Logical Volume Management (LVM)

What Is LVM?

LVM (Logical Volume Management) is a storage abstraction layer built into the Linux kernel (via the device mapper framework). Instead of working directly with fixed partitions, LVM lets you:

• Pool multiple physical drives/partitions into a single storage group
• Carve flexible logical volumes from that pool - resized, moved, or snapshotted while online
• Name volumes meaningfully: databases, backups, www

The current implementation is LVM2, which is what all modern Linux distributions use.

Why Use LVM?

Problem	LVM Solution
”Disk is full, but I can’t resize the partition”	Add a new disk to the volume group and extend the LV online
”I need to take a snapshot before upgrading”	LVM snapshots are instant and space-efficient
”I want data spread across multiple disks”	LVM striping (like RAID 0)
“I need redundancy”	LVM mirroring
”I need to migrate data off a failing disk”	pvmove migrates data live without downtime

---

LVM Architecture

LVM layers abstractions on top of physical storage:

Physical Disks / Partitions
         |
    Physical Volumes (PV)    - Raw storage marked for LVM use
         |
     Volume Group (VG)       - Pool combining one or more PVs
         |
    Logical Volumes (LV)     - Flexible "virtual partitions" carved from VG
         |
     Filesystem (ext4, xfs)  - Formatted and mounted like any partition

The Three Layers

Layer	Command prefix	Description
Physical Volume (PV)	pv...	A disk, partition, or RAID array marked for LVM. LVM writes a PV header to claim it.
Volume Group (VG)	vg...	A pool combining one or more PVs. Presents unified storage.
Logical Volume (LV)	lv...	A slice of a VG. Acts like a partition - you format and mount it.

Extents - The Allocation Unit

Each volume is carved into small, fixed-size chunks called extents:

• Physical extents (PE) on PVs
• Logical extents (LE) on LVs
• All volumes in a VG share the same extent size (default: 4 MB)

An LV is simply a mapping from logical extents to physical extents. This mapping can be:

• Linear: extents in sequence (default)
• Striped: extents alternating across multiple PVs (performance)
• Mirrored: extents duplicated across PVs (redundancy)

Because extents are the unit of allocation, LVM can expand, shrink, or move an LV simply by adjusting these mappings - even while the LV is mounted.

---

Practical Guide

1. Mark Disks as Physical Volumes

# Scan for available block devices LVM can use
sudo lvmdiskscan

# Mark devices as physical volumes
sudo pvcreate /dev/sda /dev/sdb

# Check PV status
sudo pvs                            # brief summary
sudo pvdisplay                      # detailed info
sudo pvdisplay /dev/sda             # specific PV

PV         VG   Fmt  Attr PSize   PFree
/dev/sda        lvm2 ---  200.00g 200.00g
/dev/sdb        lvm2 ---  100.00g 100.00g

Caution

pvcreate destroys existing data on the device. Always verify you’re targeting the correct device. There is no confirmation prompt.

2. Create a Volume Group

# Create VG named LVMVolGroup from both PVs
sudo vgcreate LVMVolGroup /dev/sda /dev/sdb

# View VG status
sudo vgs                            # brief
sudo vgdisplay                      # detailed
sudo vgdisplay LVMVolGroup

# Add a new PV to an existing VG later
sudo vgextend LVMVolGroup /dev/sdc

# Remove a PV from a VG (must evacuate first)
sudo pvmove /dev/sda               # move all data off /dev/sda
sudo vgreduce LVMVolGroup /dev/sda

3. Create Logical Volumes

Unlike partitions, you don’t need to know disk layout - LVM handles placement.

# Create LVs with fixed size (-L) and name (-n)
sudo lvcreate -L 10G -n projects LVMVolGroup
sudo lvcreate -L 5G  -n www      LVMVolGroup
sudo lvcreate -L 20G -n db       LVMVolGroup

# Allocate remaining free space
sudo lvcreate -l 100%FREE -n workspace LVMVolGroup

# View LV status
sudo lvs                            # brief
sudo lvdisplay                      # detailed
sudo vgs -o +lv_size,lv_name       # VG view with LV details

LV        VG          Attr       LSize
projects  LVMVolGroup -wi-a-----  10.00g
www       LVMVolGroup -wi-a-----   5.00g
db        LVMVolGroup -wi-a-----  20.00g
workspace LVMVolGroup -wi-a----- 264.99g

LV device paths:

• /dev/LVMVolGroup/projects
• /dev/mapper/LVMVolGroup-projects (same device, two paths)

4. Format and Mount

# Format
sudo mkfs.ext4 /dev/LVMVolGroup/projects
sudo mkfs.xfs  /dev/LVMVolGroup/db         # XFS is a good choice for databases

# Create mount points and mount
sudo mkdir -p /mnt/{projects,www,db,workspace}
sudo mount /dev/LVMVolGroup/projects  /mnt/projects
sudo mount /dev/LVMVolGroup/www       /mnt/www
sudo mount /dev/LVMVolGroup/db        /mnt/db
sudo mount /dev/LVMVolGroup/workspace /mnt/workspace

5. Persist Mounts in /etc/fstab

/dev/LVMVolGroup/projects  /mnt/projects  ext4  defaults,nofail  0  2
/dev/LVMVolGroup/www       /mnt/www       ext4  defaults,nofail  0  2
/dev/LVMVolGroup/db        /mnt/db        xfs   defaults,nofail  0  2
/dev/LVMVolGroup/workspace /mnt/workspace ext4  defaults,nofail  0  2

---

Resizing Logical Volumes

This is one of LVM’s most powerful features - you can grow a volume while it is mounted and in use.

Extend a Logical Volume

# 1. Extend the LV
sudo lvextend -L +10G /dev/LVMVolGroup/projects   # add 10G
sudo lvextend -L 50G  /dev/LVMVolGroup/projects   # set absolute size to 50G
sudo lvextend -l +100%FREE /dev/LVMVolGroup/projects  # use all free space

# 2. Resize the filesystem (must be done separately)
sudo resize2fs /dev/LVMVolGroup/projects           # ext4 (online resize OK)
sudo xfs_growfs /mnt/db                           # XFS (must be mounted)

# Or do both in one step (-r extends filesystem automatically)
sudo lvextend -r -L +10G /dev/LVMVolGroup/projects

Shrink a Logical Volume

Caution

Shrinking is risky and cannot be done on XFS. For ext4, you must unmount first:

# 1. Unmount
sudo umount /mnt/projects

# 2. Check filesystem
sudo e2fsck -f /dev/LVMVolGroup/projects

# 3. Shrink filesystem FIRST (leave some buffer)
sudo resize2fs /dev/LVMVolGroup/projects 8G

# 4. Then shrink the LV
sudo lvreduce -L 8G /dev/LVMVolGroup/projects

# 5. Remount
sudo mount /dev/LVMVolGroup/projects /mnt/projects

Always shrink the filesystem before shrinking the LV. Shrink the LV before removing space from the VG.

---

LVM Snapshots

A snapshot creates an instant point-in-time copy of an LV. Under the hood, LVM uses Copy-on-Write (CoW): the snapshot initially shares all data blocks with the origin. When blocks are modified in the origin, LVM copies the original version to the snapshot before overwriting.

# Create a 5G snapshot of the 'db' LV
sudo lvcreate -s -L 5G -n db_snapshot /dev/LVMVolGroup/db

# Mount the snapshot (read-only for backups is safest)
sudo mount -o ro /dev/LVMVolGroup/db_snapshot /mnt/db_snapshot

# Take a backup from the snapshot (database is still running normally)
sudo tar -czf /backup/db_$(date +%Y%m%d).tar.gz /mnt/db_snapshot

# After backup, unmount and remove snapshot
sudo umount /mnt/db_snapshot
sudo lvremove /dev/LVMVolGroup/db_snapshot

Tip

Size your snapshot based on the rate of change during the snapshot window, not the full volume size. A 5G snapshot for a 200G database that changes slowly is fine. If the snapshot fills up, it becomes invalid.

---

Migration and Maintenance

# Move data from one PV to another (online, no downtime)
sudo pvmove /dev/sda                # move all from sda to other PVs in VG
sudo pvmove /dev/sda /dev/sdc       # move sda data specifically to sdc

# Rename a VG
sudo vgrename OldName NewName

# Rename an LV
sudo lvrename LVMVolGroup old-lv new-lv

# Remove an LV (must be unmounted)
sudo umount /mnt/projects
sudo lvremove /dev/LVMVolGroup/projects

# Remove a VG (all LVs must be removed first)
sudo vgremove LVMVolGroup

# Remove PV label (must be out of all VGs first)
sudo pvremove /dev/sda

---

Quick LVM Command Reference

Category	Command	Description
PV	pvcreate /dev/sdX	Initialize PV
	pvs	Brief PV list
	pvdisplay	Detailed PV info
	pvmove /dev/sdX	Migrate data off a PV
VG	vgcreate name /dev/sdX	Create VG
	vgs	Brief VG list
	vgdisplay	Detailed VG info
	vgextend name /dev/sdX	Add PV to VG
	vgreduce name /dev/sdX	Remove PV from VG
LV	lvcreate -L size -n name VG	Create LV
	lvs	Brief LV list
	lvdisplay	Detailed LV info
	lvextend -r -L +size /dev/VG/LV	Grow LV + filesystem
	lvreduce -L size /dev/VG/LV	Shrink LV (unmount first!)
	lvcreate -s -L size -n snap /dev/VG/LV	Create snapshot
	lvremove /dev/VG/LV	Delete LV
Scan	lvmdiskscan	Find available block devices
	lvs -a	List all LVs including internal