Skip to content

Disks and Devices

The kernel presents every disk, partition, loop file and logical volume as a block device under /dev. Identifying the right device before partitioning or formatting it is the step that prevents data loss.

Track: Core · Interview weight: Med

Must-Know Facts

Fact Value Verify with
List block devices lsblk; lsblk -f adds filesystem, label, UUID and usage lsblk -o NAME,SIZE,TYPE,MOUNTPOINTS
Filesystem signatures blkid (as root for uncached devices) sudo blkid
SATA, SAS, USB disks /dev/sda, /dev/sdb; partitions sda1 lsblk
Virtio disks (KVM) /dev/vda; partitions vda1 lsblk -o NAME,TRAN
NVMe /dev/nvme0n1 (controller 0, namespace 1); partitions nvme0n1p1 nvme list
Xen disks (older EC2) /dev/xvda lsblk
Other block devices loopN (file), dm-N (LVM, LUKS), mdN (software RAID), sr0 (optical) lsblk -o NAME,TYPE
Stable names /dev/disk/by-uuid, by-id, by-path, by-label, by-partuuid ls -l /dev/disk/by-uuid
Rotational flag 1 spinning disk (and most virtual disks), 0 SSD cat /sys/block/<dev>/queue/rotational
Size in sectors /sys/block/<dev>/size, always in 512-byte units cat /sys/block/vda/size
MBR 4 primary partitions (or 3 plus an extended one), 2 TiB limit fdisk -l
GPT 128 partitions by default, backup table at the end of the disk, needed above 2 TiB and for UEFI gdisk -l
Disk health smartctl -H (SATA, SAS), nvme smart-log (NVMe); physical or passthrough disks only sudo smartctl -i /dev/sda
Loop device Presents a file as a disk: losetup -fP --show <file> losetup -l

Listing Block Devices

lsblk -o NAME,MAJ:MIN,SIZE,TYPE,ROTA,TRAN,MOUNTPOINTS
lsblk -f
sudo blkid
cat /sys/block/vda/queue/rotational /sys/block/vda/size /sys/block/vda/queue/logical_block_size

Output:

NAME MAJ:MIN SIZE TYPE ROTA TRAN   MOUNTPOINTS
vda  253:0    80G disk    1 virtio /
NAME FSTYPE FSVER LABEL UUID                                 FSAVAIL FSUSE% MOUNTPOINTS
vda  ext4   1.0         0a80853f-1c24-4c13-8293-19cd53de6bbe   72.1G     3% /
/dev/vda: UUID="0a80853f-1c24-4c13-8293-19cd53de6bbe" BLOCK_SIZE="4096" TYPE="ext4"
1
167772192
512

lsblk -o NAME,MAJ:MIN,SIZE,TYPE,ROTA,TRAN,MOUNTPOINTS
lsblk -f

Output:

NAME  MAJ:MIN   SIZE TYPE ROTA TRAN   MOUNTPOINTS
loop0   7:0   105.2M loop    1        /snap/core/17292
loop1   7:1    50.3M loop    1        /snap/snapd/27738
vda   253:0      80G disk    1 virtio /
NAME  FSTYPE   FSVER LABEL UUID                                 FSAVAIL FSUSE% MOUNTPOINTS
loop0 squashfs 4.0                                                    0   100% /snap/core/17292
loop1 squashfs 4.0                                                    0   100% /snap/snapd/27738
vda   ext4     1.0         0dc9cc86-daa7-4fcf-825d-21da512ae1f8   71.6G     4% /

The loop devices are snap packages, each a read-only squashfs image; they are always 100% full by design.

Both playgrounds format the whole Virtio disk as ext4 without a partition table, which is common for microVMs and rare on servers. The disk holds 167772192 sectors of 512 bytes, which is 80 GiB.

A virtual disk reports ROTA 1 even on SSD storage

The Virtio driver does not know what backs the disk, so ROTA says rotational. Cloud providers publish the storage type in the volume settings instead.

Stable Names

ls -l /dev/disk/by-uuid /dev/disk/by-path

Output:

/dev/disk/by-path:
total 0
lrwxrwxrwx 1 root root 9 Sep 17 07:25 pci-0000:00:01.0 -> ../../vda
lrwxrwxrwx 1 root root 9 Sep 17 07:25 virtio-pci-0000:00:01.0 -> ../../vda

/dev/disk/by-uuid:
total 0
lrwxrwxrwx 1 root root 9 Sep 17 07:25 0a80853f-1c24-4c13-8293-19cd53de6bbe -> ../../vda

Kernel names follow detection order, so sdb can become sdc after a reboot or a hot-plug. /etc/fstab and scripts use the UUID of the filesystem, or the by-id name of the disk, which contains the model and serial number.

EC2 Nitro instances rename attached volumes

A volume attached as /dev/sdf in the AWS console appears as an NVMe device such as /dev/nvme1n1 inside the instance. The volume ID is in the by-id link (nvme-Amazon_Elastic_Block_Store_vol...) and in the nvme id-ctrl serial number.

Disk Health

smartctl reads SMART data from SATA and SAS disks, and nvme from NVMe drives. A virtual disk has no SMART data:

sudo smartctl -i /dev/vda; echo "rc=$?"
sudo nvme list; echo "rc=$?"

Output:

smartctl 7.4 2023-08-01 r5530 [x86_64-linux-6.1.167] (local build)
Copyright (C) 2002-23, Bruce Allen, Christian Franke, www.smartmontools.org

/dev/vda: Unable to detect device type
Please specify device type with the -d option.

Use smartctl -h to get a usage summary

rc=1
Failed to scan topology: No such file or directory
rc=1

On a physical server, sudo smartctl -H /dev/sda prints the overall health result, and sudo smartctl -A /dev/sda lists attributes such as Reallocated_Sector_Ct and Current_Pending_Sector, whose growth predicts failure. sudo nvme smart-log /dev/nvme0 reports percentage_used and media_errors. smartd from the same package watches disks and logs changes.

Test Disks with Loop Devices

A loop device turns a file into a block device, so partitioning, LVM and RAID can be practised without spare disks. The rest of this module uses four 2 GiB sparse files:

mkdir -p /var/tmp/disks && cd /var/tmp/disks
truncate -s 2G disk1.img disk2.img disk3.img disk4.img
ls -ls
for f in disk?.img; do sudo losetup -fP --show "$f"; done
losetup -l
lsblk -o NAME,MAJ:MIN,SIZE,TYPE

Output:

total 0
0 -rw-r--r-- 1 laborant laborant 2147483648 Sep 17 07:28 disk1.img
0 -rw-r--r-- 1 laborant laborant 2147483648 Sep 17 07:28 disk2.img
0 -rw-r--r-- 1 laborant laborant 2147483648 Sep 17 07:28 disk3.img
0 -rw-r--r-- 1 laborant laborant 2147483648 Sep 17 07:28 disk4.img
/dev/loop0
/dev/loop1
/dev/loop2
/dev/loop3
NAME       SIZELIMIT OFFSET AUTOCLEAR RO BACK-FILE                DIO LOG-SEC
/dev/loop1         0      0         0  0 /var/tmp/disks/disk2.img   0     512
/dev/loop2         0      0         0  0 /var/tmp/disks/disk3.img   0     512
/dev/loop0         0      0         0  0 /var/tmp/disks/disk1.img   0     512
/dev/loop3         0      0         0  0 /var/tmp/disks/disk4.img   0     512
NAME  MAJ:MIN SIZE TYPE
loop0   7:0     2G loop
loop1   7:1     2G loop
loop2   7:2     2G loop
loop3   7:3     2G loop
vda   253:0    80G disk

The files use no blocks until data is written (0 in the first column of ls -s). -f takes the first free loop device, -P makes the kernel scan partitions, and --show prints the device name. sudo losetup -d /dev/loop0 detaches one device; attachments do not survive a reboot.

Common Errors

/dev/vda: Unable to detect device type

Cause: the device is virtual or behind a controller that smartctl cannot identify.

Fix: on hardware RAID, pass the controller type (-d megaraid,N); on a VM, check disk health on the host or in the cloud console.

losetup: cannot find an unused loop device

Cause: every loop device is in use and no new one could be created (common inside containers without /dev/loop-control).

Fix: losetup -l to find stale attachments and sudo losetup -d them; in a container, run the task on the host.

Interview Checkpoints

L1: What is the difference between /dev/sda, /dev/vda and /dev/nvme0n1?

Say first: they are disks on different drivers: SCSI layer (SATA, SAS, USB), Virtio in a KVM guest, and NVMe controller 0, namespace 1.

Proof: lsblk -o NAME,TRAN shows sata, virtio or nvme.

Follow-up: How are partitions on an NVMe disk named? (nvme0n1p1, because the name ends in a digit.)

L1: When must a disk use GPT instead of MBR?

Say first: above 2 TiB, with more than four primary partitions, and when the machine boots in UEFI mode.

Proof: sudo fdisk -l /dev/sda shows Disklabel type: gpt or dos.

Follow-up: Where does GPT keep its backup table? (In the last sectors of the disk.)

L1: Why does /etc/fstab use UUIDs instead of device names?

Say first: device names follow detection order and can change, while the filesystem UUID stays with the data.

Proof: ls -l /dev/disk/by-uuid; lsblk -f.

Follow-up: What changes the UUID? (Recreating the filesystem, or tune2fs -U and xfs_admin -U.)

L2: Find the new disk that was attached to a running server.

Say first: list block devices and pick the one with no partitions, no filesystem and no mount point, matching the expected size.

Proof: lsblk -f; sudo dmesg | tail shows the new device; ls -l /dev/disk/by-id.

Follow-up: How do you make it visible without a reboot on a SCSI bus? (echo "- - -" | sudo tee /sys/class/scsi_host/host*/scan.)

L2: Create a practice disk without a spare device.

Say first: create a sparse file and attach it as a loop device with partition scanning.

Proof: truncate -s 2G d.img; sudo losetup -fP --show d.img.

Follow-up: Why does ls -ls show 0 blocks for the file?

L2: Check whether a physical disk is failing.

Say first: read its SMART health and the reallocated and pending sector counts, and check the kernel log for I/O errors.

Proof: sudo smartctl -H -A /dev/sda; sudo nvme smart-log /dev/nvme0; journalctl -k | grep -i 'i/o error'.

Follow-up: Which RAID layer hides SMART data, and how do you reach it?

L3: A volume was attached to an EC2 instance as /dev/sdf, but the device does not exist. What happened?

Say first: on Nitro instances, EBS volumes appear as NVMe devices, so the disk is /dev/nvme1n1 or similar.

Proof: lsblk -o NAME,SIZE,SERIAL; ls -l /dev/disk/by-id/ | grep Elastic.

Follow-up: Why must /etc/fstab on such an instance use UUIDs?

Captured on Rocky Linux 10.2 and Ubuntu 24.04.4 LTS (iximiuz Labs microVMs, kernel 6.1.167), 2026-09.