disks

Drives, partitions, space and health together.

Synopsis

disks [options] [-- smartctl options]

Description

disks shows every drive in the machine as a tree. Each drive has its partitions under it, and each partition shows its size, its filesystem, where it is mounted, and how much space is used and free. One command replaces lsblk, df -h and findmnt, and lines up their answers so you do not have to match device names by eye.

With --health, it asks each drive about its own condition through SMART, the self-test data every modern drive keeps. That shows wear on an SSD, power-on hours, temperature, and the error counters that warn of a drive about to fail.

disks only reads. It never mounts, unmounts, formats, partitions or writes to any device. The SMART query in --health reads the drive's counters and starts no self-test.

How it builds the table

disks runs three read-only commands and joins what they say.

1. lsblk lists every block device with its size, type, filesystem, model and parent. The parent link makes the tree.
2. findmnt lists every mount, including the ones lsblk shows only once, such as a btrfs partition mounted at both / and /home.
3. df gives used and free space for each mount.

A partition mounted in more than one place shows all its mount points, separated by commas, as in /, /home. Used and free space come from the first one, since every mount of one filesystem shares the same space.

A drive with no partitions and no filesystem shows its model name in the mount column, so you can tell two drives of the same size apart.

Sizes

Sizes use the same short units as lsblk, in powers of 1024. 600M is 600 mebibytes, 239G is 239 gibibytes. A drive sold as 256 GB shows as about 239G, because the maker counts in powers of 1000.

The low mark

A mount that is more than 85% full gets the word low at the end of its row. --warn changes that level. tmpfs mounts never get the mark, because they live in memory and fill and empty on their own.

At 85%, most filesystems still work well. Past about 95%, ext4 and xfs slow down, btrfs can refuse writes even with free space left, and package updates fail halfway. The mark gives you time to clean up first. cleanup and bigfiles help with that.

What it hides

By default disks shows real drives and what is on them. It hides these, which -a brings back.

zram swap stays visible, because it takes real memory and shows up in free and mem.

Tree drawing

In a UTF-8 terminal the tree uses line drawing characters, ├ and └. In a plain C or POSIX locale it uses |- and `- instead, so the output stays readable over a serial console or in an old terminal.

SMART health

--health runs smartctl -H -i -A on each whole drive. The query needs root, because it talks to the drive directly. When you are not root, disks runs smartctl through sudo, which asks for your password in a terminal. With no terminal, it uses sudo -n, which fails at once instead of waiting for a password that cannot come.

For each drive it shows the lines that matter and skips the rest of smartctl's long report.

Drives behind many USB adapters cannot answer SMART at all. disks says not available over this USB bridge for those. Some adapters answer when told the protocol, with disks --health -- -d sat.

zram, loop and RAM devices are not real drives and have no SMART data, so --health skips them.

Options

disks takes no device names. It always shows every drive, because a problem is often on the drive you did not think to look at. Use lsblk /dev/sda for one device.

Pass-through

With --health, options after -- go to smartctl, before the device name. The common ones are these.

The same options go to every drive, so pass-through is most useful with one drive attached. Without --health, -- is a usage error.

Per-distro notes

• Debian and Ubuntu. smartctl is in the smartmontools package. Installing it also starts smartd, which checks drives in the background and mails root about problems.
• Fedora and RHEL. The package is smartmontools. Whether smartd starts on its own depends on the release. Check with svc smartd, and turn it on with sudo systemctl enable --now smartd.
• Arch. The package is smartmontools.
• openSUSE. The package is smartmontools.
• Alpine. lsblk and findmnt are in lsblk and findmnt packages, not in busybox. smartctl is in smartmontools.
• Virtual machines. Virtual drives such as vda have no SMART data. --health says so for each one. The host's drives are the ones to check.
• btrfs. One partition often holds several subvolumes mounted in different places. They share one pool of space, which disks shows once, with every mount point on the row.
• LVM and LUKS. A partition that holds LVM or an encrypted volume shows as lvm or luks in the FS column, with the logical or opened volumes under it in the tree.

Needs

• lsblk and findmnt, from util-linux. Every distro has them except minimal Alpine.
• df, from coreutils.
• smartctl, from smartmontools, for --health only.
• sudo, for --health when you are not root.

Examples

A laptop with a USB stick in it

$ disks
DEVICE       SIZE  FS       MOUNTED ON    USED   FREE   USE
nvme0n1      239G           SAMSUNG MZVLB256HAHQ-000L7
├ nvme0n1p1  600M  vfat     /boot/efi      19M   581M    4%
├ nvme0n1p2    1G  ext4     /boot         412M   478M   47%
└ nvme0n1p3  237G  btrfs    /, /home      214G    23G   91% low
sda           58G           SanDisk Ultra
└ sda1        58G  exfat    /run/media/khadir/USB
                                           31G    27G   54%
zram0        7.6G  swap     [SWAP]
health  run disks --health, it needs sudo

The root filesystem is 91% full, past the low mark. A mount point too long for its column moves the numbers to the next line.

A virtual machine

$ disks
DEVICE       SIZE  FS       MOUNTED ON    USED   FREE   USE
vda           50G
└ vda1        50G  ext4     /              44G   2.5G   95% low
health  run disks --health, it needs sudo

Everything, with shares and memory filesystems

$ disks -a
DEVICE       SIZE  FS       MOUNTED ON    USED   FREE   USE
sr0            1G           QEMU DVD-ROM
vda           50G
└ vda1        50G  ext4     /              44G   2.5G   95% low
tmpfs        795M  tmpfs    /run          1.2M   794M    1%
tmpfs        3.9G  tmpfs    /dev/shm      120M   3.8G    4%
tmpfs          5M  tmpfs    /run/lock       8K     5M    1%
tmpfs        3.9G  tmpfs    /tmp          3.2G   703M   83%
host_share   229G  virtiofs /mnt/host     210G    19G   92% low

The shared folder from the host is 92% full too. That space belongs to the host's drive.

Drive health

$ disks --health
[sudo] password for khadir:
nvme0n1   SAMSUNG MZVLB256HAHQ-000L7
  smart             PASSED
  wear              11% used
  power on          9,412 hours
  written           48.2 TB
  temperature       38 C
  media errors      0
  unsafe shutdowns  71
sda       SanDisk Ultra
  smart             not available over this USB bridge

This SSD has used 11% of its rated life in a little over a year of running time, and has no media errors.

A check for a cron job or a script

$ disks -q --warn 90
/ is 95% full, 2.5G free
$ echo $?
1

With -q and --warn, disks prints only the full mounts, and nothing at all when every mount is under the level. The exit status tells the script which case it is.

A USB drive that needs a protocol hint

$ disks --health -- -d sat
nvme0n1   SAMSUNG MZVLB256HAHQ-000L7
  smart             not available on this drive
sda       SanDisk Ultra
  smart             PASSED
  power on          112 hours
  temperature       30 C

-d sat made the USB adapter pass SMART through, but it also went to the NVMe drive, which does not speak that protocol. Run the command once with the stick attached and once without.

Troubleshooting

disks: --health needs root. Run it in a terminal so sudo can ask, or as root There was no terminal for sudo to ask for a password, such as in a cron job. Run sudo disks --health instead, or add a sudo rule for smartctl only.

disks: needs smartctl. Install smartmontools. The message shows the install command for your distro.

smart not available on this drive The drive is virtual, or SMART is off in its firmware. Try sudo smartctl -s on /dev/sdX to turn it on for a real drive.

The numbers do not match df -h. df -h rounds differently and counts some reserved blocks. ext4 keeps 5% of the space for root, so a full ext4 shows 95% in some tools and 100% in others. disks uses df's own percentage.

A mounted drive is missing. It is a loop device, a network share or a tmpfs. Run disks -a.

low on a mount you cannot clean. A drive meant to be full, such as a backup disk, can go over the mark on purpose. Use --warn 98 in scripts that watch it.

Exit status

See also

mem, bigfiles, cleanup, temps, lsblk(8), findmnt(8), df(1), smartctl(8)