[linux-lvm] RAID extension and lvm

[linux-lvm] RAID extension and lvm

[Ralf Eisinger]

Hi all,

during the next days we will get a new RAID-System.
We have planned to use it as one bis raid5-"disk". Its possible, to
increase the disk space by plugging in (a) new disk(s).
We want to use lvm too, because its possibility of making a snapshot.
I have two questions to that:
- when I "start" a snaptshot, where is lvm using the disk space, which is
  needed to track down the changes user will made during that time?
- Can lvm handle the changing of a disk size? (In our case, lvm will not
  recognize, that a disk is added, because the RAID-System offers the
  complete disk space as one big disk, so it seems (from the sight of
  lvm), that the disk size increases)

Is there anywhere a searchable version of the mailing list archive?

best regards

Ralf Eisinger
-------------------
IHS - University of Stuttgart, Germany
Pfaffenwaldring 10
70150 Stuttgart
Phone: (0049)-711-685-3201

Re: [linux-lvm] RAID extension and lvm

[Goetz Bock]

While I was instructed to answer RTFM more often, here you go ...

On Fri, Jul 12 '02 at 16:16, Ralf Eisinger wrote:
> - when I "start" a snaptshot, where is lvm using the disk space, which is
>   needed to track down the changes user will made during that time?
it allocates free PEs from the underlying VG

> - Can lvm handle the changing of a disk size? (In our case, lvm will not
>   recognize, that a disk is added, because the RAID-System offers the
>   complete disk space as one big disk, so it seems (from the sight of
>   lvm), that the disk size increases)
you can extend a PV (man pvresize)

-- 
Goetz Bock       (c) 2002 as     blacknet.de - Munich - Germany   /"\
IT Consultant    GNU FDL 1.1    secure mobile Linux everNETting   \ /
                                                                   X
 ASCII Ribbon Campaign against HTML email & microsoft attachments / \

Re: [linux-lvm] RAID extension and lvm

[Ralf Eisinger]

On Fri, 12 Jul 2002, Goetz Bock wrote:

>
> While I was instructed to answer RTFM more often, here you go ...
>
> On Fri, Jul 12 '02 at 16:16, Ralf Eisinger wrote:
> > - when I "start" a snaptshot, where is lvm using the disk space, which is
> >   needed to track down the changes user will made during that time?
> it allocates free PEs from the underlying VG
>
> > - Can lvm handle the changing of a disk size? (In our case, lvm will not
> >   recognize, that a disk is added, because the RAID-System offers the
> >   complete disk space as one big disk, so it seems (from the sight of
> >   lvm), that the disk size increases)
> you can extend a PV (man pvresize)
Sorry, but there is no pvresize command. I have searched through the HOWTO
too, but I find nothing.

best regard
Ralf Eisinger

Re: [linux-lvm] RAID extension and lvm

[Goetz Bock]

On Sun, Jul 14 '02 at 16:39, Ralf Eisinger wrote:
> Sorry, but there is no pvresize command. I have searched through the HOWTO
> too, but I find nothing.
Yes, looks like you're right, sorry.
I just remembered a post about this on this list and the reply was to
use pvresize. Maybe it's only part of LVM 1.1 or LVM2.
-- 
Goetz Bock       (c) 2002 as     blacknet.de - Munich - Germany   /"\
IT Consultant    GNU FDL 1.1    secure mobile Linux everNETting   \ /
                                                                   X
 ASCII Ribbon Campaign against HTML email & microsoft attachments / \

Re: [linux-lvm] RAID extension and lvm

[lembark]

> Sorry, but there is no pvresize command.

You wouldn't want it anyway. 

There isn't any way to "resize" an fault-tolerant RAID
system (1/3/5) and you can get more flexable, reliable
results with LVM than RAID0 anyway.

The way to add space into a VG is by incorporating new
PV's with vgextend. If you want to grow a fault-tolerant
system then build new RAID5 groups and add them as new
PV's.


--
Steven Lembark                              2930 W. Palmer
Workhorse Computing                      Chicago, IL 60647
                                           +1 800 762 1582

Re: [linux-lvm] RAID extension and lvm

[Ralf Eisinger]

On Sun, 14 Jul 2002 lembark@wrkhors.com wrote:

>
> > Sorry, but there is no pvresize command.
>
> You wouldn't want it anyway.
>
> There isn't any way to "resize" an fault-tolerant RAID
> system (1/3/5) and you can get more flexable,
This is not correct. You can extend most of the common hardware raid
systems with raid5. Most of it during operation without shutting down.

> reliable
> results with LVM than RAID0 anyway.
>
> The way to add space into a VG is by incorporating new
> PV's with vgextend. If you want to grow a fault-tolerant
> system then build new RAID5 groups and add them as new
> PV's.
But if I want to extend my system, I need at least three disks. If I
extend an existing raid5, I need only one disk.

best regards

Ralf Eisinger
-------------------
IHS - University of Stuttgart, Germany
Pfaffenwaldring 10
70150 Stuttgart
Phone: (0049)-711-685-3201

Re: [linux-lvm] RAID extension and lvm

[Tim]

I was about to ask why on earth a good hardware array couldn't
re-allocate the parity slices on-the-fly and act as if another disk (in
addition to the new physical volume) was present during the
reconstruction...

It is obvious to me that a *filesystem* must be resized, and the
software md drivers can hardly be expected to figure out all of this on
the fly, but a *hardware* device ought to be able to treat resizing as
just another "degraded" failure mode and rebuild.

Is that in fact what happens?  I know we used to run the Sun Storedge
arrays primarily as JBODs since we simply crammed them full to bursting,
but somewhere out there, perhaps people are buying RAIDs that aren't
already close to capacity ;-).

I still like the slice-and-dice flexibility of LVM but it is certainly a
drag to have to install a *minimum* of 3 physical devices to get an
additional RAID5 physical volume in an LVM volume group!


Quoth Ralf Eisinger:
> On Sun, 14 Jul 2002 lembark@wrkhors.com wrote:
> 
> >
> > > Sorry, but there is no pvresize command.
> >
> > You wouldn't want it anyway.
> >
> > There isn't any way to "resize" an fault-tolerant RAID
> > system (1/3/5) and you can get more flexable,
> This is not correct. You can extend most of the common hardware raid
> systems with raid5. Most of it during operation without shutting down.
> 
> > reliable
> > results with LVM than RAID0 anyway.
> >
> > The way to add space into a VG is by incorporating new
> > PV's with vgextend. If you want to grow a fault-tolerant
> > system then build new RAID5 groups and add them as new
> > PV's.
> But if I want to extend my system, I need at least three disks. If I
> extend an existing raid5, I need only one disk.
> 
> best regards
> 
> Ralf Eisinger
> -------------------
> IHS - University of Stuttgart, Germany
> Pfaffenwaldring 10
> 70150 Stuttgart
> Phone: (0049)-711-685-3201
> 
> 
> 
> _______________________________________________
> 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

-- 
  Tim Triche, Jr.
  Senior Engineer
  ConnectLive Communications, Inc.
  (202) 513-1000 || (202) 253-5666

Re: [linux-lvm] RAID extension and lvm

[Ralf Eisinger]

Sorry, I think there is/was something confusing. We want to have the
following scenario:

- we had an hardware raid5 with three disks (so we get the size from two
  of them)
- from the view of the kernel there is one big SCSI disk
- on that disk I install lvm (for snapshots) and ext3
- ...
- we are working (heavily ;-) on that disks
- ...
- Because we are running out of disk space, I plug in a new disk
  and add them to the existing raid5. So it will have now four disks.
  This adding is done by the hardware.
- Now the disk size of the "one big disk" is changed to the new
  "netto" size of the raid5.

Is it possible to handle that with lvm and when the answer is "yes", in
which way? One point was a command named "pvresize" but I didn't find it
in my distribution (SuSE 8.0) and I find nothing about that in the HOWTO.


On Sun, 14 Jul 2002, Tim wrote:

>
> I was about to ask why on earth a good hardware array couldn't
> re-allocate the parity slices on-the-fly and act as if another disk (in
> addition to the new physical volume) was present during the
> reconstruction...
>
> It is obvious to me that a *filesystem* must be resized, and the
> software md drivers can hardly be expected to figure out all of this on
> the fly, but a *hardware* device ought to be able to treat resizing as
> just another "degraded" failure mode and rebuild.
>
> Is that in fact what happens?  I know we used to run the Sun Storedge
> arrays primarily as JBODs since we simply crammed them full to bursting,
> but somewhere out there, perhaps people are buying RAIDs that aren't
> already close to capacity ;-).
>
> I still like the slice-and-dice flexibility of LVM but it is certainly a
> drag to have to install a *minimum* of 3 physical devices to get an
> additional RAID5 physical volume in an LVM volume group!
>
>
> Quoth Ralf Eisinger:
> > On Sun, 14 Jul 2002 lembark@wrkhors.com wrote:
> >
> > >
> > > > Sorry, but there is no pvresize command.
> > >
> > > You wouldn't want it anyway.
> > >
> > > There isn't any way to "resize" an fault-tolerant RAID
> > > system (1/3/5) and you can get more flexable,
> > This is not correct. You can extend most of the common hardware raid
> > systems with raid5. Most of it during operation without shutting down.
> >
> > > reliable
> > > results with LVM than RAID0 anyway.
> > >
> > > The way to add space into a VG is by incorporating new
> > > PV's with vgextend. If you want to grow a fault-tolerant
> > > system then build new RAID5 groups and add them as new
> > > PV's.
> > But if I want to extend my system, I need at least three disks. If I
> > extend an existing raid5, I need only one disk.
> >
> > best regards
> >
> > Ralf Eisinger
> > -------------------
> > IHS - University of Stuttgart, Germany
> > Pfaffenwaldring 10
> > 70150 Stuttgart
> > Phone: (0049)-711-685-3201
> >
> >
> >
> > _______________________________________________
> > 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
>
> --
>   Tim Triche, Jr.
>   Senior Engineer
>   ConnectLive Communications, Inc.
>   (202) 513-1000 || (202) 253-5666
>
> _______________________________________________
> 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
>

Mit freundlichen Grüßen

Ralf Eisinger
-------------------
IHS - University of Stuttgart, Germany
Pfaffenwaldring 10
70150 Stuttgart
Phone: (0049)-711-685-3201

Re: [linux-lvm] RAID extension and lvm

[Kirby C. Bohling]

Did you partition the disk, or did you put LVM on the raw disk 
unpartitioned?

If the disk is partitioned, run fdisk/cfdisk/disk druid to add that 
partition on the end (I've never done this, so take it with a grain of 
salt).  I don't believe that the partition table has the complete size 
of the disk, I believe it just has the sizes of the current partitions. 
  So it should be safe (the key work being: should).

Run pvcreate on the newly created partition.

Now follow the HOWTO on adding another pv to a vg.  I believe all you do 
is run vgextend then lvextend or lvcreate.

If you didn't set it up using partitions, and it is a raw block device 
you have issues.  I've got no idea how to help.  Sorry...  Report back 
if what works :-)

	Thanks,
		Kirby


Ralf Eisinger wrote:
> Sorry, I think there is/was something confusing. We want to have the
> following scenario:
> 
> - we had an hardware raid5 with three disks (so we get the size from two
>   of them)
> - from the view of the kernel there is one big SCSI disk
> - on that disk I install lvm (for snapshots) and ext3
> - ...
> - we are working (heavily ;-) on that disks
> - ...
> - Because we are running out of disk space, I plug in a new disk
>   and add them to the existing raid5. So it will have now four disks.
>   This adding is done by the hardware.
> - Now the disk size of the "one big disk" is changed to the new
>   "netto" size of the raid5.
> 
> Is it possible to handle that with lvm and when the answer is "yes", in
> which way? One point was a command named "pvresize" but I didn't find it
> in my distribution (SuSE 8.0) and I find nothing about that in the HOWTO.
> 
> 

Re: [linux-lvm] RAID extension and lvm

[Christian Limpach]

Quoting Ralf Eisinger <rei@ihs.uni-stuttgart.de>:

> Is it possible to handle that with lvm and when the answer is "yes", in
> which way? One point was a command named "pvresize" but I didn't find it
> in my distribution (SuSE 8.0) and I find nothing about that in the HOWTO.

You want to (find and) use the pvresize command, it is available at least in 
1.1-rc2: ftp://linux.msede.com:21/lvm/1.1/lvm_1.1-rc2.tar.gz
This should work well if your pv spans your raid5 device completely or at 
least your pv is at the end of your raid5 device.

If you don't want to try pvresize, you have two other options:
- if your raid5 device has a partition table, you can add the new space as a 
new partition and pvcreate a new pv on this partition and add the new pv to 
your volume group.
- you can use losetup with the -o option to create a loopback device covering 
the new space.  This is definitely not an ideal setup, but it should work and 
most of all it should work even if you couldn't use pvresize.

Then there's a few more options if you use device-mapper which lets you map 
multiple ranges of sectors into a device.

-- 
Christian Limpach <chris@Pin.LU>