2 of 3

Storage and LVMLesson 2 of 3

Logical volumes: PV, VG and LV

Turn partitions into physical volumes, pool them into a volume group with a chosen extent size, and carve out logical volumes.

  • pvcreate
  • vgcreate
  • lvcreate
  • pvs
  • vgs
  • lvs
  • pvdisplay
  • vgdisplay
  • lvdisplay
  • vgcreate -s
  • lvcreate -l

50 min5 exercises

After this lesson you can

  • Explain how physical volumes, volume groups, physical extents and logical volumes relate.
  • Create a volume group with a specified physical extent size and a logical volume sized in megabytes or in extents.
  • Put a file system on a logical volume, mount it persistently and read the layout with the *s and *display commands.

The idea

Partitions are fixed slices. LVM (Logical Volume Manager) adds a flexible layer so storage can be pooled, resized and moved while the system runs. Three layers:

LayerNameWhat it is
BottomPhysical volume (PV)A disk or partition that LVM has claimed
MiddleVolume group (VG)A pool made of one or more PVs
TopLogical volume (LV)A slice of the pool that you format and mount

A VG divides its space into equal chunks called physical extents (PE), 4 MiB by default. A logical volume is always a whole number of extents. That explains two exam favourites:

  • The PE size is set when the VG is created: vgcreate -s 8M. A task may demand a specific one.
  • Sizes can be given as -L 100M (megabytes) or -l 10 (extents). With 8 MiB extents, 100 MiB is not a multiple of 8, so LVM rounds up to 104 MiB. -l 10 gives exactly 80 MiB. When a task says “approximately”, you’re allowed this rounding; if it says extents, give extents.
TaskPVVGLV
Createpvcreate DEVvgcreate [-s SIZE] VG DEV...lvcreate -n NAME -L SIZE VG
Short listpvsvgslvs
Detailedpvdisplayvgdisplaylvdisplay
Removepvremovevgremovelvremove

The result is a device node at /dev/VG/LV, and the same one at /dev/mapper/VG-LV. Hyphens in names are doubled inside /dev/mapper, so avoid hyphens in VG and LV names.

Try it

sudo pvcreate /dev/vdb2
sudo vgcreate -s 8M vgdata /dev/vdb2
sudo vgs vgdata
  Physical volume "/dev/vdb2" successfully created.
  Volume group "vgdata" successfully created
  VG     #PV #LV #SN Attr   VSize   VFree
  vgdata   1   0   0 wz--n- 296.00m 296.00m

296 MiB rather than 300: LVM keeps some space for metadata and drops the remainder that does not fill an extent (37 × 8 MiB = 296 MiB).

sudo vgdisplay vgdata | grep -E 'PE Size|Total PE|Alloc PE|Free  PE'
  PE Size               8.00 MiB
  Total PE              37
  Alloc PE / Size       0 / 0
  Free  PE / Size       37 / 296.00 MiB

Two ways to size a volume

sudo lvcreate -n lvweb  -L 100M vgdata
sudo lvcreate -n lvlogs -l 10   vgdata
sudo lvs vgdata
  Rounding up size to full physical extent 104.00 MiB
  Logical volume "lvweb" created.
  Logical volume "lvlogs" created.
  LV     VG     Attr       LSize
  lvlogs vgdata -wi-a----- 80.00m
  lvweb  vgdata -wi-a----- 104.00m

The “Rounding up” line is LVM telling you exactly what happened. -l 100%FREE uses all remaining space; -l +50%FREE takes half of what is left.

File system and permanent mount

sudo mkfs.xfs /dev/vgdata/lvweb
sudo mkdir -p /srv/web
echo '/dev/mapper/vgdata-lvweb /srv/web xfs defaults 0 0' | sudo tee -a /etc/fstab
sudo systemctl daemon-reload
sudo mount -a
findmnt /srv/web
meta-data=/dev/vgdata/lvweb      isize=512    agcount=4, agsize=6656 blks
data     =                       bsize=4096   blocks=26624, imaxpct=25
log      =internal log           bsize=4096   blocks=1368, version=2
/srv/web /dev/mapper/vgdata-lvweb xfs rw,relatime,seclabel,attr2,inode64,logbufs=8,logbsize=32k,noquota

mount -a mounts everything in /etc/fstab that is not yet mounted. If your new line has an error, it fails now, at the prompt, rather than at boot, where a bad fstab line sends the machine into emergency mode. That is why you always run mount -a after editing it (and why findmnt --verify is worth knowing).

Undo: in reverse order

sudo umount /srv/web
sudo sed -i '\#/srv/web#d' /etc/fstab
sudo lvremove -y /dev/vgdata/lvweb /dev/vgdata/lvlogs
sudo vgremove -y vgdata
sudo pvremove -y /dev/vdb2
  Logical volume "lvweb" successfully removed.
  Logical volume "lvlogs" successfully removed.
  Volume group "vgdata" successfully removed
  Labels on physical volume "/dev/vdb2" successfully wiped.

Going further

lvcreate --type can make mirrored, striped, thin and VDO volumes; the RHCSA needs only plain (linear) volumes. lvs -o +devices shows which PV holds each LV, and pvmove migrates extents from one PV to another online, so a disk can be replaced without downtime. vgrename and lvrename rename safely if the volume is unmounted. LVM metadata is backed up automatically in /etc/lvm/backup, and vgcfgrestore can recover a deleted volume group’s layout. vgscan and lvmdevices help when disks move between machines.

Practice

0 of 5 done

Do these on your own system. They get harder. Check each result, then tick it off.

  1. Warm-up

    Read the LVM layout of a system

    On the lab VM list the physical volumes, volume groups and logical volumes with the short commands. State the volume group name that holds the root file system and how much free space it has.

    Check your result

    pvs, vgs and lvs each print a table. vgs shows VFree.

    Hint

    sudo pvs; sudo vgs; sudo lvs. lsblk shows the same tree.

    Solution
    sudo pvs
sudo vgs
sudo lvs
  2. Practice

    A volume group with 8 MiB extents

    Using partition /dev/vdb2 (300 MiB, from the previous lesson), create a physical volume, then a volume group vgdata with a physical extent size of 8 MiB. Verify the extent size and how many extents the group has.

    Check your result

    sudo vgdisplay vgdata shows PE Size 8.00 MiB and Total PE of about 37 (300 MiB ÷ 8 MiB, minus metadata).

    Hint

    pvcreate /dev/vdb2, vgcreate -s 8M vgdata /dev/vdb2.

    Solution
    sudo pvcreate /dev/vdb2
sudo vgcreate -s 8M vgdata /dev/vdb2
sudo vgdisplay vgdata | grep -E 'PE Size|Total PE|Free  PE'
  3. Practice

    Size a logical volume two ways

    Create lvweb of exactly 100 MiB using a size in megabytes, and lvlogs of 10 extents using extents. Compare their sizes with lvs. Explain why a request for 100 MiB can be rounded when the extent size is not a divisor of it.

    Check your result

    lvs vgdata shows lvweb at 104.00m (100 MiB rounded up to 13 extents of 8 MiB) and lvlogs at exactly 80.00m (10 extents).

    Hint

    lvcreate -n NAME -L 100M VG for size, lvcreate -n NAME -l 10 VG for extents. A logical volume is always a whole number of extents.

    Solution
    sudo lvcreate -n lvweb  -L 100M vgdata
sudo lvcreate -n lvlogs -l 10   vgdata
sudo lvs vgdata
sudo vgs vgdata
  4. Practice

    File system, mount point, fstab

    Put an XFS file system on lvweb, mount it on /srv/web permanently using its device-mapper path, and show that mount -a and a reboot keep it.

    Check your result

    findmnt /srv/web shows xfs on /dev/mapper/vgdata-lvweb. sudo mount -a prints nothing. After a reboot df -h /srv/web still shows it.

    Hint

    mkfs.xfs /dev/vgdata/lvweb. Both /dev/vgdata/lvweb and /dev/mapper/vgdata-lvweb name the same device. Add /dev/mapper/vgdata-lvweb /srv/web xfs defaults 0 0 to /etc/fstab.

    Solution
    sudo mkfs.xfs /dev/vgdata/lvweb
sudo mkdir -p /srv/web
echo '/dev/mapper/vgdata-lvweb /srv/web xfs defaults 0 0' | sudo tee -a /etc/fstab
sudo systemctl daemon-reload
sudo mount -a
findmnt /srv/web
df -h /srv/web
  5. Challenge

    Remove it in the right order

    Undo everything: unmount the volume and remove its fstab line, then remove the logical volumes, the volume group and the physical volume, in that order. Explain what error you get if you try to remove the volume group first.

    Check your result

    lvs, vgs and pvs no longer list lvweb, lvlogs, vgdata or /dev/vdb2. vgremove vgdata with LVs present asks for confirmation to remove them, or refuses without -f.

    Hint

    Reverse order of creation: umount, edit fstab, lvremove, vgremove, pvremove.

    Solution
    sudo umount /srv/web
sudo sed -i '\#/srv/web#d' /etc/fstab
sudo systemctl daemon-reload
sudo lvremove -y /dev/vgdata/lvweb /dev/vgdata/lvlogs
sudo vgremove -y vgdata
sudo pvremove -y /dev/vdb2