Managing Storage Pools and Snapshots with ZFS and Btrfs

Traditional filesystems like ext4 or XFS rely on separate volume managers (LVM) or hardware RAID controllers to combine drives. ZFS and Btrfs combine filesystem and storage-management features, including checksums and snapshots. A snapshot is useful for recovery but is not a backup; keep a separate tested copy of important data.

---

ZFS vs Btrfs: Which Should You Choose?

Feature	OpenZFS	Btrfs
Kernel Status	Out-of-tree kernel module; support and installation vary by distribution	Included in the upstream Linux kernel
Memory Planning	ARC uses available memory and can be tuned; ECC is recommended where supported but is not a strict ZFS requirement	Plan memory for the workload and filesystem cache
Redundancy Profiles	mirror, raidz1, or raidz2; choose based on failure and capacity requirements	RAID1/RAID10 are common choices; Btrfs RAID5/6 remains unsuitable for production data
Best For	Dedicated storage systems where its feature set is appropriate	Linux desktops, servers, and systems using its native filesystem tools

---

Step 1: Install Utilities and Identify Disks

01

Install Storage Utilities and Identify Disks

Preparation

These package commands use Debian/Ubuntu; package names and ZFS kernel-module support vary by distribution. Identify disks using persistent IDs and verify their model, serial, size, and existing signatures before continuing.

sudo apt update
sudo apt install zfsutils-linux btrfs-progs
lsblk -o NAME,SIZE,TYPE,FSTYPE,MOUNTPOINTS,SERIAL
ls -l /dev/disk/by-id/

❯ View Expected Console Output

ata-WDC_WD40EFRX_DISK1 -> ../../sdb
ata-WDC_WD40EFRX_DISK2 -> ../../sdc

Formatting these disks erases data

The commands in Step 2 create new storage on both selected devices and destroy existing filesystem contents on them. Verify the device IDs against your provisioning record and confirm you have a separate backup before proceeding. Do not run these commands against disks containing data you need.

---

Step 2: Create a Mirrored Pool or Filesystem

02

Create a Two-Disk Mirror

Pool Creation

Choose one tab and use the persistent disk IDs verified in Step 1. The ZFS command omits force options so ZFS can stop if it detects an existing signature or another condition that needs investigation. Btrfs formatting is destructive even without a force flag.

OpenZFS mirror

sudo zpool create -o ashift=12 tank mirror \
  /dev/disk/by-id/ata-WDC_WD40EFRX_DISK1 \
  /dev/disk/by-id/ata-WDC_WD40EFRX_DISK2
sudo zpool status tank

Btrfs RAID1

sudo mkfs.btrfs -L data -m raid1 -d raid1 \
  /dev/disk/by-id/ata-WDC_WD40EFRX_DISK1 \
  /dev/disk/by-id/ata-WDC_WD40EFRX_DISK2
sudo mkdir -p /data
sudo mount -o compress=zstd:1,noatime LABEL=data /data
findmnt /data

❯ View Expected Console Output

ZFS: pool tank reports ONLINE with a mirror-0 vdev.
Btrfs: findmnt /data reports the new Btrfs filesystem mounted at /data.

For Btrfs to mount after reboot, obtain its UUID with sudo blkid -s UUID -o value /dev/disk/by-id/ata-WDC_WD40EFRX_DISK1, add a corresponding UUID-based entry to /etc/fstab, then validate the file with sudo mount -a and findmnt /data before relying on it.

# Replace the value with the UUID returned by blkid.
UUID=replace-with-the-filesystem-UUID /data btrfs defaults,compress=zstd:1,noatime 0 0

---

Step 3: Organize Data into Datasets or Subvolumes

03

Separate Application Data from the Pool Root

Organization

Use separate ZFS datasets or Btrfs subvolumes for application data and snapshot destinations. Btrfs compression is enabled by the mount option from Step 2; ZFS compression is configured per dataset.

OpenZFS

sudo zfs create -o compression=lz4 tank/appdata
sudo zfs create -o compression=lz4 tank/backups
sudo zfs list

Btrfs

sudo btrfs subvolume create /data/appdata
sudo btrfs subvolume create /data/snapshots
sudo btrfs subvolume list /data

❯ View Expected Console Output

ZFS datasets are listed below the pool, such as tank/appdata.
Btrfs subvolumes include appdata and snapshots.

---

Step 4: Create and Inspect Snapshots

04

Create a Snapshot Before a Planned Change

Recovery Point

These commands snapshot only the application-data dataset or subvolume, not the operating system root. Use the snapshot to inspect or recover files from that data. A full rollback is a separate, potentially destructive recovery action; plan it against the filesystem documentation and a verified backup before proceeding.

OpenZFS

sudo zfs snapshot tank/appdata@pre-change
sudo zfs list -t snapshot
sudo zfs diff tank/appdata@pre-change
sudo zfs set snapdir=visible tank/appdata
sudo ls -la /tank/appdata/.zfs/snapshot/pre-change/
sudo mkdir -p /root/recovery
# Copy a reviewed file to a recovery location; inspect before restoring it.
sudo cp -a /tank/appdata/.zfs/snapshot/pre-change/path/to/file /root/recovery/

Btrfs

sudo btrfs subvolume snapshot -r \
  /data/appdata /data/snapshots/appdata-pre-change
sudo btrfs subvolume list /data
sudo ls -la /data/snapshots/appdata-pre-change/
sudo mkdir -p /root/recovery
# Copy a reviewed file to a recovery location; inspect before restoring it.
sudo cp -a /data/snapshots/appdata-pre-change/path/to/file /root/recovery/

❯ View Expected Console Output

ZFS: tank/appdata@pre-change
Btrfs: read-only subvolume snapshots/appdata-pre-change

---

Step 5: Scrub and Review Filesystem Health

05

Verify Checksums and Redundancy

Integrity Check

A scrub checks data against filesystem checksums. ZFS or Btrfs can repair a damaged block only when the filesystem has a valid redundant copy; checksums without redundancy can detect some damage but cannot recreate lost data. Scrubs use disk I/O, so schedule them according to workload and maintenance policy.

OpenZFS

sudo zpool scrub tank
sudo zpool status tank

Btrfs

sudo btrfs scrub start /data
sudo btrfs scrub status /data

❯ View Expected Console Output

Review the scrub summary for repaired, uncorrectable, or otherwise reported errors. A healthy result does not replace a separate backup.

Snapshots do not replace backups

Snapshots are stored in the same pool or filesystem as the data. They do not protect against loss of the pool, theft, ransomware with administrative access, or a mistaken destructive command. Maintain a separate backup and test restoring it.