XFS Explained

XFS Explained

XFS is a high-performance 64-bit journaling file system originally developed by SGI in 1993. It is the default file system in RHEL/CentOS and is increasingly used on Ubuntu for data volumes, database storage, and log aggregation. XFS excels at large file handling and high-throughput sequential writes. Understanding when to choose XFS over ext4 — and what its unique properties mean in practice — helps you make better infrastructure decisions.

Why choose XFS

Choose XFS for: database storage (PostgreSQL, MySQL data directories), large log volumes, streaming media storage, or when building storage stacks compatible with RHEL/CentOS systems.

Creating and mounting XFS

# Install XFS tools
sudo apt install -y xfsprogs

# Create an XFS file system
sudo mkfs.xfs /dev/sdb1

# Create with a label
sudo mkfs.xfs -L data /dev/sdb1

# Create with custom block size (for databases, 4096 matches OS page size)
sudo mkfs.xfs -b size=4096 /dev/sdb1

# Mount XFS
sudo mkdir -p /data
sudo mount -t xfs /dev/sdb1 /data

# Mount with performance options for databases
sudo mount -t xfs -o noatime,nobarrier /dev/sdb1 /data
# nobarrier: disable write barriers (faster, safe only with UPS/battery-backed cache)

# Add to /etc/fstab
UUID=$(sudo blkid -s UUID -o value /dev/sdb1)
echo "UUID=$UUID  /data  xfs  defaults,noatime  0  2" | sudo tee -a /etc/fstab

XFS-specific tools

# Show XFS file system info
sudo xfs_info /data
sudo xfs_info /dev/sdb1

# Check an XFS file system (must be unmounted or mounted read-only)
sudo umount /data
sudo xfs_check /dev/sdb1    # Basic check
sudo xfs_repair -n /dev/sdb1  # Dry-run repair check

# Repair an XFS file system
sudo xfs_repair /dev/sdb1

# Dump XFS file system metadata (for backup/forensics)
sudo xfs_metadump /dev/sdb1 /tmp/sdb1.metadump

# Grow an XFS file system (XFS can grow online, on a mounted volume)
sudo xfs_growfs /data    # Grows to fill the underlying device/LV

# Freeze/thaw XFS for consistent snapshots
sudo xfs_freeze -f /data    # Freeze (blocks writes for snapshot)
# --- take snapshot here ---
sudo xfs_freeze -u /data    # Thaw

⚠️ WARNING: XFS cannot be shrunk once created. If you format a 500 GB partition as XFS and later want to shrink it to 200 GB, you cannot — you must backup, reformat, and restore. Plan XFS volume sizes carefully, or use LVM so you can add space later. Growing is always possible (xfs_growfs); shrinking is not.

XFS performance tuning

# For SSD-backed XFS volumes, enable TRIM
# In fstab, add 'discard' option:
UUID=...  /data  xfs  defaults,noatime,discard  0  2

# Or run periodic fstrim instead (less write amplification):
sudo fstrim -v /data

# For databases: align to storage block size and disable access time
UUID=...  /data  xfs  defaults,noatime,nodiratime  0  2

# For Elasticsearch/log storage: improve throughput with larger block size
# Use -d agcount=N when creating to set allocation group count
sudo mkfs.xfs -d agcount=8 /dev/sdb1    # More AGs = better parallelism

XFS limitations

• Cannot shrink: Once formatted, XFS cannot be reduced in size
• Small file performance: Slightly worse than ext4 for workloads with millions of tiny files
• Metadata-only journaling: Data writes are not journaled — there is a small window during which data written but not synced can be lost in a crash (same as ext4 in ordered mode)
• Not default in Ubuntu: ext4 is the Ubuntu default; XFS tooling must be installed separately

Conclusion

XFS is the right choice for Ubuntu data volumes that serve databases, large log files, or media storage — especially if you are also managing RHEL/CentOS systems and want consistent behavior. It cannot be shrunk, so plan volume sizes with room to grow. Always install xfsprogs before creating XFS volumes. Use xfs_growfs for online expansion (no downtime required) and schedule periodic fstrim for SSD-backed volumes.