[linux-lvm] Root on LV

[linux-lvm] Root on LV

[dobos_s]

Hi!

I cannot understand something.
My kernel has no modules.
Why should I create an ext2 partition for holding an initrd which is not
doing any essential things except vgscan and vgchange -ay ?
Why cant these things built into the kernel?

Maybe I didnt understand something...


Please CC to me, because I am not member on list...

Thanks


Dobos Sandor
IBCnet Hungary Ltd.

Re: [linux-lvm] Root on LV

[josv]

vgscan and vgchange -a y *ARE* quite essential.
It is currently *not* possible to perform these two essential
commands in the kernel.

AFAIK that is...

++Jos

And thus it came to pass that dobos_s@IBCnet.hu wrote:
(on Thu, Sep 06, 2001 at 04:03:49PM +0200 to be exact)

> Hi!
> 
> I cannot understand something.
> My kernel has no modules.
> Why should I create an ext2 partition for holding an initrd which is not
> doing any essential things except vgscan and vgchange -ay ?
> Why cant these things built into the kernel?
> 
> Maybe I didnt understand something...
> 
> 
> Please CC to me, because I am not member on list...
> 
> Thanks
> 
> 
> Dobos Sandor
> IBCnet Hungary Ltd.
> 
> 
> 
> 
> 
> _______________________________________________
> linux-lvm mailing list
> linux-lvm@sistina.com
> http://lists.sistina.com/mailman/listinfo/linux-lvm
> read the LVM HOW-TO at http://www.sistina.com/lvm/Pages/howto.html

-- 
Klingon coding rule #12: "Specifications are for the weak and timid!"

Re: [linux-lvm] Root on LV

[svetljo]

it probably will hold the kernel also
and as far i know there is no bootloader which can access logical volume
am i wrong

dobos_s@IBCnet.hu wrote:

>Hi!
>
>I cannot understand something.
>My kernel has no modules.
>Why should I create an ext2 partition for holding an initrd which is not
>doing any essential things except vgscan and vgchange -ay ?
>Why cant these things built into the kernel?
>
>Maybe I didnt understand something...
>
>
>Please CC to me, because I am not member on list...
>
>Thanks
>
>
>Dobos Sandor
>IBCnet Hungary Ltd.
>
>
>
>
>
>_______________________________________________
>linux-lvm mailing list
>linux-lvm@sistina.com
>http://lists.sistina.com/mailman/listinfo/linux-lvm
>read the LVM HOW-TO at http://www.sistina.com/lvm/Pages/howto.html
>

Re: [linux-lvm] Root on LV

[Andreas Dilger]

On Sep 06, 2001  16:42 +0200, svetljo wrote:
> it probably will hold the kernel also
> and as far i know there is no bootloader which can access logical volume
> am i wrong

Yes, you are wrong.  LILO+patch can get a kernel from a logical volume.
Now that Linus has accepted my patch to fix LV_USER_BMAP from causing
an oops I will submit the patch to the LILO maintainer.  It can also be
found in the archives for this mailing list (have posted several times).

> dobos_s@IBCnet.hu wrote:
> >My kernel has no modules.
> >Why should I create an ext2 partition for holding an initrd which is not
> >doing any essential things except vgscan and vgchange -ay ?

You don't need vgscan at boot time unless the disks have moved around.
You DO need "vgchange -a y" in order to load the LVM data into the kernel.
If you are running EVMS you don't need that.  I think this is also on the
TODO list for the next LVM version.

Cheers, Andreas
-- 
Andreas Dilger  \ "If a man ate a pound of pasta and a pound of antipasto,
                 \  would they cancel out, leaving him still hungry?"
http://www-mddsp.enel.ucalgary.ca/People/adilger/               -- Dogbert

Re: [linux-lvm] Root on LV

[svetljo]

Andreas Dilger wrote:

>On Sep 06, 2001  16:42 +0200, svetljo wrote:
>
>>it probably will hold the kernel also
>>and as far i know there is no bootloader which can access logical volume
>>am i wrong
>>
>
>Yes, you are wrong.  LILO+patch can get a kernel from a logical volume.
>Now that Linus has accepted my patch to fix LV_USER_BMAP from causing
>an oops I will submit the patch to the LILO maintainer.  It can also be
>found in the archives for this mailing list (have posted several times).
>
does this mean that i can have my kernel on LV of any type ( can i have 
it on striped one)
i think i have read some time ago that lilo can boot from linear RAID or 
RAID1 but not from RAID-0

Re: [linux-lvm] Root on LV

[lembark]

-- dobos_s@IBCnet.hu on 09/06/01 16:03:49 +0200

> Hi!
> 
> I cannot understand something.
> My kernel has no modules.
> Why should I create an ext2 partition for holding an initrd which is not
> doing any essential things except vgscan and vgchange -ay ?
> Why cant these things built into the kernel?

Don't put an initrd on it.  Create the first few partitions on 
your drive at 128MB (/), 64 or 128MB (primary swap), 320MB (/var).
Leave the rest on LVM.  With these three you can normally start
up and recover a damaged LVM system; if the root and primary swap
at least aren't in non-LVM storage you'll be unable to boot if 
LVM fries for any reason; /var is usually a pain to loose and 
doesn't grow all that much so I leave it on a partition also.

sl

Re: [linux-lvm] Root on LV

[Luca Berra]

On Thu, Sep 06, 2001 at 06:35:39PM +0200, svetljo wrote:
> Andreas Dilger wrote:
> >Yes, you are wrong.  LILO+patch can get a kernel from a logical volume.
> does this mean that i can have my kernel on LV of any type ( can i have 
> it on striped one)
> i think i have read some time ago that lilo can boot from linear RAID or 
> RAID1 but not from RAID-0
NO!!!!
lilo uses your pc bios to boot, it cannot do anything fancy,
it just reads from a disk into memory then jumps there.

in order to boot kernel lilo needs to store a mapfile on
the /boot fs, the disk offset of this file is stored in
lilo 1-st stage loader a 512 byte program, and you cannot read
a striped lv from a 512 byte proggie yet.

L.

-- 
Luca Berra -- bluca@comedia.it
        Communication Media & Services S.r.l.
 /"\
 \ /     ASCII RIBBON CAMPAIGN
  X        AGAINST HTML MAIL
 / \

Re: [linux-lvm] Root on LV

[Heinz J . Mauelshagen]

On Thu, Sep 06, 2001 at 09:37:29AM -0600, Andreas Dilger wrote:
> On Sep 06, 2001  16:42 +0200, svetljo wrote:
> > it probably will hold the kernel also
> > and as far i know there is no bootloader which can access logical volume
> > am i wrong
> 
> Yes, you are wrong.  LILO+patch can get a kernel from a logical volume.
> Now that Linus has accepted my patch to fix LV_USER_BMAP from causing
> an oops I will submit the patch to the LILO maintainer.  It can also be
> found in the archives for this mailing list (have posted several times).

Congratulations that the patch is in :-)

> 
> > dobos_s@IBCnet.hu wrote:
> > >My kernel has no modules.
> > >Why should I create an ext2 partition for holding an initrd which is not
> > >doing any essential things except vgscan and vgchange -ay ?
> 
> You don't need vgscan at boot time unless the disks have moved around.
> You DO need "vgchange -a y" in order to load the LVM data into the kernel.
> If you are running EVMS you don't need that.

But you need EVMS tools which are not available fo far :-(

> I think this is also on the
> TODO list for the next LVM version.

Yes, for one of the next versions ;-)

> 
> Cheers, Andreas
> -- 
> Andreas Dilger  \ "If a man ate a pound of pasta and a pound of antipasto,
>                  \  would they cancel out, leaving him still hungry?"
> http://www-mddsp.enel.ucalgary.ca/People/adilger/               -- Dogbert
> 
> 
> _______________________________________________
> linux-lvm mailing list
> linux-lvm@sistina.com
> http://lists.sistina.com/mailman/listinfo/linux-lvm
> read the LVM HOW-TO at http://www.sistina.com/lvm/Pages/howto.html

-- 

Regards,
Heinz    -- The LVM Guy --

*** Software bugs are stupid.
    Nevertheless it needs not so stupid people to solve them ***

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

Heinz Mauelshagen                                 Sistina Software Inc.
Senior Consultant/Developer                       Am Sonnenhang 11
                                                  56242 Marienrachdorf
                                                  Germany
Mauelshagen@Sistina.com                           +49 2626 141200
                                                       FAX 924446
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

Re: [linux-lvm] Root on LV

[Heinz J . Mauelshagen]

On Thu, Sep 06, 2001 at 04:10:11PM +0200, josv@osp.nl wrote:
> vgscan and vgchange -a y *ARE* quite essential.
> It is currently *not* possible to perform these two essential
> commands in the kernel.
> 
> AFAIK that is...

That's true. We will have in kernel LVM metadata read support in a
post 1.0 version of LVM which will avoid to run userspace tools to activate
choosen VGs at all.

Regards,
Heinz    -- The LVM Guy --

> 
> ++Jos
> 
> And thus it came to pass that dobos_s@IBCnet.hu wrote:
> (on Thu, Sep 06, 2001 at 04:03:49PM +0200 to be exact)
> 
> > Hi!
> > 
> > I cannot understand something.
> > My kernel has no modules.
> > Why should I create an ext2 partition for holding an initrd which is not
> > doing any essential things except vgscan and vgchange -ay ?
> > Why cant these things built into the kernel?
> > 
> > Maybe I didnt understand something...
> > 
> > 
> > Please CC to me, because I am not member on list...
> > 
> > Thanks
> > 
> > 
> > Dobos Sandor
> > IBCnet Hungary Ltd.
> > 
> > 
> > 
> > 
> > 
> > _______________________________________________
> > linux-lvm mailing list
> > linux-lvm@sistina.com
> > http://lists.sistina.com/mailman/listinfo/linux-lvm
> > read the LVM HOW-TO at http://www.sistina.com/lvm/Pages/howto.html
> 
> -- 
> Klingon coding rule #12: "Specifications are for the weak and timid!"
> 
> _______________________________________________
> linux-lvm mailing list
> linux-lvm@sistina.com
> http://lists.sistina.com/mailman/listinfo/linux-lvm
> read the LVM HOW-TO at http://www.sistina.com/lvm/Pages/howto.html

*** Software bugs are stupid.
    Nevertheless it needs not so stupid people to solve them ***

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

Heinz Mauelshagen                                 Sistina Software Inc.
Senior Consultant/Developer                       Am Sonnenhang 11
                                                  56242 Marienrachdorf
                                                  Germany
Mauelshagen@Sistina.com                           +49 2626 141200
                                                       FAX 924446
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

Re: [linux-lvm] Root on LV

[Steve Pratt]

>> > dobos_s@IBCnet.hu wrote:
>> > >My kernel has no modules.
>> > >Why should I create an ext2 partition for holding an initrd which is
not
>> > >doing any essential things except vgscan and vgchange -ay ?
>>
>> You don't need vgscan at boot time unless the disks have moved around.
>> You DO need "vgchange -a y" in order to load the LVM data into the
kernel.

> If you are running EVMS you don't need that.

>But you need EVMS tools which are not available fo far :-(

Actually that is not entirely true.  All of the volume group and physical
volumes tools are done.  We have started on all of the logical volume
commands.  We have stand alone utilities which minic the LVM ones, as well
as a GUI which integrates LVM function with other volume management
functions such as partitioning.


EVMS Development - http://www.sf.net/projects/evms
Linux Technology Center - IBM Corporation
(512) 838-9763  EMAIL: SLPratt@US.IBM.COM

Re: [linux-lvm] Root on LV

[lembark]

-- Steve Pratt <slpratt@us.ibm.com> on 09/17/01 09:26:12 -0500

>>> > dobos_s@IBCnet.hu wrote:
>>> > > My kernel has no modules.
>>> > > Why should I create an ext2 partition for holding an initrd which is
> not
>>> > > doing any essential things except vgscan and vgchange -ay ?
>>> 
>>> You don't need vgscan at boot time unless the disks have moved around.
>>> You DO need "vgchange -a y" in order to load the LVM data into the
> kernel.
> 
>> If you are running EVMS you don't need that.
> 
>> But you need EVMS tools which are not available fo far :-(
> 
> Actually that is not entirely true.  All of the volume group and physical
> volumes tools are done.  We have started on all of the logical volume
> commands.  We have stand alone utilities which minic the LVM ones, as well
> as a GUI which integrates LVM function with other volume management
> functions such as partitioning.

Q:  Why bother to put 64MB on an LVM?

The root volume doesn't grow much (make it 128MB and it'll never need to
grow).
If you ever have a problem with LVM you'll end up with an unbootable
system...

Re: [linux-lvm] Root on LV

[Steve Pratt]

>.Q:  Why bother to put 64MB on an LVM?

>The root volume doesn't grow much (make it 128MB and it'll never need to
>grow).

Sure, you could always do this.

>If you ever have a problem with LVM you'll end up with an unbootable
>system...

Right, which is why volume managers need to be really stable and robust.
Bottom line is that if this is the confidence level you have in your volume
manager (whether it is LVM or EVMS), I would question putting any data on
it.

Steve

EVMS Development - http://www.sf.net/projects/evms
Linux Technology Center - IBM Corporation
(512) 838-9763  EMAIL: SLPratt@US.IBM.COM

Re: [linux-lvm] Root on LV

[lembark]

-- Steve Pratt <slpratt@us.ibm.com> on 09/19/01 13:54:23 -0500

>> .Q:  Why bother to put 64MB on an LVM?
> 
>> The root volume doesn't grow much (make it 128MB and it'll never need to
>> grow).
> 
> Sure, you could always do this.
> 
>> If you ever have a problem with LVM you'll end up with an unbootable
>> system...
> 
> Right, which is why volume managers need to be really stable and robust.
> Bottom line is that if this is the confidence level you have in your
> volume manager (whether it is LVM or EVMS), I would question putting any
> data on it.

That's like saying you don't need ABS brakes because tire makers
should make ones with really, really good traction.  If you look
at the commercial LVM systems (e.g., HP's) they get around LVM
failures by having a non-trivial boot system that allows side-
stepping LVM at boot time via the command line.  This is done on
HP by requring /dev/vg00/lvol[123] to be contiguous and starting
at offset 0 on the VG.  You can boot vmunix -lm if LVM gets toasted
and at least get the boot, swap and root volumes on line to fix
the problem.  Lacking that much intelligence in the hardware level 
of Intel boot rom's you are one helluvalot sa[fn]er to put the
root and primary swap on /dev/[hs]da[12] and go from there.  

sl

Re: [linux-lvm] Root on LV

[Adrian Phillips]

>>>>> "Steve" == Steve Pratt <slpratt@us.ibm.com> writes:

    >> .Q: Why bother to put 64MB on an LVM?  The root volume doesn't
    >> grow much (make it 128MB and it'll never need to grow).

    Steve> Sure, you could always do this.

    >> If you ever have a problem with LVM you'll end up with an
    >> unbootable system...

    Steve> Right, which is why volume managers need to be really
    Steve> stable and robust.  Bottom line is that if this is the
    Steve> confidence level you have in your volume manager (whether
    Steve> it is LVM or EVMS), I would question putting any data on
    Steve> it.

Well, I disagree with what you seem to be implying (that if you can't
trust to have the whole disk with LVM then it isn't robust). As
somebody whos been running RS6000 with AIX (a highly robust OS IMO)
for several years I can say that you can never guarantee that a new
(ie. bugfix version) doesn't screw something up. At least with Linux
one has a nice fast, neat option of booting from a non LVM root
partition to fix any problems. Try waiting 10 minutes for a tape to
boot the operating system enough times to realise that having this
(root partition) option is very useful. And yes, I know, one can in
theory have a diskette or even CD with a version of the OS and all the
tools on (although using floppy these days is becoming more difficult
due to the size of everything), but a non LVM root of 128MB on a many
GB disks seems little to pay for the possiblities it gives (for
example, remote fixing of problems without having to burn a new CD for
each new machine).

On the other hand AIX was very easy to recover when an upgrade did
screw up, just roll back the upgrade. This is one thing I miss with
Linux (Debian at least) so far.

Sincerely,

Adrian Phillips

-- 
Your mouse has moved.
Windows NT must be restarted for the change to take effect.
Reboot now?  [OK]