system install grub to the mbr.  For more 
information, see the grub documentation.

B. Or how can i configure lilo to get access to my Windows??

I do not know.

C. how do i make my display normal??

We will probably need more information, but a start would 
be 'dpkg-reconfigure xserver-xorg' which should prompt you about what 
resolutions you'd like available.
-- 
Boyd Stephen Smith Jr. ,= ,-_-. =. 
[EMAIL PROTECTED]  ((_/)o o(\_))
ICQ: 514984 YM/AIM: DaTwinkDaddy   `-'(. .)`-' 
http://iguanasuicide.org/  \_/ 


signature.asc
Description: This is a digitally signed message part.

problem with grub

[Gilles Guiot]

Hello all,

I'm a newbie to Linux. I have a Dell server with two raid arrays (sda 
and sdb), both raid 1.  I created and LVM on sdb. After having copied 
the filesystem from sda onto sdb, I want to configure grub so that it 
boots on sdb.  when i type grub root (sdb,0) , i get the message : 
syntax error near unexpected token '('.

Could someone tell me what's my mistake and how to do it properly ?

thanks a lot in advance

--
Gilles Guiot
Responsable exploitation informatique
Tél. : 01 53 23 02 20
[EMAIL PROTECTED]


--
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]

Re: problem with grub

[hh . eu]

Am 2008-05-15 um 11:14 schrieb Gilles Guiot:

Hello all,

I'm a newbie to Linux. I have a Dell server with two raid arrays  
(sda and

sdb), both raid 1.  I created and LVM on sdb. After having copied the
filesystem from sda onto sdb, I want to configure grub so that it  
boots on

sdb.


Could you clarify what your setup is?

Do you have two disks, sda and sdba, which together form *one* RAID1  
array, or
do you have two partitions on each of the two disks and use two RAID1  
arrays,
or...? I don't quite understand why you only create LVM on sdb, not  
on sda,

and what you mean by copying the filesystem from sda to sdb.

You probably mean you have two partitions on disks sda and sdb, one  
small

partition and a larger partition. Then you put the two small partitions
together in RAID1 and use it for /boot. You also put the two larger  
partitions

together into another RAID1 array and use this for LVM.

And now you want to make sure you can boot your system either from  
sda or sdb,

i.e. you can start the system even if one disk has failed, correct?


when i type grub root (sdb,0) , i get the message : syntax error near
unexpected token '('.
Could someone tell me what's my mistake and how to do it properly ?


Do you type grub root (sdb,0) (without the ) all in one line? If  
so, you
have misunderstood the instructions! You first need to type grub,  
then press
enter, this will get you to the GRUB command, i.e. you will see a  
command
prompt of grub at the left. Only then you have to enter root (sdb, 
0) and
all the other commands (each followed by pressing enter and without  
the ).


-Moritz


--
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]

Re: problem with grub

[Gilles Guiot]

[EMAIL PROTECTED] a écrit :


Am 2008-05-15 um 11:14 schrieb Gilles Guiot:

Hello all,

I'm a newbie to Linux. I have a Dell server with two raid arrays (sda 
and

sdb), both raid 1.  I created and LVM on sdb. After having copied the
filesystem from sda onto sdb, I want to configure grub so that it 
boots on

sdb.


Could you clarify what your setup is?

Do you have two disks, sda and sdba, which together form *one* RAID1 
array, or
do you have two partitions on each of the two disks and use two RAID1 
arrays,

or...?


my server has two raid1 arrays, each with two disks, for a total of four 
hd. sda1 relates to the biggest partion on the first array (sda) and 
sdb1 relates to the only partition on the second raid array. point is 
this sdb1 is in lvm so to speak, ie there is an lv using all of the 
partition.



I don't quite understand why you only create LVM on sdb, not on sda,
and what you mean by copying the filesystem from sda to sdb.
This is an installed and working server. Because we needed more space 
for backuppc, the initial plan was to create an lvm on the second raid 
array, boot to it and see if it worked, if so extend it by incorporating 
the first raid array.


You probably mean you have two partitions on disks sda and sdb, one small
partition and a larger partition. Then you put the two small partitions
together in RAID1 and use it for /boot. You also put the two larger 
partitions

together into another RAID1 array and use this for LVM.

And now you want to make sure you can boot your system either from sda 
or sdb,

i.e. you can start the system even if one disk has failed, correct?


when i type grub root (sdb,0) , i get the message : syntax error near
unexpected token '('.
Could someone tell me what's my mistake and how to do it properly ?


Do you type grub root (sdb,0) (without the ) all in one line? If 
so, you
have misunderstood the instructions! You first need to type grub, 
then press

enter, this will get you to the GRUB command, i.e. you will see a command
prompt of grub at the left. Only then you have to enter root 
(sdb,0) and
all the other commands (each followed by pressing enter and without 
the ).
I typed the first grub to enter the grub command, omitted to specifiy it 
. Apologies.




-Moritz





--
Gilles Guiot
Responsable exploitation informatique
Tél. : 01 53 23 02 20
[EMAIL PROTECTED]


--
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]

Re: problem with grub

[hh . eu]

2008-05-15 17:02+0200 Gilles Guiot:

my server has two raid1 arrays, each with two disks, for a total of  
four hd. sda1 relates to the biggest partion on the first array  
(sda) and sdb1 relates to the only partition on the second raid  
array. point is this sdb1 is in lvm so to speak, ie there is an  
lv using all of the partition.



2008-05-15 16:43+0200 hh.eu:



I don't quite understand why you only create LVM on sdb, not on sda,
and what you mean by copying the filesystem from sda to sdb.
This is an installed and working server. Because we needed more  
space for backuppc, the initial plan was to create an lvm on the  
second raid array, boot to it and see if it worked, if so extend it  
by incorporating the first raid array.


I am using software RAID, which gives me device names such as md0 and  
md1 for

the RAID arrays rather than sda and sdb, so are you using some sort of
hardware RAID (either with a real hardware RAID controller or 'fake  
hardware

RAID' using the embedded controller on the motherboard)?

Anyway, if I understand you correctly, you have two partitions on the  
old RAID
set and just one partition (with LVM) on the new RAID set. The  
problem is
that (legacy) GRUB cannot deal with LVM directly. GRUB can understand  
RAID,
but not LVM. That's why people usually have a small RAID1 set on two  
disks
that contains just a partition on each disk for /boot, and another  
larger
RAID1 set on the same two disks that is entirely filled with LVM.  
That's what

I was talking about:

You probably mean you have two partitions on disks sda and sdb,  
one small
partition and a larger partition. Then you put the two small  
partitions
together in RAID1 and use it for /boot. You also put the two  
larger partitions

together into another RAID1 array and use this for LVM.


In other words: You need two partitions on the new RAID set, too. The  
size for
the /boot partitions can be quite small, I chose 512 MB which should  
be much
more than enough. (I didn't want to make it smaller because hard disk  
space is
cheap nowadays and making it bigger isn't so convenient because it's  
not on

LVM!).

when i type grub root (sdb,0) , i get the message : syntax error  
near

unexpected token '('.
Could someone tell me what's my mistake and how to do it properly ?


Do you type grub root (sdb,0) (without the ) all in one line?  
If so, you
have misunderstood the instructions! You first need to type  
grub, then press
enter, this will get you to the GRUB command, i.e. you will see a  
command
prompt of grub at the left. Only then you have to enter root  
(sdb,0) and
all the other commands (each followed by pressing enter and  
without the ).
I typed the first grub to enter the grub command, omitted to  
specifiy it . Apologies.


I was asking because the error message seems very much like an error  
message

from your normal shell, e.g. bash, which you shouldn't get from the GRUB
shell. (I am not in front of a Debian system so I can't test...)

You also need to specify the hard disk name/number in GRUB notation  
(which is
something like hd0 or hd1, even for SCSI disks), not in the normal  
notation,

i.e. root (sdb,0) doesn't work.

Unfortunately, GRUB uses some sort of guessing to map device names  
(e.g. sdb
-- hd0, sda -- hd1) which is not very reliable. See a recent post  
from me

for tips and details to find out about the mapping:

http://lists.debian.org/debian-user/2008/05/msg01006.html

Did you read the manual http://www.gnu.org/software/grub/manual/ 
grub.html?


-Moritz


--
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]

Re: problem with grub

[Wackojacko]

[EMAIL PROTECTED] wrote:


2008-05-15 17:02+0200 Gilles Guiot:

my server has two raid1 arrays, each with two disks, for a total of 
four hd. sda1 relates to the biggest partion on the first array (sda) 
and sdb1 relates to the only partition on the second raid array. point 
is this sdb1 is in lvm so to speak, ie there is an lv using all of 
the partition.



2008-05-15 16:43+0200 hh.eu:



I don't quite understand why you only create LVM on sdb, not on sda,
and what you mean by copying the filesystem from sda to sdb.
This is an installed and working server. Because we needed more space 
for backuppc, the initial plan was to create an lvm on the second raid 
array, boot to it and see if it worked, if so extend it by 
incorporating the first raid array.


I am using software RAID, which gives me device names such as md0 and 
md1 for

the RAID arrays rather than sda and sdb, so are you using some sort of
hardware RAID (either with a real hardware RAID controller or 'fake 
hardware

RAID' using the embedded controller on the motherboard)?

Anyway, if I understand you correctly, you have two partitions on the 
old RAID

set and just one partition (with LVM) on the new RAID set. The problem is
that (legacy) GRUB cannot deal with LVM directly. GRUB can understand RAID,
but not LVM. That's why people usually have a small RAID1 set on two disks
that contains just a partition on each disk for /boot, and another larger
RAID1 set on the same two disks that is entirely filled with LVM. That's 
what

I was talking about:

You probably mean you have two partitions on disks sda and sdb, one 
small

partition and a larger partition. Then you put the two small partitions
together in RAID1 and use it for /boot. You also put the two larger 
partitions

together into another RAID1 array and use this for LVM.


In other words: You need two partitions on the new RAID set, too. The 
size for
the /boot partitions can be quite small, I chose 512 MB which should be 
much
more than enough. (I didn't want to make it smaller because hard disk 
space is

cheap nowadays and making it bigger isn't so convenient because it's not on
LVM!).


when i type grub root (sdb,0) , i get the message : syntax error near
unexpected token '('.
Could someone tell me what's my mistake and how to do it properly ?


Do you type grub root (sdb,0) (without the ) all in one line? If 
so, you
have misunderstood the instructions! You first need to type grub, 
then press
enter, this will get you to the GRUB command, i.e. you will see a 
command
prompt of grub at the left. Only then you have to enter root 
(sdb,0) and
all the other commands (each followed by pressing enter and without 
the ).
I typed the first grub to enter the grub command, omitted to specifiy 
it . Apologies.


I was asking because the error message seems very much like an error 
message

from your normal shell, e.g. bash, which you shouldn't get from the GRUB
shell. (I am not in front of a Debian system so I can't test...)

You also need to specify the hard disk name/number in GRUB notation 
(which is
something like hd0 or hd1, even for SCSI disks), not in the normal 
notation,

i.e. root (sdb,0) doesn't work.

Unfortunately, GRUB uses some sort of guessing to map device names (e.g. 
sdb
-- hd0, sda -- hd1) which is not very reliable. See a recent post 
from me

for tips and details to find out about the mapping:

http://lists.debian.org/debian-user/2008/05/msg01006.html

Did you read the manual 
http://www.gnu.org/software/grub/manual/grub.html?


-Moritz




Grub does tab completion so if you type root (tab it should give you a 
list of the available drives as grub sees them.  The file 
/boot/grub/device.map shows how these translate to real hard drives.


HTH

Wackojacko


--
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]

Re: chroot problem with grub

[Philippe De Ryck]

On Thu, 2006-04-13 at 12:13 -0600, Justin Guerin wrote:
 Hi,
 
 I've had to move my install to a new physical disk.  I made an image of my 
 two partitions (/boot and /), and they restored properly.  Now, I only need 
 to run grub-install to install the boot loader.  
 
 When I boot from Knoppix, I can mount the / to /mnt/target, then mount /boot 
 to /mnt/target/boot, and /proc to /mnt/target/proc, but I can't get 
 grub-install to work properly.  When I chroot /mnt/target, and run grub, 
 grub can't see the drives (error 21).  However, when I back out of the 
 chroot, grub sees the drives just fine.
 
 Can anyone tell me how grub accesses the bios to find out information about 
 drives?  I'm not passing something through the chroot, but I have no idea 
 what.  The device nodes are available in the chroot, and so is proc.  I'm 
 running as root, and I know I have access to the device nodes.
 
 Any help is appreciated.
 
 Justin

Justin,

I don't think it is necessary to chroot at all. The knoppix disk has
grub on board, so you can use that command. The command also has a
command line switch to specify a device (/dev/hda for instance) and you
can also specify a root-dir. If you specify as root-dir the mount point
of your system (/mnt/target) grub will take the config file
from /mnt/target/boot/...) and everything should work just fine.

If you search the internet (or the manual perhaps) for this specific
info you'll find a lot more.

Good luck

Philippe De Ryck


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]

Re: chroot problem with grub

[Andrei Popescu]

On Thu, 13 Apr 2006 21:21:44 +0200
Philippe De Ryck [EMAIL PROTECTED] wrote:

 On Thu, 2006-04-13 at 12:13 -0600, Justin Guerin wrote:
  Hi,
  
  I've had to move my install to a new physical disk.  I made an image of my 
  two partitions (/boot and /), and they restored properly.  Now, I only need 
  to run grub-install to install the boot loader.  
  
  When I boot from Knoppix, I can mount the / to /mnt/target, then mount 
  /boot 
  to /mnt/target/boot, and /proc to /mnt/target/proc, but I can't get 
  grub-install to work properly.  When I chroot /mnt/target, and run grub, 
  grub can't see the drives (error 21).  However, when I back out of the 
  chroot, grub sees the drives just fine.
  
  Can anyone tell me how grub accesses the bios to find out information about 
  drives?  I'm not passing something through the chroot, but I have no idea 
  what.  The device nodes are available in the chroot, and so is proc.  I'm 
  running as root, and I know I have access to the device nodes.
  
  Any help is appreciated.
  
  Justin
 
 Justin,
 
 I don't think it is necessary to chroot at all. The knoppix disk has
 grub on board, so you can use that command. The command also has a
 command line switch to specify a device (/dev/hda for instance) and you
 can also specify a root-dir. If you specify as root-dir the mount point
 of your system (/mnt/target) grub will take the config file
 from /mnt/target/boot/...) and everything should work just fine.
 
 If you search the internet (or the manual perhaps) for this specific
 info you'll find a lot more.
 
 Good luck
 
 Philippe De Ryck

This will work, but will install the grub version from Knoppix. If you
still want to do via chroot here's the recipe. Commands indented for
visibility (# means the root prompt, you can use sudo instead) and
assuming hda2 is your '/' and hda1 is your '/boot':

1. mount your / /boot AND /dev:
#mount /dev/hda2 /mnt/target
#mount /dev/hda1 /mnt/target/boot
#mount -o remount /dev /mnt/target/dev

2. chroot:
#chroot /mnt/target
3. mount /proc:
#mount /proc
4. install grub:
#grub-install /dev/hda

This should do it. The first part is the one that can create problems.
If you mount / via Knoppix's fstab then you won't be able to
mount /dev due to the nodev option ;) (This is documented on Knoppix's
site)

HTH
Andrei
-- 
If you can't explain it simply, you don't understand it well enough.
(Albert Einstein)


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]

chroot problem with grub

[Justin Guerin]

Hi,

I've had to move my install to a new physical disk.  I made an image of my 
two partitions (/boot and /), and they restored properly.  Now, I only need 
to run grub-install to install the boot loader.  

When I boot from Knoppix, I can mount the / to /mnt/target, then mount /boot 
to /mnt/target/boot, and /proc to /mnt/target/proc, but I can't get 
grub-install to work properly.  When I chroot /mnt/target, and run grub, 
grub can't see the drives (error 21).  However, when I back out of the 
chroot, grub sees the drives just fine.

Can anyone tell me how grub accesses the bios to find out information about 
drives?  I'm not passing something through the chroot, but I have no idea 
what.  The device nodes are available in the chroot, and so is proc.  I'm 
running as root, and I know I have access to the device nodes.

Any help is appreciated.

Justin


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]

Re: chroot problem with grub

[Justin Guerin]

On Thursday 13 April 2006 13:21, Philippe De Ryck wrote:
 On Thu, 2006-04-13 at 12:13 -0600, Justin Guerin wrote:
  Hi,
 
  [snip problem]

 Justin,

 I don't think it is necessary to chroot at all. The knoppix disk has
 grub on board, so you can use that command. The command also has a
 command line switch to specify a device (/dev/hda for instance) and you
 can also specify a root-dir. If you specify as root-dir the mount point
 of your system (/mnt/target) grub will take the config file
 from /mnt/target/boot/...) and everything should work just fine.

 If you search the internet (or the manual perhaps) for this specific
 info you'll find a lot more.

 Good luck

 Philippe De Ryck

You're right, it wasn't necessary to chroot.  I simply mounted the drives 
and issued the command grub-install --root-directory=/mnt/target /dev/hda 
and it worked.  For good measure, before I rebooted, I chrooted and ran 
update-grub, but I'm not certain that was necessary.  Now, all my kernels 
are back and working.

Thanks Philippe!


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]

Re: Problem mit Grub

[Matthias Houdek]

Hallo Andreas Krummrich, hallo auch an alle anderen

Am Freitag, 3. Februar 2006 22:43 schrieb Andreas Krummrich:
 [...]
 Hat aber auch nicht geklappt. Ein update-grup mach die Änderung
 wieder rückgängig?!?

 Was mache ich hier falsch?

Schreib deine Zeilen außerhalb des automatisch generierten Bereichs 
(hinter '## ## END DEFAULT OPTIONS ##'), wenn es sich eh um zusätzliche 
Einträge handelt.

Oder schreib die Parameter nicht direkt in den jeweiligen Menüeintrag, 
sondern nutze die Zeilen im oberen Bereich der menu.lst (Jeweils die 
mit einem '#' auskommentierten Zeilen, ist ganz vernünftig erklärt in 
der Datei).

-- 
Gruß
MaxX

Bitte beachten: Diese Mailadresse nimmt nur Listenmails entgegen.
Für PM bitte den Empfänger gegen den Namen in der Sig tauschen.

Problem mit Grub

[Andreas Krummrich]

Hallo zusammen,

ich habe ein Problem mit Grub. Ich  habe das OpenBSD auf meiner Soekris
NET4801 in Rente geschickt und Debian Sarge installiert.
Nun habe ich ein Problem beim booten bzw. mit den Ausgaben über die
serielle Konsole. Also der Grub und später das login prompt werden auf
der Konsole ausgegeben. Nur die Bootmeldungen nicht :-(

Folgendes habe ich in die menu.lst eingertragen:

serial --unit=0 --speed=19200 --word=8 --parity=no --stop=1
terminal serial

Das funktioniert auch soweit. Na ja, bis auf die Bootmeldungen.

Dann habe ich versucht die Zeile

kernel  /boot/vmlinuz-2.6.8-2-686 root=/dev/hda1 ro

in

kernel  /boot/vmlinuz-2.6.8-2-686 root=/dev/hda1 ro
console=ttyS0,1920n81

zu ändern.

Hat aber auch nicht geklappt. Ein update-grup mach die Änderung wieder
rückgängig?!?

Was mache ich hier falsch?

Gruß,
Andreas


-- 
Haeufig gestellte Fragen und Antworten (FAQ): 
http://www.de.debian.org/debian-user-german-FAQ/

Zum AUSTRAGEN schicken Sie eine Mail an [EMAIL PROTECTED]
mit dem Subject unsubscribe. Probleme? Mail an [EMAIL PROTECTED] (engl)