This is an internal documentation. There is a good chance you’re looking for something else. See Disclaimer.
LVM¶
Work through this document from top to bottom. A LVM is set up first, then various features are shown and, finally, LVM is removed again.
Tip
lvm2 and thin-provisioning-tools need to be installed:
apt install lvm2 thin-provisioning-tools
Terminology¶
PV (Physical Volume): a disk used for storage
VG (Volume Group): group of one or more disks (PVs)
LV (Logical Volume): a volume that can be accessed as virtual disk (similar to a partition)
Thin provisioning: Storage that’s not fully backed (only space used by FS is used on disk)
Set up Test Drives¶
Create storage for virtual disks:
truncate -s 2G disk1 disk2
Create virtual disks:
l1=$(losetup --show -f disk1)
l2=$(losetup --show -f disk2)
Creating Groups and Volumes¶
Create Volume Group (name main):
lvm vgcreate main $l1
Create a Logical Volume (named vol1):
lvm lvcreate -l 20%FREE -n vol1 main
Check VG/LV/PV status:
lvm vgs
lvm lvs
lvm pvs
Or verbose status:
lvm vgdisplay
lvm lvdisplay
lvm pvdisplay
Create pool for thin provisioning (named pool):
lvm lvcreate -l 60%FREE --thin main/pool
Create thin volume in pool (named vol2):
lvm lvcreate -V 1G --thin main/pool -n vol2
Show volumes:
lvm lvs
You should now see the thin pool main/pool and main/vol2. Both should use 0.00% data yet.
Create filesystem:
mkfs.ext4 /dev/mapper/main-vol2
Mount volume:
mkdir m
mount -o discard /dev/mapper/main-vol2 m
Tip
-o discard is needed to tell LVM about freed blocks.
Alternatively, fstrim m can be used to free space
on the underlying Logical Volume.
Write a file:
dd if=/dev/urandom of=m/file bs=1M count=50
Show volumes again:
lvm lvs
You should now see that some Data is used, on both the pool and volume.
Thin Volume Snapshots¶
Create and activate snapshot:
lvm lvcreate -k n -n vol2-snapshot1 -s main/vol2
A snapshot can be mounted just like any other volume:
mkdir n
mount -o discard /dev/mapper/main-vol2--snapshot1 n
Remove file again:
rm m/file
Umount volumes:
umount m n
Non-Thin Snapshots¶
Non-thin snapshots work slightly differently. You create a snapshot and give it a certain size. The size determines how much the origin can divert from the snapshot. Modifications to the snapshot and the origin are written (CoW-style) to the snapshot. Once a snapshot is no longer needed, it should be merged backed to the origin to make storage linear again.
Create and activate snapshot:
lvm lvcreate -k n -n vol1-snapshot1 -s main/vol1 -L 1G
Warning
Be sure that the specified size is large enough to hold any changes to the origin and snapshot volume. You’ll see an out-of-disk-space error otherwise.
Show snapshots and their origins:
lvm lvs
Merge snapshot back into origin:
lvconvert --merge main/vol1-snapshot1
Resize Physical Volume¶
In case the size of the underlying device changes, the corresponding PV needs to be resized too.
Resize disk:
truncate -s 4G $l1
losetup --set-capacity $l1
Resize PV to the size of underlying device:
lvm pvresize $l1
Check size:
lvm pvs
Volume group should now show some more free space:
lvm vgs
Resize Volume¶
Check available space (VFree):
lvm vgs
If no free space is available, check if spare disks are available within the RAID, add more disks, or replace existing disks with larger disks and grow the raid.
Resize Volume:
lvm lvresize main/vm1 -L +200M
You’ll also have to resize the filesystem:
# ext4
resize2fs /dev/mapper/main-vm1
# btrfs
btrfs filesystem resize max ${mountpoint}
Add Drive to VG¶
Add drive:
lvm vgextend main $l2
Show PVs:
lvm pvs
There should be some free space in the VG now:
lvm vgs
Starting LVM manually¶
Usually LVMs are started automatically. Should this not be the case for any reason, do it manually.
Scan drives for LVMs:
lvm pvscan
Once it detected the LVM you can:
activate all LVs in all VGs:
lvm vgchange --activate y
activate all LVs of a specific VG:
lvm vgchange --activate y main
activate a specific LV:
lvm lvchange --activate y main/vm1
Clean Up¶
Clean up virtual disks:
losetup -d $l1 $l2
rm disk[1-2]