All technical articles

Edit fstab Safely: UUIDs, nofail and Testing With mount -a

Write fstab entries with UUIDs and nofail, test with mount -a and findmnt --verify before rebooting, and avoid a stuck boot.

Edit fstab Safely: UUIDs, nofail and Testing With mount -a

One wrong line in /etc/fstab can drop a server into emergency mode, exactly when you have no console. This guide gives a routine that tests every change before the reboot.

1. Back it up first

sudo cp -a /etc/fstab /etc/fstab.bak

That one line is your quick way back. Without it, a broken edit must be rebuilt from memory.

2. Use UUIDs, not device names

lsblk -f
sudo blkid

Names like /dev/sdb1 can change when disk order changes, while the UUID stays the same. A sample line:

UUID=xxxx-xxxx  /data  ext4  defaults,nofail  0  2

Replace xxxx-xxxx with the value blkid printed.

3. nofail for non-critical disks

Without nofail, a missing data disk makes boot wait and then drop to emergency mode. For disks other than the root filesystem add nofail, and optionally x-systemd.device-timeout=10s so the wait stays short. Never put nofail on the root filesystem or /boot.

4. Test before you reboot

sudo findmnt --verify
sudo mount -a
sudo systemctl daemon-reload

The first command reports structural errors; the second really mounts every entry. If neither complains, a reboot is much safer.

5. If boot is stuck

In emergency mode or from your provider’s console, remount root writable (mount -o remount,rw /), restore the backup file and reboot. For services that depend on a mount, see writing a systemd unit, and use a safe rsync copy before moving data.

6. Illustrative example, not a measurement

Suppose a second disk was mounted at /data by UUID, and after a disk swap the server sat in emergency mode because nofail was missing. With nofail the same server boots and only /data is empty. This scenario is an invented illustration, not a measurement.

7. Common mistakes

  • Using /dev/sdX for non-root disks.
  • Rebooting without running mount -a.
  • Putting nofail on root or /boot.
  • Skipping the fstab backup.

8. Frequently asked questions

What is the last number on an fstab line? The fsck order: 1 for root, usually 2 for others, and 0 for filesystems that skip fsck.

Is nofail safe? It keeps boot moving, but an application that depends on the path may write to the root disk instead; check for that.

Does mount -a remount everything? It mounts only the entries that are not mounted yet.

9. Quick checklist

  1. fstab was backed up.
  2. UUID read with blkid.
  3. nofail set on non-critical disks.
  4. findmnt –verify and mount -a ran clean.
  5. A console recovery path is ready.

If you need help reviewing performance or maintaining a server, see the server management services.

More server and WordPress guides are collected in the article archive.

Official references

fstab(5) man page · findmnt(8) man page · mount(8) man page