Re: [linux-lvm] Re: your mail

Re: [linux-lvm] Re: your mail

[Heinz Mauelshagen]

> 
> On Wed, 17 Feb 1999, Michael Marxmeier wrote:
> 
> > Heinz wrote:
> > > /boot must be in a regular disk partition containing basic commands,
> > > device specials and LVM driver/commands.
> > > About 20-30MB should be sufficient to set this up.
> > 
> > IMO this could even be less if initrd is used.
> > We may need about 1-2 MB per kernel image + 1-2 MB for the initial
> > utilities (shell, vgchange, mount).
> 
> How about a special boot area inside a boot-configured
> root volume group? That way, lilo could be modified to run the
> second stage boot loader which could then be as heavyweight
> as anyone needs it.

LILO is able to deal with sectors scattered over a boot disk limited
by the 1024 cylinder BIOS rsstriction.
This implies no need for a special boot area _if_ we lay all
sectors LILO needs into the LV parts on that one disk.
Or did you mean that with voot area?

If LILO would be able to deal with sectors scattered over many disks,
why to have a special boot area anyway?

> 
> Isn't this sort of how hpux handles it?
> 


HP-UX's 2nd stage bootstrap loader is only able to deal with a contiguous
HFS formated filesystem.
That's the reason they mark the boot/root LV contiguous.
Without tricks you are stuck with the initial created size.
AFAIK this hasn't been changed including 11.X.


Heinz
--

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

Systemmanagement C/S                             Deutsche Telekom AG
                                                 Entwicklungszentrum Darmstadt
Heinz Mauelshagen                                Otto-Roehm-Strasse 71c
Senior Systems Engineer                          Postfach 10 05 41
                                                 64205 Darmstadt
mge@ez-darmstadt.telekom.de                      Germany
                                                 +49 6151 886-425
                                                          FAX-386
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-


--

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

Systemmanagement C/S                             Deutsche Telekom AG
                                                 Entwicklungszentrum Darmstadt
Heinz Mauelshagen                                Otto-Roehm-Strasse 71c
Senior Systems Engineer                          Postfach 10 05 41
                                                 64205 Darmstadt
mge@ez-darmstadt.telekom.de                      Germany
                                                 +49 6151 886-425
                                                          FAX-386
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

Re: [linux-lvm] Re: your mail

[Daniel Whicker]

>LILO is able to deal with sectors scattered over a boot disk limited
>by the 1024 cylinder BIOS rsstriction.
>This implies no need for a special boot area _if_ we lay all
>sectors LILO needs into the LV parts on that one disk.
>Or did you mean that with voot area?
>
>If LILO would be able to deal with sectors scattered over many disks,
>why to have a special boot area anyway?

Okay.  Let me make sure I understand this.  I think that what you're saying
is that at this point, LILO should be able handle an LV like any other
partition as far as booting is concerned?  If so, then the real problem
lies in getting the LVM turned on early enough that we can finish the boot
process, correct?  And if that were the case, we should be able to treat
the LVM like any other necessary kernel driver and install it with an
initrd.  Sorry to repeat things, but I want to make sure that I have a firm
understanding of what you're saying.

--Daniel
									-Daniel
		------------------------------------------
		Daniel Whicker  (heimdall@mail.org)

Re: [linux-lvm] Re: your mail

[Shawn Leas]

On Wed, 17 Feb 1999, Daniel Whicker wrote:

> Okay.  Let me make sure I understand this.  I think that what you're saying
> is that at this point, LILO should be able handle an LV like any other
> partition as far as booting is concerned?  If so, then the real problem
> lies in getting the LVM turned on early enough that we can finish the boot
> process, correct?  And if that were the case, we should be able to treat
> the LVM like any other necessary kernel driver and install it with an
> initrd.  Sorry to repeat things, but I want to make sure that I have a firm
> understanding of what you're saying.

I guess now that I think of it, I was kindof thinking of
an area on the root VG that lilo could find, which would
hold a kernel, and an initrd image so the extra partition
isn't needed.

-Shawn
<=========== America Held Hostage ===========>
   Day 2219 for the poor and the middle class. 
   Day 2238 for the rich and the dead.
   703 days remaining in the Raw Deal.
<============================================> 

Re: [linux-lvm] Re: your mail

[Heinz Mauelshagen]

> 
> On Thu, 18 Feb 1999, Heinz Mauelshagen wrote:
> 
> > This implies no need for a special boot area _if_ we lay all
> > sectors LILO needs into the LV parts on that one disk.
> > Or did you mean that with voot area?
> 

On Wed, 17 Feb 1999 22:34:49 -0600, Shawn Leas wrote:
> Mirrored configurations would work ok then I guess, right?

> This (mirroring) would be my main reason for root on an LV.

Yes, but LVM mirroring is future 8*(

If we will have mirroring support in the LVM some day it should not be the
big issue to activate a mirror at an early stage while initializing
the LVM driver (static bound into the kernel).


> By voot,
> what did you mean? (Sorry, it's probably in some M somewhere that I
> should be RTFing...)

Sorry, just an early morning hours typo coming much too late from work.

voot should have been boot (area).

RTFing beer on my bill ;*)

> 
> > If LILO would be able to deal with sectors scattered over many disks,
> > why to have a special boot area anyway?
> 
> Probably none. /me says KISS! ;P

As i said in another posting, LVM had to take care of LILO single disk
and 1024 cylinder limitations.

> 
> > HP-UX's 2nd stage bootstrap loader is only able to deal with a contiguous
> > HFS formated filesystem.
> > That's the reason they mark the boot/root LV contiguous.
> > Without tricks you are stuck with the initial created size.
> > AFAIK this hasn't been changed including 11.X.
> 
> Geez... Well, commercial eunices are driven my market demand and
> not quality or grace or programmer's pride. Luckily, there's you,
> Linus, Alan, Richard Gooch, SCT, HJ Lu, HPA, Andreas, and all the
> rest of the code wariors out there doing it better!
> 

Heinz
--

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

Systemmanagement C/S                             Deutsche Telekom AG
                                                 Entwicklungszentrum Darmstadt
Heinz Mauelshagen                                Otto-Roehm-Strasse 71c
Senior Systems Engineer                          Postfach 10 05 41
                                                 64205 Darmstadt
mge@ez-darmstadt.telekom.de                      Germany
                                                 +49 6151 886-425
                                                          FAX-386
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

Re: [linux-lvm] Re: your mail

[Heinz Mauelshagen]

> 
> Heinz:
> >LILO is able to deal with sectors scattered over a boot disk limited
> >by the 1024 cylinder BIOS rsstriction.
> >This implies no need for a special boot area _if_ we lay all
> >sectors LILO needs into the LV parts on that one disk.
> >Or did you mean that with voot area?
> >
> >If LILO would be able to deal with sectors scattered over many disks,
> >why to have a special boot area anyway?
> 

Daniel:
> Okay.  Let me make sure I understand this.  I think that what you're saying
> is that at this point, LILO should be able handle an LV like any other
> partition as far as booting is concerned?

No, only if the logical extents belonging to the boot logical volume
all are on _one_ bootable physical disk _and_ don't suffer from a 1024 cylinder
limit or anything similar.

LVM had to care about keeping up those restrictions.

> If so, then the real problem
> lies in getting the LVM turned on early enough that we can finish the boot
> process, correct?

Yes.
It would imply an extension to the LVM driver init stuff to
deal with such a kind of early activation of the boot volume group if
the driver is static bound into the kernel.


> And if that were the case, we should be able to treat
> the LVM like any other necessary kernel driver and install it with an
> initrd.

If we had a boot logical volume we could have a static LVM bound kernel
or an initrd solution.

> Sorry to repeat things, but I want to make sure that I have a firm
> understanding of what you're saying.

You are welcome!

It helps me to be more precise and others to join in 8*)


Regards,
Heinz
--

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

Systemmanagement C/S                             Deutsche Telekom AG
                                                 Entwicklungszentrum Darmstadt
Heinz Mauelshagen                                Otto-Roehm-Strasse 71c
Senior Systems Engineer                          Postfach 10 05 41
                                                 64205 Darmstadt
mge@ez-darmstadt.telekom.de                      Germany
                                                 +49 6151 886-425
                                                          FAX-386
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

Re: [linux-lvm] Re: your mail

[Shawn Leas]

On Thu, 18 Feb 1999, Heinz Mauelshagen wrote:

> This implies no need for a special boot area _if_ we lay all
> sectors LILO needs into the LV parts on that one disk.
> Or did you mean that with voot area?

Mirrored configurations would work ok then I guess, right? This
(mirroring) would be my main reason for root on an LV. By voot,
what did you mean? (Sorry, it's probably in some M somewhere that I
should be RTFing...)

> If LILO would be able to deal with sectors scattered over many disks,
> why to have a special boot area anyway?

Probably none. /me says KISS! ;P

> HP-UX's 2nd stage bootstrap loader is only able to deal with a contiguous
> HFS formated filesystem.
> That's the reason they mark the boot/root LV contiguous.
> Without tricks you are stuck with the initial created size.
> AFAIK this hasn't been changed including 11.X.

Geez... Well, commercial eunices are driven my market demand and
not quality or grace or programmer's pride. Luckily, there's you,
Linus, Alan, Richard Gooch, SCT, HJ Lu, HPA, Andreas, and all the
rest of the code wariors out there doing it better!

-Shawn
<=========== America Held Hostage ===========>
   Day 2219 for the poor and the middle class. 
   Day 2238 for the rich and the dead.
   703 days remaining in the Raw Deal.
<============================================> 

Re: [linux-lvm] Re: your mail

[Shawn Leas]

On Wed, 17 Feb 1999, Michael Marxmeier wrote:

> Heinz wrote:
> > /boot must be in a regular disk partition containing basic commands,
> > device specials and LVM driver/commands.
> > About 20-30MB should be sufficient to set this up.
> 
> IMO this could even be less if initrd is used.
> We may need about 1-2 MB per kernel image + 1-2 MB for the initial
> utilities (shell, vgchange, mount).

How about a special boot area inside a boot-configured
root volume group? That way, lilo could be modified to run the
second stage boot loader which could then be as heavyweight
as anyone needs it.

Isn't this sort of how hpux handles it?

-Shawn
<=========== America Held Hostage ===========>
   Day 2219 for the poor and the middle class. 
   Day 2238 for the rich and the dead.
   703 days remaining in the Raw Deal.
<============================================> 

Re: [linux-lvm] Re: your mail

[Thomas Haas]

Hi

I am using the md raid stuff
(http://www.kernel.org/pub/linux/daemons/raid/) right now. For the md
stuff both problems (booting md root device and shutdown) were solved in
one way or another. The last time I checked initrd had been used for
booting. I have not tracked how the shutdown problem has been resolved.
Some cleanup may be moved into the unmount call.


As I used AIX before I am waiting for LVM since a long time. I am glad it
is finally here. Now, only JFS must be implemented. Anyone on more details
on this (URLs)?

tnx for the great work
- tom

Michael Marxmeier wrote:

> Heinz wrote:
> > /boot must be in a regular disk partition containing basic commands,
> > device specials and LVM driver/commands.
> > About 20-30MB should be sufficient to set this up.
>
> IMO this could even be less if initrd is used.
> We may need about 1-2 MB per kernel image + 1-2 MB for the initial
> utilities (shell, vgchange, mount).
>
> Michael
>
> --
> Michael Marxmeier           Marxmeier Software GmbH
> E-Mail: mike@msede.com      Besenbruckstrasse 9
> Voice : +49 202 2431440     42285 Wuppertal, Germany
> Fax   : +49 202 2431420     http://www.msede.com/

--
* Thomas Haas                    <mailto:haas@softwired-inc.com>
* SoftWired AG                   <http://www.softwired-inc.com/>
* Technoparkstr. 1  ***  CH-8005 Zurich  ***  +41-1-4452370

Re: [linux-lvm] Re: your mail

[Michael Marxmeier]

Heinz wrote:
> /boot must be in a regular disk partition containing basic commands,
> device specials and LVM driver/commands.
> About 20-30MB should be sufficient to set this up.

IMO this could even be less if initrd is used.
We may need about 1-2 MB per kernel image + 1-2 MB for the initial
utilities (shell, vgchange, mount).


Michael

--
Michael Marxmeier           Marxmeier Software GmbH
E-Mail: mike@msede.com      Besenbruckstrasse 9
Voice : +49 202 2431440     42285 Wuppertal, Germany
Fax   : +49 202 2431420     http://www.msede.com/

[linux-lvm] Re: your mail

[Heinz Mauelshagen]

> 
> Hello,
> 
> Firstly  great product,

Thanks.

> but have a couple of questions
> 
> 1.  will we be able to use this as our boot disk [vg00]

Yes, but not supported with the standard distribution.
Hope to have supprt in the next release.

>  
> 2.  has anyone tested if lilo can handle /dev/vg00/lvol1 so it knows how to

No, it can't because it deals with sectors on a single boot disk.

You have to have a (small) boot partition (for eg. /dev/sda5) to boot
from and have to change to the real-root-dev in a LV after activating
volume groups.

>       boot the kernel,  I would assume changes would need to be made to allow
>       lilo to issue a vgchange on the [vg00] volume group.

LILO had to be enhanced to be able to load sectors from multiple disks
if the root logical volume is not continuous.
If it must stay continuous for LILO to be able to address kernel sectors
to load there would not be a real enhancement compared
to booting from a hard disk partition.
On Intel compatible PCs BIOS restrictions limiting the number of
BIOS addressable disks apply too (LILO uses the BIOS).


> Or could we leave
>       /boot  as a disk section and build into the kernel the ability to active
>       [vg00] ??

Yes, we could boot with LILO form a partition based boot FS and enhance
the LVM driver init stuff to activate LV root.
But this driver overhead is not necessary, if we boot from a small
partition based fs anyway.
OTOH it would make sense if LILO could directly load the kernel image from a LV.

> 
> 3.  are there plans for LVM-MIRRORING

Yes, but no fixed time frame 8*(

> 
> 4.  for allowing root on lvm, is their anny recommended file system sizes and
>         should contingous be set for /boot ?

/boot must be in a regular disk partition containing basic commands,
device specials and LVM driver/commands.
About 20-30MB should be sufficient to set this up.

> 
> Sorry if some of these questions have been duplicated, at time of writing
> couldn't see anything in the archive listing.
> 

Maybe you like to subscribe to our new linux-lvm maling list at
linux-lvm@msede.com by sending a "subscribe linux-lvm" in the body of the mail
to majordomo@msede.com.


Regards,
Heinz

--

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

Systemmanagement C/S                             Deutsche Telekom AG
                                                 Entwicklungszentrum Darmstadt
Heinz Mauelshagen                                Otto-Roehm-Strasse 71c
Senior Systems Engineer                          Postfach 10 05 41
                                                 64205 Darmstadt
mge@ez-darmstadt.telekom.de                      Germany
                                                 +49 6151 886-425
                                                          FAX-386
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-