️ Snapshots — LVM/ZFS/Btrfs

If you like this project, consider supporting me on Buy Me a Coffee ☕️

tags:

• backups
• s3
• sysadmin

• linux

  Filesystem snapshots (LVM, ZFS, Btrfs) provide instant point-in-time copies of data at the block or extent level using Copy-on-Write (CoW) technology. They are used for consistent backups of running databases, quick rollbacks before risky changes, and disaster recovery. Snapshots are not a replacement for off-site backups — they protect against logical errors but not hardware failures. All three technologies are actively maintained and widely used in production. / Снапшоты файловых систем (LVM, ZFS, Btrfs) обеспечивают мгновенные копии данных на уровне блоков с использованием Copy-on-Write (CoW). Используются для консистентных бэкапов работающих баз данных, быстрых откатов и аварийного восстановления. Снапшоты не заменяют внешние резервные копии.

Technology Comparison

Choosing the right snapshot technology depends on your storage stack and OS.

LVM Snapshots

Create Snapshot

lvcreate -L 10G -s -n snap1 /dev/vg0/lv_root     # Create 10 GB snapshot / Создать снапшот 10 ГБ
lvcreate -L 5G  -s -n snap_www /dev/vg0/lv_www   # Snapshot specific LV / Снапшот конкретного LV

[!IMPORTANT] The snapshot size (-L) controls the CoW (Copy-on-Write) extent pool, not the data size. If the source LV changes more than the snapshot size, the snapshot becomes invalid. Size to at least 10–20% of the source LV's expected change rate.

List Snapshots

lvs                                                # List all LVs / Список всех LV
lvs | grep snapshot                               # Show only snapshots / Только снапшоты
lvdisplay /dev/vg0/snap1                          # Snapshot details / Детали снапшота
lvs -o lv_name,data_percent                       # Show usage % / Использование снапшота

Sample output of lvs | grep snapshot:

  snap1   vg0 swi-ao----  10.00g                                     lv_root 45.23
  snap_www vg0 swi-ao----   5.00g                                     lv_www   12.01

Mount Snapshot

mkdir -p /mnt/snap1                               # Create mount point / Создать точку монтирования
mount /dev/vg0/snap1 /mnt/snap1 -o ro            # Mount read-only / Монтировать только для чтения
ls /mnt/snap1                                     # Browse files / Просмотр файлов
umount /mnt/snap1                                 # Unmount / Размонтировать

Remove Snapshot

[!WARNING] Removing a snapshot is permanent and immediate. Ensure no backup jobs are reading from it first.

lvremove /dev/vg0/snap1                           # Remove snapshot (interactive) / Удалить снапшот (интерактивно)
lvremove -f /dev/vg0/snap1                        # Force remove (no confirmation) / Принудительно удалить

Merge Snapshot

lvconvert --merge /dev/vg0/snap1                  # Merge snapshot into origin / Слить снапшот с оригиналом
# If LV is active, merge completes on next activation (reboot)
#

[!CAUTION] Merging a snapshot replaces the current state of the origin LV with the snapshot state. This is a destructive rollback — all changes since the snapshot was taken will be lost.

Extend Snapshot

lvextend -L +5G /dev/vg0/snap1                   # Extend by 5 GB / Расширить на 5 ГБ

[!TIP] Monitor snapshot fill with lvs -o lv_name,data_percent. If it reaches 100%, the snapshot is invalidated automatically.

ZFS Snapshots

Create Snapshot

zfs snapshot pool/dataset@snap1                   # Create snapshot / Создать снапшот
zfs snapshot -r pool/dataset@snap1                # Recursive snapshot (all children) / Рекурсивный снапшот
zfs snapshot pool/dataset@$(date +%Y%m%d)        # Snapshot with date suffix / Снапшот с суффиксом даты

List Snapshots

zfs list -t snapshot                              # List all snapshots / Список всех снапшотов
zfs list -t snapshot pool/dataset                 # Dataset-specific snapshots / Снапшоты датасета
zfs list -t snapshot -o name,used,referenced      # Custom columns / Кастомные колонки
zfs list -t snapshot -s creation                  # Sort by creation time / Сортировка по времени

Sample output of zfs list -t snapshot -o name,used,referenced:

NAME                      USED   REFER
pool/data@20240101       1.23G   45.6G
pool/data@20240102        234M   46.1G

Rollback Snapshot

[!CAUTION] zfs rollback destroys all snapshots newer than the target and reverts the dataset to the snapshot state. Use -r to also remove child snapshots. This is irreversible.

zfs rollback pool/dataset@snap1                   # Rollback to snapshot / Откатить к снапшоту
zfs rollback -r pool/dataset@snap1                # Recursive rollback / Рекурсивный откат
zfs rollback -rf pool/dataset@snap1               # Force rollback (unmounts if needed) / Принудительный откат

Clone Snapshot

zfs clone pool/dataset@snap1 pool/clone1          # Clone snapshot to new dataset / Клонировать снапшот
zfs promote pool/clone1                           # Promote clone (reverse dependency) / Повысить клон
# After promote: clone becomes independent, original depends on it
#

Delete Snapshot

zfs destroy pool/dataset@snap1                    # Delete single snapshot / Удалить снапшот
zfs destroy -r pool/dataset@snap1                 # Recursive delete / Рекурсивное удаление
zfs destroy pool/dataset@snap1%snap3              # Delete range snap1 → snap3 / Удалить диапазон
zfs destroy -nv pool/dataset@snap1                # Dry run — preview what will be freed / Пробный запуск

Send/Receive — Replication

# Full send to remote host
zfs send pool/dataset@snap1 | ssh <USER>@<HOST> zfs receive backup/dataset

# Incremental send (much faster after initial)
zfs send -i pool/dataset@snap1 pool/dataset@snap2 | ssh <USER>@<HOST> zfs receive backup/dataset

# Recursive send (all children)
zfs send -R pool/dataset@snap1 | zfs receive backup/dataset

[!TIP] Add mbuffer between zfs send and SSH to buffer network fluctuations: zfs send pool/dataset@snap1 | mbuffer -s 128k -m 1G | ssh <USER>@<HOST> zfs receive backup/dataset

Snapshot Holds

zfs hold keep pool/dataset@snap1                  # Place hold (prevents destroy) / Удержание (запрет удаления)
zfs holds pool/dataset@snap1                      # List holds / Список удержаний
zfs release keep pool/dataset@snap1               # Release hold / Освободить удержание

ZFS Auto-Snapshot

apt install zfs-auto-snapshot                     # Install auto-snapshot / Установить авто-снапшот

zfs set com.sun:auto-snapshot=true pool/dataset              # Enable auto-snapshot / Включить
zfs set com.sun:auto-snapshot:frequent=false pool/dataset    # Disable 15-min snapshots / Отключить частые
zfs set com.sun:auto-snapshot:hourly=true pool/dataset       # Enable hourly / Включить почасовые
zfs set com.sun:auto-snapshot:daily=true pool/dataset        # Enable daily / Включить ежедневные
zfs set com.sun:auto-snapshot:weekly=true pool/dataset       # Enable weekly / Включить еженедельные
zfs set com.sun:auto-snapshot:monthly=true pool/dataset      # Enable monthly / Включить ежемесячные

Btrfs Snapshots

Create Snapshot

btrfs subvolume snapshot /mnt/data /mnt/data_snap1      # Read-write snapshot / Снапшот для чтения-записи
btrfs subvolume snapshot -r /mnt/data /mnt/data_snap1   # Read-only snapshot (required for send) / Только чтение (нужен для send)

[!IMPORTANT] Read-only snapshots (-r) are required when using btrfs send. Read-write snapshots can be accidentally modified, making incremental send impossible.

List Snapshots

btrfs subvolume list /mnt                         # List all subvolumes / Список всех подтомов
btrfs subvolume list -s /mnt                      # List only snapshots / Только снапшоты
btrfs subvolume show /mnt/data_snap1              # Snapshot details / Детали снапшота

Delete Snapshot

btrfs subvolume delete /mnt/data_snap1            # Delete snapshot / Удалить снапшот

Restore from Snapshot — Production Runbook

[!CAUTION] This procedure replaces the live subvolume. Ensure all services writing to /mnt/data are stopped first.

1. Stop services that use the subvolume:

   systemctl stop <SERVICE>

2. Rename the current (broken) subvolume:

   mv /mnt/data /mnt/data_old

3. Create a new snapshot from the restore point:

   btrfs subvolume snapshot /mnt/data_snap1 /mnt/data

4. Restart services:

   systemctl start <SERVICE>

5. Verify, then remove the old subvolume:

   btrfs subvolume delete /mnt/data_old

Send/Receive — Replication

# Full send to remote
btrfs send /mnt/data_snap1 | ssh <USER>@<HOST> btrfs receive /backup

# Incremental send (-p specifies parent snapshot)
btrfs send -p /mnt/data_snap1 /mnt/data_snap2 | ssh <USER>@<HOST> btrfs receive /backup

Space Usage

btrfs filesystem df /mnt                          # Filesystem usage by type / Использование файловой системы
btrfs qgroup show /mnt                            # Quota group usage (if enabled) / Использование квот групп

Snapshot Strategies

Retention Policy Reference

Snapshot-Based Backup Workflow (LVM)

# 1. Create snapshot
lvcreate -L 5G -s -n backup_snap /dev/vg0/lv_data

# 2. Mount read-only and archive
mount /dev/vg0/backup_snap /mnt/snap -o ro
tar -czf /backup/data-$(date +%Y%m%d).tar.gz -C /mnt/snap .
umount /mnt/snap

# 3. Remove snapshot
lvremove -f /dev/vg0/backup_snap

Sysadmin Patterns

Daily LVM Snapshot Script

/usr/local/bin/daily-lvm-snapshot.sh

#!/bin/bash
# Daily LVM snapshot with 7-day rolling retention
#

set -euo pipefail

VG="vg0"
LV="lv_data"
SNAP_NAME="snap_$(date +%Y%m%d)"
SNAP_SIZE="10G"
KEEP=7

# Create snapshot
lvcreate -L "$SNAP_SIZE" -s -n "$SNAP_NAME" "/dev/$VG/$LV"
echo "Created snapshot: /dev/$VG/$SNAP_NAME"

# Remove oldest snapshots beyond retention count
lvs --noheadings -o lv_name "$VG" | awk '{print $1}' | grep '^snap_' | sort | head -n "-$KEEP" | while read -r snap; do
  echo "Removing old snapshot: /dev/$VG/$snap"
  lvremove -f "/dev/$VG/$snap"
done

chmod +x /usr/local/bin/daily-lvm-snapshot.sh    # Make executable / Сделать исполняемым

Cron for LVM Snapshot / Cron

/etc/cron.d/lvm-snapshot

# Run LVM daily snapshot at 01:00
0 1 * * * root /usr/local/bin/daily-lvm-snapshot.sh >> /var/log/lvm-snapshot.log 2>&1

ZFS Snapshot + Remote Replication Script

/usr/local/bin/zfs-snapshot-replicate.sh

#!/bin/bash
# ZFS snapshot with incremental remote replication
#

set -euo pipefail

DATASET="pool/data"
REMOTE_HOST="<HOST>"
REMOTE_USER="<USER>"
REMOTE_DATASET="backup/data"
SNAP_NAME="daily_$(date +%Y%m%d)"
KEEP=30

# Create new snapshot
zfs snapshot "${DATASET}@${SNAP_NAME}"
echo "Created: ${DATASET}@${SNAP_NAME}"

# Get previous snapshot for incremental send
PREV_SNAP=$(zfs list -t snapshot -o name -s creation | grep "^${DATASET}@daily" | tail -n 2 | head -n 1 | cut -d@ -f2)

if [ -n "$PREV_SNAP" ]; then
  echo "Incremental send: @${PREV_SNAP} → @${SNAP_NAME}"
  zfs send -i "${DATASET}@${PREV_SNAP}" "${DATASET}@${SNAP_NAME}" \
    | ssh "${REMOTE_USER}@${REMOTE_HOST}" zfs receive "${REMOTE_DATASET}"
else
  echo "Full send: ${DATASET}@${SNAP_NAME}"
  zfs send "${DATASET}@${SNAP_NAME}" \
    | ssh "${REMOTE_USER}@${REMOTE_HOST}" zfs receive "${REMOTE_DATASET}"
fi

# Prune old snapshots beyond retention
zfs list -t snapshot -o name -s creation | grep "^${DATASET}@daily" | head -n "-${KEEP}" | while read -r snap; do
  echo "Destroying old snapshot: $snap"
  zfs destroy "$snap"
done

chmod +x /usr/local/bin/zfs-snapshot-replicate.sh

Cron for ZFS Replication / Cron

/etc/cron.d/zfs-snapshot

# ZFS daily snapshot + replication at 02:00
0 2 * * * root /usr/local/bin/zfs-snapshot-replicate.sh >> /var/log/zfs-snapshot.log 2>&1

Btrfs Automated Snapshot Script

/usr/local/bin/btrfs-snapshot.sh

#!/bin/bash
# Btrfs daily read-only snapshots with 30-day retention
#

set -euo pipefail

SUBVOL="/mnt/data"
SNAP_DIR="/mnt/snapshots"
SNAP_NAME="data_$(date +%Y%m%d_%H%M%S)"
KEEP_DAYS=30

# Ensure snapshot directory exists
mkdir -p "$SNAP_DIR"

# Create read-only snapshot
btrfs subvolume snapshot -r "$SUBVOL" "${SNAP_DIR}/${SNAP_NAME}"
echo "Created: ${SNAP_DIR}/${SNAP_NAME}"

# Remove snapshots older than retention period
find "$SNAP_DIR" -maxdepth 1 -type d -name "data_*" -mtime "+${KEEP_DAYS}" | while read -r snap; do
  echo "Deleting old snapshot: $snap"
  btrfs subvolume delete "$snap"
done

chmod +x /usr/local/bin/btrfs-snapshot.sh

Cron for Btrfs Snapshots / Cron

/etc/cron.d/btrfs-snapshot

# Btrfs daily snapshot at 01:30
30 1 * * * root /usr/local/bin/btrfs-snapshot.sh >> /var/log/btrfs-snapshot.log 2>&1

Logrotate for Snapshot Logs / Logrotate

/etc/logrotate.d/snapshots

/var/log/lvm-snapshot.log
/var/log/zfs-snapshot.log
/var/log/btrfs-snapshot.log {
    daily
    rotate 30
    compress
    delaycompress
    missingok
    notifempty
    create 640 root root
    dateext
}

Troubleshooting

LVM Issues

# Error: "Insufficient free space" / "Недостаточно
vgs                                               # Check VG free space / Проверить свободное место VG
lvextend -L +5G /dev/vg0/snap1                   # Extend snapshot / Расширить снапшот

# Snapshot fill percentage
lvs -o lv_name,data_percent                       # Check all LV usage / Проверить использование LV

# Emergency: snapshot is full (100%) — snapshot is now invalid
#
lvremove -f /dev/vg0/snap1                        # Remove invalid snapshot / Удалить недействительный снапшот

[!CAUTION] When an LVM snapshot fills to 100%, it becomes automatically invalidated — all data in it is lost. Monitor usage regularly and extend or remove promptly.

ZFS Issues

# Error: "Cannot destroy snapshot: dataset is busy"
#
zfs holds pool/dataset@snap1                      # Check active holds / Проверить удержания
zfs release keep pool/dataset@snap1               # Release hold / Освободить удержание

# Error: "Cannot receive incremental stream" (remote diverged)
#
zfs rollback pool/dataset@snap1                   # Rollback to common snapshot / Откатить к общему снапшоту
zfs destroy -nv pool/dataset@snap1                # Dry run before destroy / Пробный запуск перед удалением

Btrfs Issues

# Error: "Cannot delete subvolume" / "Невозможно
btrfs subvolume list /mnt                         # List subvolumes to find IDs / Список подтомов
umount /mnt/data_snap1                            # Unmount if mounted / Размонтировать если смонтирован
btrfs subvolume delete /mnt/data_snap1            # Retry delete / Повторить удаление

Space Management

# LVM — snapshot space usage / LVM —
lvs -o lv_name,data_percent                       # Usage in % / Использование в %

# ZFS — per-snapshot space usage / ZFS —
zfs list -o name,used,refer -t snapshot           # Space used / Занятое место
zfs destroy -nv pool/dataset@snap1                # Preview space freed / Просмотр освобождаемого места

# Btrfs — overall + quota groups / Btrfs —
btrfs filesystem df /mnt                          # Filesystem usage / Использование файловой системы
btrfs qgroup show /mnt                            # Quota group usage (if enabled) / Квота групп

Documentation

• LVM Snapshots: https://man7.org/linux/man-pages/man8/lvcreate.8.html
• Btrfs Snapshots: https://btrfs.readthedocs.io/en/latest/Snaphots.html