Signs of filesystem problems

• Errors such as Read-only file system when you save files: the kernel switched the disk to read-only to protect it.
• Messages about ext4 or XFS errors in dmesg or the system journal.
• Files or folders that cannot be opened, or show question marks in ls -l.
• The server stops at boot and asks you to run fsck manually.

dmesg -T | grep -iE "ext4|xfs|i/o error" | tail -n 20
journalctl -k -p err -b --no-pager | tail -n 20

Step 1: find the device and filesystem type

lsblk -f          # devices, filesystem types and mount points
df -hT            # mounted filesystems with their types

Note the device for the problem partition, for example /dev/sda3, /dev/vda1 or an LVM volume such as /dev/mapper/almalinux-home, and its type.

Step 2: back up first

A repair can delete damaged files or move them to lost+found. Copy important data off the server, or take a snapshot, before you continue. See our backups guide.

Step 3: check and repair a data partition

For a partition other than /, such as /home or an extra disk, stop the services that use it, then unmount it:

# Find what is using the partition
lsof +f -- /home | head

# Unmount it
umount /home

ext4

fsck -n /dev/sda3     # check only, no changes
fsck -y /dev/sda3     # repair, answering yes to every fix

XFS

xfs_repair -n /dev/sda3     # check only, no changes
xfs_repair /dev/sda3        # repair

If xfs_repair says the log must be replayed, mount and unmount the filesystem once, then run it again. Use the -L option only as a last resort, because it discards the log and can lose recent changes.

Mount the partition again with mount /home and check your data.

Step 4: the root filesystem

You cannot unmount / while the server runs from it. You have two options:

• Rescue mode: boot the server into your provider’s rescue system, then run the check on the root device, which is not mounted there. This works for both ext4 and XFS.
• Check at next boot (ext4): add fsck.mode=force fsck.repair=yes to the kernel command line for one boot, then remove it again.

# AlmaLinux and Rocky Linux: force a check at the next boot, then reboot
grubby --update-kernel=ALL --args="fsck.mode=force fsck.repair=yes"
reboot

# After the server is back, remove the option again
grubby --update-kernel=ALL --remove-args="fsck.mode=force fsck.repair=yes"

Have console access ready in case the check needs your input. XFS is checked by xfs_repair from rescue mode, not at boot.

Read the fsck result

Check the code with echo $? right after fsck. Files fsck could not place are moved to lost+found on that filesystem.

When the disk itself is failing

Repeated errors can mean the disk is failing, not just the filesystem. On a dedicated server, check the disk health:

smartctl -H /dev/sda
smartctl -a /dev/sda | grep -iE "reallocated|pending|uncorrectable"

If the disk reports problems, copy your data off at once and contact support to replace it. On a VPS, the disk is virtual, so report repeated I/O errors to support.