Why LVM

LVM has three layers. Each layer is a different unit of abstraction.

Raw disk or partition → Physical Volume (PV) → Volume Group (VG) → Logical Volume (LV) → filesystem.

• Physical Volume (PV): a disk or partition that has been initialised for LVM. It is the raw ingredient. Any block device works: /dev/sdb, /dev/sdc1, even a software RAID device (/dev/md0).
• Volume Group (VG): a pool that aggregates one or more PVs. The VG is where disk space lives. Adding a PV to a VG grows the pool. Removing a PV drains it.
• Logical Volume (LV): a named slice carved out of the VG’s pool. An LV looks like a block device (/dev/myvg/data) and holds a filesystem. You grow, shrink, snapshot, and move LVs without touching partitions or disks directly.

# Check what LVM knows about each layer
pvs         # list physical volumes
vgs         # list volume groups
lvs         # list logical volumes

# More detail
pvdisplay /dev/sdb
vgdisplay myvg
lvdisplay /dev/myvg/data

Create a PV: initialise a block device for LVM.

# First, identify the disk (do NOT do this to a disk with data you need)
lsblk
# NAME   MAJ:MIN RM  SIZE RO TYPE MOUNTPOINTS
# sdb      8:16   0   50G  0 disk
# sdc      8:32   0   50G  0 disk

# Initialise /dev/sdb as a physical volume
sudo pvcreate /dev/sdb
#   Physical volume "/dev/sdb" successfully created.

# And /dev/sdc for a two-disk pool
sudo pvcreate /dev/sdc
#   Physical volume "/dev/sdc" successfully created.

# Verify
sudo pvs
#   PV         VG   Fmt  Attr PSize   PFree
#   /dev/sdb        lvm2 ---  <50.00g <50.00g
#   /dev/sdc        lvm2 ---  <50.00g <50.00g
# VG column is empty — not yet assigned to a group

pvcreate writes an LVM metadata header to the device. The disk’s data is effectively replaced — do this only on a blank or dedicated disk.

Create a VG: pool the PVs together.

# Create a volume group named "datavg" spanning both PVs
sudo vgcreate datavg /dev/sdb /dev/sdc
#   Volume group "datavg" successfully created

sudo vgs
#   VG     #PV #LV #SN Attr   VSize    VFree
#   datavg   2   0   0 wz--n- <99.99g <99.99g

# The VG now has ~100 GB of pooled space (2 × 50 GB)
# #LV is 0 — nothing carved out yet

# You can add more PVs to an existing VG later:
# sudo vgextend datavg /dev/sdd

The VG name is how all subsequent LVM commands refer to this pool. Choose something meaningful: vg0, datavg, prod-vg. The name is embedded in device paths.

Create an LV: carve a named slice from the VG.

# Create a 30 GB logical volume named "appdata" in datavg
sudo lvcreate -L 30G -n appdata datavg
#   Logical volume "appdata" created.

# Create a second LV using 20% of remaining free space
sudo lvcreate -l 20%FREE -n logs datavg
#   Logical volume "logs" created.

sudo lvs
#   LV      VG     Attr       LSize
#   appdata datavg -wi-a----- 30.00g
#   logs    datavg -wi-a-----  14.00g

# The LV is now a block device at two equivalent paths:
ls -l /dev/datavg/appdata
# lrwxrwxrwx  /dev/datavg/appdata -> ../dm-0
ls -l /dev/mapper/datavg-appdata
# Same device, different symlink

# Format and mount the LV exactly like any block device:
sudo mkfs.ext4 /dev/datavg/appdata
sudo mkdir /mnt/appdata
sudo mount /dev/datavg/appdata /mnt/appdata

# Add to /etc/fstab using the device mapper path (stable):
# /dev/datavg/appdata  /mnt/appdata  ext4  defaults,nofail  0  2

Understand what LVM buys you over raw partitions.

The naming benefit is underrated. /dev/myvg/postgres survives disk replacements and server migrations. /dev/sda2 is hardware topology baked into your config.

# See the complete picture: which PVs back which LVs
sudo pvdisplay --maps
#   --- Physical volume ---
#   PV Name               /dev/sdb
#   VG Name               datavg
#   ...
#   --- Physical Segments ---
#   Physical extent 0 to 7679:
#     Logical volume    /dev/datavg/appdata
#     Logical extents   0 to 7679