These three classic Unix utilities like to get their hands dirty with file systems.
****************************************************
df – display disk space usage on mounted filesystems
****************************************************
Every newbie sysadmin learns df right off the bat; it displays mounted file systems and the disk space usage on each. By default (in linux) it displays the statistics as measured in 1k blocks. For the less mathematically inclined, try the -h (human readable) option:
[usr-3@srv-3 usr-3]$ df -h Filesystem Size Used Avail Use% Mounted on /dev/hda1 1.9G 108M 1.6G 6% / /dev/hda7 10G 2.4G 7.2G 25% /home /dev/hda5 99M 435k 93M 1% /tmp /dev/hda6 2.9G 1.8G 972M 66% /usr /dev/hda3 243M 93M 138M 40% /var none 251M 0 251M 0% /dev/shm eel:/share 169G 74G 87G 46% /share
You’ll notice that both local and nfs filesystems are included. Sometimes you’ll get wacky messages that a device is full but df shows plenty of space. If your filesystem holds many small files, chances are you’re out of inodes. To check this use the -i option:
[usr-3@srv-3 usr-3]$ df -i Filesystem Inodes IUsed IFree IUse% Mounted on /dev/hda1 251392 17692 233700 8% / /dev/hda7 1359872 63108 1296764 5% /home /dev/hda5 26104 50 26054 1% /tmp /dev/hda6 384768 100644 284124 27% /usr /dev/hda3 64512 629 63883 1% /var none 64304 1 64303 1% /dev/shm eel:/share 0 0 0 - /share
The number of inodes is set at filesystem creation time, and cannot be changed once the filesystem is made. So if you’re creating a file system which you know will be holding many small files of 1-10K, use the options in mke2fs to increase the number of inodes beyond the default. Finally, use the -T option to include file system types in the listing:
[usr-3@srv-3 usr-3]$ df -T Filesystem Type 1k-blocks Used Available Use% Mounted on /dev/hda1 ext3 1976492 110504 1765584 6% / /dev/hda7 ext3 10689180 2509896 7636292 25% /home /dev/hda5 ext2 101089 435 95435 1% /tmp /dev/hda6 ext3 3028080 1878800 995460 66% /usr /dev/hda3 ext2 248895 94291 141752 40% /var none tmpfs 257216 0 257216 0% /dev/shm eel:/share nfs 177265344 77013672 91247124 46% /share
************************************************
du – estimate disk space usage by directory/file
************************************************
Now that you’ve determined your disk space usage, how will you figure out who or what is hogging up all that disk space? One way is to use du to summarize disk space usage by directory/file. For a quick summary of the space usage in a directory:
[root@srv-3 home]# du -h --max-depth=1 16k ./lost+found 2.2G ./usr-3 188M ./eore 21M ./web 2.4G . Again, -h makes it human readable, and the --max-depth option tells du to summarize directories which are only n or fewer levels down from the command line argument (in this case /home, our current working directory because we did not specify a directory on the command line.) If we forgo the -h option, we can pipe it through sort and find the piggiest files:
[root@srv-3 usr-3]# du | sort -g
the biggest directories show up at the bottom:
429812 ./code/blh/base/src 513876 ./music/zz 532640 ./code/blh/base 541740 ./music 612196 ./tools/general 781756 ./tools 823384 ./code/blh 862448 ./code 2264696 .
So let's investigate that pesky music directory -- that doesn't look legit for a user on one of *my* systems! Use the -a option to list counts for files as well as directories:
[root@srv-3 music]# du -a | sort -n 28 ./songlists 4240 ./shake-your-ass/Bodyrock.mp3 4552 ./shake-your-ass/11_Southern_girls.mp3 4924 ./shake-your-ass/Chemical Brothers - Block Rockin' Beat.mp3 5400 ./shake-your-ass/VIOLENTLY_HAPPY_Fluke(Well_Tempered).mp3 24300 ./zz/6girlinformme.wav 27828 ./shake-your-ass 28420 ./zz/95rottenworldblues.wav 34100 ./zz/91musiqueautomatic.wav 34276 ./zz/92ratormole.wav 34524 ./zz/3caringiscreepy.wav 513876 ./zz 541740 .
Uh huh. I could go have a talk with this user! Especially if I'm backing her music files to tape with my file server backup! Or perhaps I should just mount her home directory on my workstation.
******************************************************************
dd – copy and convert files
******************************************************************
The dd utility is used to copy and convert files block by block. It reads from standard in by default and writes to standard out.
[usr-3@srv-3 tmp]$ cat upper I AM A BUNCH OF UPPER CASE CHARACTERS. [usr-3@srv-3 tmp]$ dd conv=lcase < upper > lower 0+1 records in 0+1 records out [usr-3@srv-3 tmp]$ cat lower i am a bunch of upper case characters.
One can also specify input files and output files at the command line:
[usr-3@srv-3 tmp]$ dd conv=lcase if=upper of=lower 0+1 records in 0+1 records out
There are several conversion options which can be discovered by reading the fine dd man page; to be honest I've never used dd for this much at all. I use it to make images of file systems. example: make a Red Hat boot floppy.
[root@srv-3 images]# dd if=bootnet.img of=/dev/fd0 2880+0 records in 2880+0 records out [root@srv-3 images]# mount /dev/fd0 /mnt/floppy -t msdos [root@srv-3 images]# cd /mnt/floppy/ [root@srv-3 floppy]# ls boot.msg initrd.img param.msg snake.msg vmlinuz general.msg ldlinux.sys rescue.msg syslinux.cfg
And there you have it. A bootable Red Hat install floppy. Add a kickstart config file and you're on your way to having unattended automated installs. It's useful for larger file systems as well (this takes a few minutes and works the system):
[root@srv-3 home]# dd if=/dev/hda3 of=/home/tmp/var.img 514080+0 records in 514080+0 records out [root@srv-3 tmp]# mount var.img var -o loop [root@srv-3 tmp]# cd var [root@srv-3 var]# ls cache gdm local log mail opt run tmp yp db lib lock lost+found nis preserve spool www [root@srv-3 tmp]# df Filesystem 1k-blocks Used Available Use% Mounted on /dev/hda1 1976492 110508 1765580 6% / /dev/hda7 10689180 2767220 7378968 28% /home /dev/hda5 101089 435 95435 1% /tmp /dev/hda6 3028080 1878800 995460 66% /usr /dev/hda3 248895 94291 141752 40% /var none 257216 0 257216 0% /dev/shm eel:/share 177265344 77013676 91247120 46% /share /home/tmp/var.img 248895 94291 141752 40% /home/tmp/var
The -o loop option to mount allows us to mount files which contain file system images as filesystems. It is handy, let me tell you! This article has been brought to you by the letter D.
