Disks & mounting

A server is running out of space, so your boss plugs in a brand-new 10 GB disk. Linux sees the disk, but you can't store anything on it yet. You have to partition it, format it and mount it. By the end of this lesson you'll have done all three, and made the disk survive a reboot.

Disks are files too

On Linux, hardware shows up as special files in /dev. The first disk is /dev/sda, the second is /dev/sdb, and so on. A disk is divided into partitions, which get numbers: /dev/sdb1 is the first partition on the second disk. (Fast NVMe drives use names like /dev/nvme0n1p1, and cloud servers often use /dev/vda. Same idea.)

lsblk              # tree of disks and partitions, and where they're mounted
sudo fdisk -l      # details of every disk
sudo dmesg | tail  # kernel messages: did Linux notice the new disk?
df -hT             # how full each mounted filesystem is, and its type

NAME          MAJ:MIN RM  SIZE RO TYPE MOUNTPOINTS
sda             8:0    0   20G  0 disk
├─sda1          8:1    0    1G  0 part /boot
└─sda2          8:2    0   19G  0 part
  ├─rl-root   253:0    0   17G  0 lvm  /
  └─rl-swap   253:1    0    2G  0 lvm  [SWAP]
sdb             8:16   0   10G  0 disk          ← the new disk: no partitions, not mounted

dmesg shows the kernel's own messages, like hardware being plugged in. On Ubuntu it needs sudo, and on Rocky it works without it, although sudo never hurts.

Step 1: partition with fdisk

fdisk is interactive: it asks questions and nothing is saved until you press w (write). Press q to quit without saving if you get confused.

sudo fdisk /dev/sdb

Then answer its questions like this:

Command (m for help): n          # n = new partition
Select (default p): p            # primary
Partition number (1-4, default 1): # Enter = accept the default
First sector …:                  # Enter
Last sector …:                   # Enter = use the whole disk
Command (m for help): p          # print, to check before saving
Command (m for help): w          # write and exit

Step 2: format with mkfs

A filesystem is the way files are organized on a partition. This is where the families differ:

sudo mkfs.xfs /dev/sdb1

sudo mkfs.ext4 /dev/sdb1

Both families can use both filesystems. Just install the tools (xfsprogs on Ubuntu). Sticking to the default is a good habit, because it's what the next admin will expect.

Step 3: mount it

Windows gives a new disk a letter like E:. Linux has only one tree, so you mount the disk onto an empty folder, called the mount point. After that, whatever you save in that folder lands on the new disk.

sudo mkdir /data
sudo mount /dev/sdb1 /data
df -hT /data           # check it: size, type, where
sudo umount /data      # unmount (note the spelling: umount, no n!)

Step 4: survive a reboot with /etc/fstab

A plain mount only lasts until the next reboot. To mount a disk automatically at boot, add a line to /etc/fstab (the “file system table”). It has six columns:

UUID=e645634b-9477-…   /data   ext4   defaults   0 0
# what (disk ID)        where   type   options    dump fsck-order

Use the UUID, the disk's unique ID, rather than /dev/sdb1. Device names can change if you add another disk, but the UUID never does. Get it with blkid:

sudo blkid /dev/sdb1
echo "UUID=$(sudo blkid -s UUID -o value /dev/sdb1) /data xfs defaults 0 0" | sudo tee -a /etc/fstab
sudo systemctl daemon-reload
sudo mount -a

sudo blkid /dev/sdb1
echo "UUID=$(sudo blkid -s UUID -o value /dev/sdb1) /data ext4 defaults 0 0" | sudo tee -a /etc/fstab
sudo systemctl daemon-reload
sudo mount -a

$( ) runs the command inside it and pastes its output right there, so you don't have to copy the long UUID by hand. You could also open /etc/fstab in nano and type the line yourself.

systemd turns fstab lines into “mount units” behind the scenes, which is why mount reminds you to run systemctl daemon-reload after editing it (Linux Basics, lesson 15).

Try it: add a new disk 💽

This server has a brand-new, empty 10 GB disk. Make it usable at /data, permanently.

Quick check