4.2. LVM Configuration

4.2. LVM Configuration
Prev	Chapter 4. Hardware and Software Configuration	Next

4.2.1. Logical Volume Management (LVM)

The Logical Volume Management framework allows one to group different block devices (disks, partitions, RAID arrays...) together as a single larger device, the volume group (VG).

Individual devices used to form a volume group are referred to as physical volumes (PV). Physical volumes once regrouped into a volume group lose their individual character.

Instead the entire volume group is divided into physical extents (PE) of fixed size (4MB by default) from which logical volumes (LV) are created. A logical volume can be thought of as a partition.

4.2.2. Kernel and software components

The LVM kernel module is lvm-mod.o. The software tools are installed by the lvm package which provides in particular /sbin/vgscan. This command will start the LVM environment by scanning all the volume groups and build the /etc/lvmtab as well as databases in /etc/lvmtab.d/ which are used by all the other LVM tools.

Table 4.1. Main LVM tools

PV tools	pvcreate, pvmove, pvchange, pvdisplay, pvscan ...
VG tools	vgcreate, vgremove, vgchange, vgdisplay, vgscan ...
LV tools	lvcreate, lvextend, lvreduce, lvremove, lvchange, lvscan ...

We won't need to use or know all the above tools. We will rather focus on the various LVM components (as depicted in the diagram) and the commands needed to create these components: pvcreate, vgcreate and lvcreate.

Create a volume group called volumeA with three physical volumes (3 partitions in this case) and create a logical volume called lv0 of size 150MB initially.

Run vgscan to create the /etc/lvmtab file
Create three new partitions (say /dev/hda16, /dev/hda17, /dev/hda18) of 100MB each. Make sure you toggle the partition type to 8E (Linux LVM). Then reboot.

Prepare the physical volumes

pvcreate /dev/hda16
pvcreate /dev/hda17
pvcreate /dev/hda18

Create a volume group called volumeA with the above physical volumes:
```
vgcreate volumeA /dev/hda16 /dev/hda17 /dev/hda18
```
This will create a directory called /dev/volumeA/. The default PE size of 4MB will be used, one can change this with the -s option.
Create a logical volume called lv0 of size 150MB on this volume group
```
lvcreate -L 150M -n lv0 volumeA
```
This will create the block device /dev/volume1/lv0

Make a filesystem on lv0 and mount it on /mnt/lvm

mkfs -t ext3 /dev/volumeA/lv0
mkdir /mnt/lvm
mount /dev/volumeA/lv0 /mnt/lvm

This wouldn't be very different from other partition types if it weren't for the possibility to change the logical volume's size at any time.

Let's first show how to reduce the existing 150MB logical volume lv0 with the e2fsadm tool installed by the lvm package.

umount /mnt/lvm
e2fsadm -L 25 /dev/volumeA/lv0

	Note
	The `-L` option refers to size in megabytes. The is the case with most LVM tools. The `-l` option can be used to specify logical extents (LE) instead. The default size of an LE is 4MB.

The next section will show how to add a new physical volume (a disk) to a volume group and demonstrates how an existing logical volume can be made larger by including physical extents available in the volume group to itself. Once this is done the e2fsadm tool will resize the filesystem across the logical volume.

4.2.3. Extending the Volume Group with a RAID 0 device

So far we have

VG = /dev/hda16 +
/dev/hda17 + /dev/hda18

and we would like to add a RAID0 device to this

Create three more partitions (eg /dev/hda19, /dev/hda20 and /dev/hda21) of size 50MB and of type "Linux raid autodetect" (FD) - reboot!

Edit /etc/mtab to add the following RAID 0 device:

raiddev /dev/md1
        raid-level 0
        nr-raid-disks 3
        nr-spare-disks 0
        persistent-superblock 1
        chunk-size 4
        device /dev/hda19
        raid-disk 0
        device /dev/hda20
        raid-disk 1
        device /dev/hda21
        raid-disk 2

Start the raid meta device:
```
mkraid /dev/md1
raidstart /dev/md1
```
Add this device to the Volume Group volumeA
Before adding the device to the volume group run pvscan to see which physical volumes are available. Notice that /dev/md1 is not listed.
We now prepare /dev/md1 as a PV (physical volume):
```
pvcreate /dev/md1
```
When running pvscan again the output should look like the following. Notice that /dev/md1 is now listed.
```
pvscan
pvscan -- reading all physical volumes (this may take a while...)
pvscan -- ACTIVE  PV "/dev/md1"   is in no VG [305.62 MB]
pvscan -- ACTIVE  PV "/dev/hda10" of VG "volumeA"[96 MB / 0 free]
pvscan -- ACTIVE  PV "/dev/hda11" of VG "volumeA"[96 MB / 0 free]
pvscan -- ACTIVE  PV "/dev/hda12" of VG "volumeA"[96 MB / 84 MB free]
pvscan -- total:4[611.46 MB]  /in use:3[305.83 MB]  /in no VG:1 [305.62 MB]
```
We next add the device /dev/md1 to the volume group volumeA:
```
vgextend volumeA /dev/md1
```
At this stage the volume group has four devices:
```
VolumeA = /dev/hda10 +
/dev/hda11 + /dev/hda12 +
/dev/md1
```
We can take 50MB from /dev/md1 and add them to lv0 (unmount the volume first)
```
lvextend -L +50 /dev/volumeA/lv0 /dev/md1
```
The original lv0 volume had 150 megabytes. The + flag in front of the requested size has added 50MB to the logical volume, making it about 200 megabytes. But we haven't extended the filesystem across the entire logical volume yet.
The output of lvscan will show 80MB available. This corresponds to the 25 megabytes resizing done with e2fsadm on p. XXXX plus the 50MB added by lvextend above.
```
lvscan
lvscan -- ACTIVE            "/dev/volumeA/lv0" [80 MB]
lvscan -- 1 logical volumes with 80 MB total in 1 volume group
lvscan -- 1 active logical volume
```
The next command will extend the filesystem to 80 megabytes:
```
e2fsadm -L 80 /dev/volume/lv0
```
If you remount this volume on /mnt/lvm you can see the new available space with df.

Reboot Warning

	Reboot Warning
The LVM tools need the `lvm-mod.o` module and in our case the metadevice `/dev/md1`. You need to create a new `initrd` with mkinitrd or add the following lines to a new `initrd`: insmod /lib/lvm-mod.o raidautorun /dev/md1 The volume group is then activated with vgscan from the rc.sysinit script.

The LVM tools need the lvm-mod.o module and in our case the metadevice /dev/md1. You need to create a new initrd with mkinitrd or add the following lines to a new initrd:

insmod /lib/lvm-mod.o
raidautorun /dev/md1

The volume group is then activated with vgscan from the rc.sysinit script.

4.2.4. Booting from a logical volume root device

As with software RAID we are going to investigate some issues we need to consider when using LVM on the root device.

First make sure the volume we have created previously is mounted. If it isn't then do

mount /dev/volumeA/lv0 /mnt/lvm

Next we archive the root device in the same way as we did for RAID:

tar clvf - / | (cd /mnt/lvm/; tar xvf -)

Edit /mnt/lvm/etc/fstab and enter

/dev/volumeA/lv0   /     ext2    defaults 0 1

Edit /etc/lilo.conf or /etc/grub.conf to add a new entry where the kernel points to the new root logical volume. For a 2.4.22 kernel an additional entry in /etc/grub.conf looks like this:

title lvm-root
        root (hd0,1)
        kernel /vmlinuz-2.4.22 ro root=LABEL=/ 
        initrd /initrd-2.4.22-lvm.img

All we need is the initrd initrd-2.4.22-lvm.img.

Once again we will run mkinitrd with --fstab=fstab which we will use to make the script read our new fstab file /mnt/lvm/etc/fstab.

We test this:

mkinitrd --fstab=/mnt/lvm/etc/fstab /boot/initrd-lvm.img $(uname -r)

If we mount this initial ram disk we can see that this is going to work by looking at the linuxrc script.

# linuxrc
echo "Loading lvm-mod.o module"
insmod /lib/lvm-mod.o
echo Creating block devices
mkdevices /dev
echo Scanning logical volumes
vgscan
echo Activating logical volumes
vgchange -ay
----snip---