[linux-lvm] Newbie question
Heinz J. Mauelshagen
Mauelshagen at sistina.com
Wed Jul 18 15:25:07 UTC 2001
On Tue, Jul 17, 2001 at 09:58:11PM +0200, Luca Berra wrote:
> sorry if i am getting way offtopic here but reply is a must
>
> On Tue, Jul 17, 2001 at 10:31:59AM +0200, josv at osp.nl wrote:
> > Hi all,
> >
> > Just for the record, HP-UX's vg layout is not in /stand. Information about
> > essential logical volumes (boot (/stand), root (/), primary swap and dump)
> > is stored in a LIF file (LABEL) in the boot area's of the disk (use the lifls
> > command to display the structure of the boot area). The lvlnboot command is used
> i was referring to:
> (from the lvlnboot manpage)
> FILES
> /stand/rootconf Contains the location of the root volume.
> Used during maintenance-mode boots (see
> hpux(1M)) to locate the root volume for
> volume groups with separate boot and root
> volumes.
>
> ok, i was not exactly correct but if you follow me you will maybe
> pardon me.
:)
> Steven stated that hpux -lm worked because HP9000 were able to
> understand LVM in firmware.
> i stated that hpux -lm works because lvm info is stored in a file in
> stand.
Luca is right: PDC (Processor Dependant Code; that's their kind of BIOS) doesn't
know anything about LVM. It understands LIF (Logical Interchange Format; kind
of a very simple filesystem)
>
> The real truth:
> /stand/rootconf does not contain LVM info
> it just contains info about the location on disk of the root lv
> meening starting offset and lenght (it is used only on split
> boot/root configurations when booting in LVM mainteinance mode)
>
> on my workstation
> # od -x /stand/rootconf
> 0000000 dead beef 000d 5b60 0004 0000
> | magic | offset | size
>
> /stand/rootconf is created/updated using lvlnboot -c
>
>
> lvlnboot updates both the LABEL file and the BRDA on the PV
> the kernel during normal boot reads the BRDA, not the LABEL file
> the LABEL file is used by the hpux utility, it is also used
> by offline diagnostics.
Sorry for sounding picky:
rootconf contains that information *but* the info is used to
switch to the real *root* filsystem which is necessary in HP-UX in order
to have root journaled (supported since HP-UX 10.20 IIRC).
The offset and length above is for sure the one of the root LV, because the
root LV needs to be contiguous with its LEs in sequential ascending order.
*But* as stated below, the LVM driver is deactivated in LVM maintenance
mode. Therefore the kernel mainly reads the offset info from /stand/rootconf
(BTW: stand is mounted as root at that point in time) in order to switch
to the 'real' root filesystem.
BTW: as briefly said above, seperate stand/root filesystem are
necessary in HP-UX in order to have root journaled.
If you need to recreate such a stand/root configuration after a crash
and you don't do that with *exactly the same* VG parameters for the root
VG, this leads to a different root LV offset. An unconditional restore to
/stand will cause an unbootable system, because the restored offset in
/stand/rootconf is incorrect then :-(
In this case only a nice echo trick to recreate /stand/rootconf
(knowing HP VGDA size params) can bring your system back to life again ;-)
>
> > to query and change the LABEL file. The mapping between VG and PV's is stored
> > in the /etc/lvmtab file. This file is used by the vgchange command to activate
> > volume groups. At boot time, /etc/lvmrc is run from the inittab, and it
> > executes a 'vgchange -a y' for all volume groups.
> this is not strictly correct (lvmrc is a script anc can/must be customized
> not to activate all VGs in a clustered environment)
>
> > The "-lm" boot option in HP-UX disables the lv driver (major 64), so only the
> > /stand, /, pri swap and dump "volumes" are available. Not as logical volumes,
> > but directly from the disk because these volumes must be contiguous. This
> this is just plain wrong
> only root is available to the system in lvm maint mode (via /stand/rootconf)
> /stand and pri swap are not, let alone dump.
> /stand must be contiguous at the beginning of the boot disk so the
> hpux utility can read a kernel off it
> / must also be contiguous on the boot disk so the kernel can map it
> swap and dump can be on any disk (the latter only if IODC is able to
> map it) and must be both contiguous. (yes primary swap does not need
> to be on lvol2 and on the first disk, altough this is the default)
>
> > option is necessary in HP-UX since it basically only supports LVM disk setups
> > (whole disk layout is also supported, but almost never used...). If Linux
> AFAIRC whole disk works only on disks < 1GB or so
>
> > would ever get to the point that LVM (or its successor) is *the* choice for
> > partitioning all disks (even the boot disk), we would need such an option as
> > well because otherwise there would be no path to recovery in case of LVM
> > data corruption.
>
> honestly i believe having a resizable root volume is better than being able to
> mount root without using lvm. since i can stick enough tools to recover lvm in
> an initrd or on a emergency cd-rom.
Agreed.
>
> ....
> > ++Jos
> > "Who could not resist commenting on this, sorry..."
>
> me neither, sorry again....
:)
>
> L.
> --
> Luca Berra -- bluca at comedia.it
> Communication Media & Services S.r.l.
> /"\
> \ / ASCII RIBBON CAMPAIGN
> X AGAINST HTML MAIL
> / \
>
> _______________________________________________
> linux-lvm mailing list
> linux-lvm at 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 at Sistina.com +49 2626 141200
FAX 924446
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
More information about the linux-lvm
mailing list