Stage 3 · Build

Storage & Filesystems

ext4, XFS, and Modern Filesystems

Journaling, delayed allocation, metadata checksums, and choosing between ext4 and XFS.

7 min readLinux Deep DivesBuild

ext4 Overview

ext4 is the most widely used Linux filesystem. It supports volumes up to 1 EB, files up to 16 TB, and has been the default for most distributions for over a decade. ext4 is mature, well-tested, and optimized for general-purpose workloads.

ext4 filesystem information


# View filesystem type and mount options
mount | grep " / "
 # /dev/sda2 on / type ext4 (rw,relatime,seclabel)

 # Detailed filesystem stats
tune2fs -l /dev/sda1 | head -20
 # Filesystem volume name:   <none>
 # Filesystem magic number:  0xEF53
 # Filesystem features:      has_journal extents huge_file dir_nlink extra_isize
 # Block size:               4096
 # Blocks per group:         32768
 # Inodes per group:         8192

 # Check filesystem usage
df -hT
 # Filesystem     Type  Size  Used  Avail Use% Mounted on
 # /dev/sda2      ext4  499G  45G   430G  10% /

ext4 uses extents (contiguous block ranges) instead of indirect block mapping, significantly improving performance for large files.

XFS Overview

XFS is a high-performance 64-bit journaling filesystem designed for parallel I/O. It excels with large files and large directories. XFS is the default for RHEL/CentOS and is used by many databases and big data platforms.

XFS filesystem information

  # View XFS filesystem info
  xfs_info /dev/sda1
  # meta-data=/dev/sda1  isize=512  agcount=32, agsize=32768 blks
  #          =           sectsz=512  attr=2, projid32bit=1
  # data     =           bsize=4096  blocks=1048576, imaxpct=25
  #          =           unit=1      blocks=1048576
  # naming   =version 2  bsize=4096  ascii-ci=0 ftype=1
  # log      =internal   bsize=4096  blocks=1024, version=2
  #          =           sectsz=512  sunit=0 blks, lazy-count=1

  # XFS repair tool
  xfs_repair -n /dev/sda1  # Dry run

XFS allocates space in allocation groups (AGs). Multiple AGs allow parallel I/O across different parts of the filesystem, making XFS excellent for concurrent workloads.

Journaling

Journaling writes metadata changes to a log before applying them to the filesystem. If the system crashes, the journal replays uncommitted changes, ensuring consistency. ext4 and XFS both journal metadata by default.

Journal configuration

  # ext4 journal info
  tune2fs -l /dev/sda1 | grep journal
  # Journal inode:            8
  # Journal backup:          blocks
  # Journal features:        journal_incompat_revoke

  # XFS log information
  xfs_info /dev/sda1 | grep log
  # log      =internal   bsize=4096  blocks=1024, version=2

  # Disable journaling (DANGEROUS — data loss risk)
  tune2fs -O ^has_journal /dev/sda1

  # External journal for write-heavy workloads
  mkfs.ext4 -J device=/dev/sdb1 /dev/sda1

External journals on a separate device can improve performance for write-heavy workloads by reducing contention on the main device.

Delayed Allocation

Delayed allocation (delalloc) defers block allocation until data is flushed to disk. This allows the filesystem to make better allocation decisions and reduces fragmentation. It is enabled by default on both ext4 and XFS.

Delayed allocation behavior

  # ext4 delalloc is default (mount option: delalloc)
  mount | grep " / "
  # ... defaults,delalloc ...

  # Disable delalloc for specific needs
  mount -o nodelalloc /dev/sda1 /mnt

  # Check allocation statistics
  cat /proc/fs/ext4/sda2/session_write_kbytes
  # 123456

  # XFS also uses delayed allocation by default
  # No mount option needed

Delayed allocation can cause space issues if the system crashes before flushing. Applications that need guaranteed space (databases) may disable it.

Metadata Checksums

Modern filesystems compute checksums on metadata blocks to detect corruption. ext4 uses metadata_csum feature; XFS has always had metadata checksums. Data checksums are separate and available via dm-integrity or filesystem features.

Enable and verify metadata checksums

  # Check ext4 checksum feature
  tune2fs -l /dev/sda1 | grep checksum
  # Filesystem features: has_journal extents huge_file ...

  # Enable data checksumming on ext4
  tune2fs -O metadata_csum /dev/sda1

  # XFS always has metadata checksums
  xfs_info /dev/sda1 | grep crc
  # sectsz=512  crc=1

  # Check for filesystem errors
  dmesg | grep -i "ext4|xfs|error|corrupt"

Metadata checksums protect against silent corruption. Always enable them on new filesystems. For data integrity, use dm-integrity or application-level checksums.

Choosing Between ext4 and XFS

Feature	ext4	XFS
Max volume size	1 EB	8 EB
Max file size	16 TB	8 EB
Parallel I/O	Limited	Excellent
Small file performance	Good	Slower
Large file performance	Good	Excellent
Quota support	User/group/project	User/group/project
Online defrag	e4defrag	xfs_fsr
Default distribution	Debian/Ubuntu	RHEL/CentOS

Use XFS for large files and databases

XFS handles large files and parallel I/O better than ext4. For databases, virtual machines, and container storage, XFS is often the better choice. ext4 excels for general-purpose desktop and small-file workloads.

Never use noatime for databases

The noatime mount option disables access time updates, reducing write I/O. Add it to /etc/fstab for any database or I/O-heavy workload. Example: /dev/sda1 / ext4 defaults,noatime 0 1

Mark this lesson complete to store local progress and unlock a cleaner resume path the next time you visit.