181

How to know the size of a directory? Including subdirectories and files.

11 Answers 11

234
du -s directory_name

Or to get human readable output:

du -sh directory_name

The -s option means that it won't list the size for each subdirectory, only the total size.

  • 7
    Actually du's default unit is 512-byte blocks according to POSIX, and kilobytes on Linux (unless the environment variable POSIXLY_CORRECT is set) or with du -k. – Gilles 'SO- stop being evil' Oct 12 '10 at 17:49
  • 3
    @Gilles: Good catch. I've removed the "number of bytes" bit from my answer. – sepp2k Oct 12 '10 at 17:53
  • 1
    worked as prescribed – skidadon May 28 '15 at 19:17
  • 1
    if the directory is very big and have lots of subdirectories, it takes lots of time... almost 1 min.. is that normal? is there a way to get the size more rapidly? – yeahman Oct 15 '15 at 19:59
  • 2
    I needed to calculate the size of my folder "bag", du -sh bag worked perfectly! – António Almeida Mar 4 '16 at 12:15
7

While using a separate package such as ncdu may work well, the same comparison of many folders can be done, to some degree, by just giving du a list of folders to size up. For example to compare top-level directories on your system...

cd /    
sudo du -sh ./*
  • 2
    More simply, du -sh /* – roaima Sep 10 '15 at 17:00
7

GNU du takes a -b option.

See the man page and the info page for more help:

-b, --bytes is equivalent to --apparent-size --block-size=1

1
du -csh

-c produces grand total

  • 1
    The -c doesn't make sense to use together with -s, right? -s only displays the size of the specified directory, that is the total size of the directory. – Andreas Storvik Strauman Jun 5 '18 at 10:43
1

Try

du -hax --max-depth=1 / | grep '[0-9]G' | sort -nr

This helps find large directories to then sift through using du -sh ./*

1

You can use "file-size.sh" from the awk Velour library:

ls -ARgo "$@" | awk '{q += $3} END {print q}'
  • This gives a more accurate count than du. Unpack a tarball on two servers and use "du -s" (with or without --bytes) and you will likely see different totals, but using this technique the totals will match. – Angelo Babudro 15 hours ago
0

you can also use ls -ldh:

ls -ldh /etc drwxr-xr-x 145 root root 12K 2012-06-02 11:44 /etc

-l is for long listing ; -d is for displaying dir info, not the content of the dir, -h is for displaying size in huma readable format.

  • 4
    This isn't correct, the person asking is clearly looking for footprint of a directory and it's contents on disk. @sepp2k's answer is correct. – blong Jun 5 '12 at 13:16
  • 1
    The ls -ldh command only shows the size of inode structure of a directory. The metric is a reflection of size of the index table of file names, but not the actual size of the file content within the directory. – linbianxiaocao Mar 28 '16 at 18:19
0

I always install the "ncdu" package and see all the output of all directories with graphical representation. This is because I usually need to know what's taking up the most disk space on my machines, regardless of how much a single directory sums up.

Usage: sudo ncdu / (You do not need sudo for folders on which you have read permission).

It will take a while to scan disk usage statistics on the whole file system. It has a nice command line graphical representation and included keyboard navigation using the arrow keys, like going deeper or higher in the scanned path. You can also delete items by pressing D.

0

du -hd1

will list in human-readable format the sizes of all the directories, e.g.

656K    ./rubberband
2.2M    ./lame
652K    ./pkg-config
0

I tried with below command since already best answer has been provided

sudo find . -maxdepth 1 -exec du -shk {} \;| awk 'NR >1'| awk 'BEGIN{sum=1}{sum=sum+$1}END{print sum}'

output

 sudo find . -maxdepth 1 -exec du -shk {} \;| awk 'NR >1'| awk 'BEGIN{sum=1}{sum=sum+$1}END{print sum}'

679445
0

The original question asked the size, but did not specify if it was the size on disk or the actual size of data.

I have found that the calculation of 'du' can vary between servers with the same size partition using the same file system. If file system characteristics differ this makes sense, but otherwise I can't figure why. The 'ls|awk" answer that Steven Penny gave yields a more consistent answer, but still gave me inconsistent results with very large file lists.

Using 'find' gave consistent results for 300,000+ files, even when comparing one server using XFS and another using EXT4. So if you want to know the total bytes of data in all files then I suggest this is a good way to get it:

find /whatever/path -type f -printf "%s\n"|awk '{q+=$1} END {print q}'

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.