RHCSA V10
Part of RHCSA V10. Builds directly on maintain-basic-storage — LVM sits on top of raw partitions, and RHEL’s default install layout is already LVM (see installing-rhel), so this is the normal state of things, not an advanced edge case.
Why LVM exists
Plain partitions are fixed the moment you create them — growing /var when it fills up means backup, repartition, restore. LVM (Logical Volume Manager) adds a layer of abstraction between physical disks and the filesystems on top of them, so volumes can be resized, moved, and combined across physical disks without touching the filesystem layout underneath, mostly live.
Physical Volume (PV) → Volume Group (VG) → Logical Volume (LV)
raw disk/partition pool of storage the thing you
initialized for LVM combining 1+ PVs actually format
and mount
Building the stack
sudo pvcreate /dev/sdb /dev/sdc # initialize one or more disks/partitions as PVs
sudo pvs # list PVs, quick summary
sudo pvdisplay # list PVs, full detail
sudo vgcreate data-vg /dev/sdb /dev/sdc # combine PVs into a volume group — the "pool"
sudo vgs # list VGs
sudo vgdisplay # full detail, including free extents
sudo lvcreate -L 20G -n data-lv data-vg # carve a 20GB logical volume out of the VG
sudo lvcreate -l 100%FREE -n data-lv data-vg # or take ALL remaining free space instead of a fixed size
sudo lvs # list LVs
sudo lvdisplay # full detailLVs show up as devices at /dev/data-vg/data-lv (or /dev/mapper/data--vg-data--lv) — format and mount them exactly like a regular partition:
sudo mkfs.xfs /dev/data-vg/data-lv
sudo mount /dev/data-vg/data-lv /mnt/dataGrowing a volume group
Ran out of space in the VG entirely? Add another PV:
sudo pvcreate /dev/sdd
sudo vgextend data-vg /dev/sddResizing a logical volume — and the filesystem on top of it
Resizing the LV and resizing the filesystem inside it are two separate steps. Growing the LV without growing the filesystem just leaves unused space the filesystem doesn’t know about.
sudo lvextend -L +10G /dev/data-vg/data-lv # grow the LV by 10GB
sudo lvextend -l +100%FREE /dev/data-vg/data-lv # or grow to consume all free space in the VGThen grow the filesystem to match — the command depends on the filesystem type:
sudo xfs_growfs /mnt/data # XFS — note: takes the MOUNTPOINT, not the device, and the fs must be mounted
sudo resize2fs /dev/data-vg/data-lv # ext4 — takes the DEVICE, works mounted or unmountedA common one-liner combining both steps:
sudo lvextend -L +10G -r /dev/data-vg/data-lv # -r = automatically resize the filesystem after growing the LVShrinking
This is where the XFS-can’t-shrink limitation from maintain-basic-storage fully bites. For an XFS volume that needs to be smaller, there’s no in-place shrink — the only path is: back up the data, destroy and recreate the LV at the smaller size, format fresh, restore the data.
ext4 can shrink, but order matters and it’s the reverse of growing:
sudo umount /mnt/data
sudo e2fsck -f /dev/data-vg/data-lv # forced check — required before resize2fs will shrink
sudo resize2fs /dev/data-vg/data-lv 15G # shrink the FILESYSTEM first
sudo lvreduce -L 15G /dev/data-vg/data-lv # THEN shrink the LV to matchShrinking the LV before the filesystem risks truncating live filesystem data that’s still sitting in the space being removed — this ordering (filesystem first, then LV) is the exam-relevant gotcha to remember.
Replacing or removing a physical volume safely
Growing (above) is the common case; the other direction — a PV needs to come out of a VG, because the underlying disk is failing, undersized, or just being decommissioned — needs its data moved off first, not just deleted out from under the filesystem sitting on top of it.
pvmove -A y /dev/sdb1 # move all data OFF this PV onto other PVs with free space in the same VG/dev/sdb1: Moved: 2.74%
...
/dev/sdb1: Moved: 100.00%
-A y auto-backs-up the VG’s metadata (via vgcfgbackup) once the move finishes — worth doing every time, since pvmove needs the other PVs in the VG to have enough free space to absorb what’s being moved, and a power loss mid-move can leave the VG in an inconsistent, data-losing state — back up the actual data on the LVs beforehand too, not just the metadata.
Once the PV is empty, remove it from the VG, then wipe it:
vgreduce vg01 /dev/sdb1 # remove the now-empty PV from the volume group
pvremove /dev/sdb1 # wipe the PV label — the device is now free for any other useThis pvmove → vgreduce → pvremove sequence — evacuate, then detach, then wipe — is the safe order specifically because it never removes a PV while it might still hold live data.
Cleaning up
sudo umount /mnt/data
sudo lvremove /dev/data-vg/data-lv
sudo vgremove data-vg
sudo pvremove /dev/sdb /dev/sdcOrder matters here too — you can’t remove a VG that still has LVs in it, or a PV that’s still part of a VG.
Stratis — the newer alternative
RHEL 8+ ships Stratis, a volume-managing filesystem that wraps LVM + XFS complexity behind a simpler CLI, aimed at making thin provisioning and snapshots easier without hand-managing PV/VG/LV layers directly.
sudo systemctl enable --now stratisd
sudo stratis pool create data-pool /dev/sdb
sudo stratis filesystem create data-pool data-fsMounts at /stratis/data-pool/data-fs. Worth knowing it exists and roughly what it’s for — LVM is still the primary thing tested and used in practice, but Stratis shows up in newer RHEL objectives as the direction storage management is heading.
Quick reference: which layer am I resizing?
| Symptom | Fix at this layer |
|---|---|
| VG has no free space | pvcreate + vgextend — add more physical storage |
| VG has free space, LV is small | lvextend |
| LV is the right size, filesystem still reports old size | xfs_growfs (mounted, XFS) or resize2fs (ext4, mounted or not) |
Software engineering
Infrastructure & security
Computer science
Hardware & engineering