shield_lock
JinDaGe
← Back to skills
extension
Category: Development & EngineeringNo API key required

lvm

LVM (Logical Volume Manager) administration: physical volumes, volume groups, logical volumes, thin provisioning, snapshots, resizing, pvmove, and disaster recovery. Triggers on: LVM, logical volume, pvcreate, vgcreate, lvcreate, volume group, LVM snapshot, LVM thin, pvmove, vgextend, lvextend, lvreduce, thin pool, physical volume, VG, LV, PV.

personAuthor: jakexiaohubgithub
open_in_newRepositorydownloadDownload package

Identity

  • Kernel modules: dm-mod (device mapper core), dm-thin-pool (thin provisioning)
  • CLI tools: pvcreate, pvdisplay, pvs, pvremove, pvmove; vgcreate, vgdisplay, vgs, vgextend, vgreduce, vgrename, vgchange, vgscan; lvcreate, lvdisplay, lvs, lvextend, lvreduce, lvresize, lvrename, lvchange, lvscan, lvremove; lvmconfig, lvm
  • Config: /etc/lvm/lvm.conf, /etc/lvm/profile/ (per-VG overrides)
  • Metadata backups: /etc/lvm/backup/ (latest), /etc/lvm/archive/ (history)
  • Distro install: apt install lvm2 / dnf install lvm2
  • Module load: modprobe dm-mod (usually auto-loaded on first use)

Key Operations

| Operation | Command | |-----------|---------| | List physical volumes (brief) | pvs | | List physical volumes (verbose) | pvdisplay | | List volume groups (brief) | vgs | | List volume groups (verbose) | vgdisplay | | List logical volumes (brief) | lvs | | List logical volumes (verbose) | lvdisplay | | List all LVM devices | lvscan / vgscan | | Initialize a disk as PV | pvcreate /dev/sdX | | Create a volume group | vgcreate myvg /dev/sdX | | Create a linear LV | lvcreate -L 20G -n mylv myvg | | Create a striped LV (2 disks, 64K stripe) | lvcreate -L 20G -i 2 -I 64 -n mylv myvg | | Create a thin pool | lvcreate -L 50G --thinpool mypool myvg | | Create a thin volume | lvcreate -V 100G --thin myvg/mypool -n mythinlv | | Extend a VG (add disk) | vgextend myvg /dev/sdY | | Extend an LV (size + filesystem) | lvextend -L +10G /dev/myvg/mylv && resize2fs /dev/myvg/mylv | | Extend an LV to fill free space | lvextend -l +100%FREE /dev/myvg/mylv | | Reduce an LV (ext4, offline only) | e2fsck -f /dev/myvg/mylv && resize2fs /dev/myvg/mylv 15G && lvreduce -L 15G /dev/myvg/mylv | | Create a snapshot | lvcreate -L 5G -s -n mysnap /dev/myvg/mylv | | Merge a snapshot back | lvconvert --merge /dev/myvg/mysnap | | Remove a snapshot | lvremove /dev/myvg/mysnap | | Move data off a PV | pvmove /dev/sdX | | Remove a PV from VG | vgreduce myvg /dev/sdX | | Rename a volume group | vgrename oldvg newvg | | Rename a logical volume | lvrename myvg oldlv newlv | | Activate all LVs in a VG | vgchange -ay myvg | | Deactivate all LVs in a VG | vgchange -an myvg | | Scan and import VG from foreign disks | vgscan && vgimport myvg | | Check thin pool utilization | lvs -o+data_percent,metadata_percent myvg/mypool |

Expected State

  • All VGs and LVs should be active (lvs shows a in the Attr column)
  • Mounted LVs listed in /etc/fstab with noatime or appropriate options
  • Thin pool data usage below the thin_pool_autoextend_threshold (default 100%)

Health Checks

  1. vgs --units g — confirms VGs exist and shows free space; VFree should be non-zero if more LVs are planned
  2. lvs -o+lv_attr,data_percent,metadata_percent — a in attr means active; data/metadata percentages matter for thin pools
  3. pvs -o+pv_used,pv_free — confirms PVs are allocated as expected and not reporting errors

Common Failures

| Symptom | Likely cause | Check/Fix | |---------|--------------|-----------| | LV not activating on boot | Missing fstab entry or not in initrd | Add to /etc/fstab; on Debian/Ubuntu run update-initramfs -u | | No space left on device inside VG | VG is full | vgs to confirm; vgextend myvg /dev/sdY to add disk, or lvreduce another LV | | Thin pool showing 100% data usage | Over-provisioned thin volumes wrote more than pool holds | lvextend -L +20G myvg/mypool; enable thin_pool_autoextend in lvm.conf | | Snapshot LV filling up | COW space exhausted by writes to origin since snapshot was taken | Merge or remove the snapshot immediately; next time use a larger -L for the snap | | pvmove hangs or times out | Background kernel move stalled | Check lvs --all for [pvmove0] progress; pvmove --abort to cancel safely | | VG name collision after disk copy | Two VGs with the same UUID/name imported simultaneously | vgrename <uuid> newname to disambiguate before importing | | device not found after UUID change | LVM cached old device UUIDs | pvscan --cache; check filter in lvm.conf is not excluding the device | | LVM cache (dm-cache) misconfiguration | Cache LV and origin LV roles swapped, or chunk size mismatch | lvconvert --splitcache myvg/myoriginlv to detach; reconfigure with correct --cachemode and --chunksize | | Can't open /dev/sdX exclusively | Another process holds the device | fuser -m /dev/sdX; unmount or stop the consumer before LVM operations |

Pain Points

  • Thin provisioning can silently over-commit: you can allocate 10TB of thin volumes on a 1TB pool. The pool fills when writes arrive, not when the LV is created. Monitor data percentage continuously — a full thin pool causes I/O errors on all thin volumes simultaneously.
  • Snapshots are copy-on-write and not free: every write to the origin after snapshot creation doubles the I/O (original write + COW copy). Old snapshots on busy volumes fill up fast; an exhausted snapshot is auto-deactivated and becomes unreadable.
  • pvmove is slow and disruptive: moving data across spindles at full speed can saturate I/O. There is no built-in throttle. If interrupted (power loss, pvmove --abort), LVM resumes on next boot but the mirror state must be clean.
  • LV extension requires two commands: lvextend grows the block device; the filesystem (ext4: resize2fs, XFS: xfs_growfs, Btrfs: btrfs filesystem resize) must be grown separately. Forgetting the second command leaves the filesystem at the old size with no error.
  • VG metadata area limits on large PV counts: the default metadata area (1 MiB) caps the number of PEs the VG can describe. With many small PVs or very large disks with a small PE size, the metadata area fills. Increase with pvresize and pvchange --metadatacopies.
  • LVM cache (dm-cache, dm-writecache) is unrelated to ZFS ARC/L2ARC: LVM cache is a block-level cache layer sitting below the filesystem. Combining it with ZFS ARC/L2ARC on the same volume creates redundant caching layers and can cause cache coherency surprises. Use one caching layer per storage stack.

References

See references/ for:

  • lvm-patterns.md — task-oriented command sequences for common workflows
  • lvm.conf.annotated — key directives in /etc/lvm/lvm.conf explained
  • docs.md — man pages and official documentation links