Yoran Heling
home - git - @ayo
= pgp = only used for releases
key - mit
7446 0D32 B808 10EB A9AF A2E9 6239 4C69 8C27 39FA

Cute decorative scissors, cutting through your code.

Ncdu Manual


ncdu - NCurses Disk Usage


ncdu [options] dir


ncdu (NCurses Disk Usage) is a curses-based version of the well-known 'du', and provides a fast way to see what directories are using your disk space.


Mode Selection

-h, --help

Print a short help message and quit.

-v, -V, --version

Print ncdu version and quit.


Load the given file, which has earlier been created with the -o option. If FILE is equivalent to -, the file is read from standard input.

For the sake of preventing a screw-up, the current version of ncdu will assume that the directory information in the imported file does not represent the filesystem on which the file is being imported. That is, the refresh, file deletion and shell spawning options in the browser will be disabled.


Scan the given directory.


Export all necessary information to FILE instead of opening the browser interface. If FILE is -, the data is written to standard output. See the examples section below for some handy use cases.

Be warned that the exported data may grow quite large when exporting a directory with many files. 10.000 files will get you an export in the order of 600 to 700 KiB uncompressed, or a little over 100 KiB when compressed with gzip. This scales linearly, so be prepared to handle a few tens of megabytes when dealing with millions of files.

-e, --extended, --no-extended

Enable/disable extended information mode. This will, in addition to the usual file information, also read the ownership, permissions and last modification time for each file. This will result in higher memory usage (by roughly ~30%) and in a larger output file when exporting.

When using the file export/import function, this flag will need to be added both when exporting (to make sure the information is added to the export), and when importing (to read this extra information in memory). This flag has no effect when importing a file that has been exported without the extended information.

This enables viewing and sorting by the latest child mtime, or modified time, using 'm' and 'M', respectively.


Do not attempt to load any configuration files.

Scan Options

These options affect the scanning progress, and have no effect when importing directory information from a file.

-x, --one-file-system

Do not cross filesystem boundaries, i.e. only count files and directories on the same filesystem as the directory being scanned.


Do cross filesystem boundaries. This is the default, but can be specified to overrule a previously given -x.

--exclude PATTERN

Exclude files that match PATTERN. The files will still be displayed by default, but are not counted towards the disk usage statistics. This argument can be added multiple times to add more patterns.

-X FILE, --exclude-from FILE

Exclude files that match any pattern in FILE. Patterns should be separated by a newline.

--include-caches, --exclude-caches

Include (default) or exclude directories containing CACHEDIR.TAG. The directories will still be displayed, but their contents will not be scanned or counted towards the disk usage statistics. http://www.brynosaurus.com/cachedir/

-L, --follow-symlinks, --no-follow-symlinks

Follow (or not) symlinks and count the size of the file they point to. As of ncdu 1.14, this option will not follow symlinks to directories and will count each symlinked file as a unique file (i.e. unlike how hard links are handled). This is subject to change in later versions.

--include-kernfs, --exclude-kernfs

(Linux only) Include (default) or exclude Linux pseudo filesystems, e.g. /proc (procfs), /sys (sysfs).

The complete list of currently known pseudo filesystems is: binfmt, bpf, cgroup, cgroup2, debug, devpts, proc, pstore, security, selinux, sys, trace.

Interface options


Don't give any feedback while scanning a directory or importing a file, other than when a fatal error occurs. Ncurses will not be initialized until the scan is complete. When exporting the data with -o, ncurses will not be initialized at all. This option is the default when exporting to standard output.


Similar to -0, but does give feedback on the scanning progress with a single line of output. This option is the default when exporting to a file.

In some cases, the ncurses browser interface which you'll see after the scan/import is complete may look garbled when using this option. If you're not exporting to a file, -2 is probably a better choice.


Provide a full-screen ncurses interface while scanning a directory or importing a file. This is the only interface that provides feedback on any non-fatal errors while scanning.

-q, --slow-ui-updates, --fast-ui-updates

Change the UI update interval while scanning or importing. Ncdu will update the screen 10 times a second by default (--fast-ui-updates), this can be decreased to once every 2 seconds with -q or --slow-ui-updates. This feature can be used to save bandwidth over remote connections. This option has no effect when -0 is used.

--enable-shell, --disable-shell

Enable or disable shell spawning from the browser. This feature is enabled by default when scanning a live directory and disabled when importing from file.

--enable-delete, --disable-delete

Enable or disable the built-in file deletion feature. This feature is enabled by default when scanning a live directory and disabled when importing from file. Explicitly disabling the deletion feature can work as a safeguard to prevent accidental data loss.

--enable-refresh, --disable-refresh

Enable or disable directory refreshing from the browser. This feature is enabled by default when scanning a live directory and disabled when importing from file.


Read-only mode. When given once, this is an alias for --disable-delete, when given twice it will also add --disable-shell, thus ensuring that there is no way to modify the file system from within ncdu.

--si, --no-si

List sizes using base 10 prefixes, that is, powers of 1000 (KB, MB, etc), as defined in the International System of Units (SI), instead of the usual base 2 prefixes, that is, powers of 1024 (KiB, MiB, etc).

--disk-usage, --apparent-size

Select whether to display disk usage (default) or apparent sizes. Can also be toggled in the browser with the 'a' key.

--show-hidden, --hide-hidden

Show (default) or hide "hidden" and excluded files. Can also be toggled in the browser with the 'e' key.

--show-itemcount, --hide-itemcount

Show or hide (default) the item counts column. Can also be toggled in the browser with the 'c' key.

--show-mtime, --hide-mtime

Show or hide (default) the last modification time column. Can also be toggled in the browser with the 'm' key. This option is ignored when not in extended mode (see -e).

--show-graph, --hide-graph

Show (default) or hide the relative size bar column. Can also be toggled in the browser with the 'g' key.

--show-percent, --hide-percent

Show (default) or hide the relative size percent column. Can also be toggled in the browser with the 'g' key.

--graph-style OPTION

Change the way that the relative size bar column is drawn. Recognized values are hash to draw ASCII # characters (default and most portable), half-block to use half-block drawing characters or eighth-block to use eighth-block drawing characters. Eighth-block characters are the most precise but may not render correctly in all terminals.

--shared-column OPTION

Set to off to disable the shared size column for directories, shared (default) to display shared directory sizes as a separate column or unique to display unique directory sizes as a separate column. These options can also be cycled through in the browser with the 'u' key.

--sort COLUMN

Change the default column to sort on. Accepted values are disk-usage (the default), name, apparent-size, itemcount or mtime. The latter only makes sense in extended mode, see -e.

The column can be suffixed with -asc or -desc to set the order to ascending or descending, respectively. e.g. --sort=name-desc will sort by name in descending order.

--enable-natsort, --disable-natsort

Enable (default) or disable natural sort when sorting by file name.

--group-directories-first, --no-group-directories-first

Sort (or not) directories before files.

--confirm-quit, --no-confirm-quit

Require a confirmation before quitting ncdu. Very helpful when you accidentally press 'q' during or after a very long scan.

--confirm-delete, --no-confirm-delete

Require a confirmation before deleting a file or directory. Enabled by default, but can be disabled if you're absolutely sure you won't accidentally press 'd'.

--color SCHEME

Select a color scheme. The following schemes are recognized: off to disable colors, dark for a color scheme intended for dark backgrounds and dark-bg for a variation of the dark color scheme that also works in terminals with a light background.

The default is dark-bg unless the NO_COLOR environment variable is set.


Ncdu can be configured by placing command-line options in /etc/ncdu.conf or $HOME/.config/ncdu/config. If both files exist, the system configuration will be loaded before the user configuration, allowing users to override options set in the system configuration. Options given on the command line will override options set in the configuration files. The files will not be read at all when --ignore-config is given on the command line.

The configuration file format is simply one command line option per line. Lines starting with # are ignored. Example configuration file:

# Always enable extended mode

# Disable file deletion

# Exclude .git directories
--exclude .git



Show help + keys + about screen

up, down, j, k

Cycle through the items

right, enter, l

Open selected directory

left, <, h

Go to parent directory


Order by filename (press again for descending order)


Order by filesize (press again for descending order)


Order by number of items (press again for descending order)


Toggle between showing disk usage and showing apparent size.


Order by latest child mtime, or modified time. (press again for descending order) Requires the -e flag.


Delete the selected file or directory. An error message will be shown when the contents of the directory do not match or do not exist anymore on the filesystem.


Toggle dirs before files when sorting.


Toggle between showing percentage, graph, both, or none. Percentage is relative to the size of the current directory, graph is relative to the largest item in the current directory.


Toggle display of the shared / unique size column for directories that share hard links. This column is only visible if the current listing contains directories with shared hard links.


Toggle display of child item counts.


Toggle display of latest child mtime, or modified time. Requires the -e flag.


Show/hide 'hidden' or 'excluded' files and directories. Please note that even though you can't see the hidden files and directories, they are still there and they are still included in the directory sizes. If you suspect that the totals shown at the bottom of the screen are not correct, make sure you haven't enabled this option.


Show information about the current selected item.


Refresh/recalculate the current directory.


Spawn shell in current directory.

Ncdu will determine your preferred shell from the NCDU_SHELL or SHELL variable (in that order), or will call /bin/sh if neither are set. This allows you to also configure another command to be run when he 'b' key is pressed. For example, to spawn the vifm(1) file manager instead of a shell, run ncdu as follows:

export NCDU_SHELL=vifm

Ncdu will set the NCDU_LEVEL environment variable or increment it before spawning the shell. This variable allows you to detect when your shell is running from within ncdu, which can be useful to avoid nesting multiple instances of ncdu. Ncdu itself does not (currently) warn when attempting to run nested instances.




Entries in the browser interface may be prefixed by a one-character flag. These flags have the following meaning:


An error occurred while reading this directory.


An error occurred while reading a subdirectory, so the indicated size may not be correct.


File or directory is excluded from the statistics by using exclude patterns.


Directory is on another filesystem.


Directory is excluded from the statistics due to being a Linux pseudo filesystem.


This is neither a file nor a folder (symlink, socket, ...).


Same file was already counted (hard link).


Empty directory.


To scan and browse the directory you're currently in, all you need is a simple:


If you want to scan a full filesystem, your root filesystem, for example, then you'll want to use -x:

ncdu -x /

Since scanning a large directory may take a while, you can scan a directory and export the results for later viewing:

ncdu -1xo- / | gzip >export.gz
# ...some time later:
zcat export.gz | ncdu -f-

To export from a cron job, make sure to replace -1 with -0 to suppress any unnecessary output.

You can also export a directory and browse it once scanning is done:

ncdu -o- | tee export.file | ./ncdu -f-

The same is possible with gzip compression, but is a bit kludgey:

ncdu -o- | gzip | tee export.gz | gunzip | ./ncdu -f-

To scan a system remotely, but browse through the files locally:

ssh -C user@system ncdu -o- / | ./ncdu -f-

The -C option to ssh enables compression, which will be very useful over slow links. Remote scanning and local viewing has two major advantages when compared to running ncdu directly on the remote system: You can browse through the scanned directory on the local system without any network latency, and ncdu does not keep the entire directory structure in memory when exporting, so you won't consume much memory on the remote system.


Every disk usage analysis utility has its own way of (not) counting hard links. There does not seem to be any universally agreed method of handling hard links, and it is even inconsistent among different versions of ncdu. This section explains what each version of ncdu does.

ncdu 1.5 and below does not support any hard link detection at all: each link is considered a separate inode and its size is counted for every link. This means that the displayed directory sizes are incorrect when analyzing directories which contain hard links.

ncdu 1.6 has basic hard link detection: When a link to a previously encountered inode is detected, the link is considered to have a file size of zero bytes. Its size is not counted again, and the link is indicated in the browser interface with a 'H' mark. The displayed directory sizes are only correct when all links to an inode reside within that directory. When this is not the case, the sizes may or may not be correct, depending on which links were considered as "duplicate" and which as "original". The indicated size of the topmost directory (that is, the one specified on the command line upon starting ncdu) is always correct.

ncdu 1.7 and later has improved hard link detection. Each file that has more than two links has the "H" mark visible in the browser interface. Each hard link is counted exactly once for every directory it appears in. The indicated size of each directory is therefore, correctly, the sum of the sizes of all unique inodes that can be found in that directory. Note, however, that this may not always be same as the space that will be reclaimed after deleting the directory, as some inodes may still be accessible from hard links outside it.


Directory hard links and firmlinks (MacOS) are not supported. They will not be detected as being hard links, and may thus be scanned and counted multiple times.

Some minor glitches may appear when displaying filenames that contain multibyte or multicolumn characters.

The unique and shared directory sizes are calculated based on the assumption that the link count of hard links does not change during a filesystem scan or in between refreshes. If it does, for example after deleting a hard link, then these numbers will be very much incorrect and a full refresh by restarting ncdu is needed to get correct numbers again.

All sizes are internally represented as a signed 64bit integer. If you have a directory larger than 8 EiB minus one byte, ncdu will clip its size to 8 EiB minus one byte. When deleting or refreshing items in a directory with a clipped size, the resulting sizes will be incorrect. Likewise, item counts are stored in a 32-bit integer, so will be incorrect in the unlikely event that you happen to have more than 4 billion items in a directory.

Please report any other bugs you may find at the bug tracker, which can be found on the web site at https://dev.yorhel.nl/ncdu


Written by Yoran Heling <projects@yorhel.nl>.