df shows full disk, even after removing lots of files

lennih · 03-17-2012, 10:11 PM

Hey, this is my first thread here.

A backup routine filled my hard drive. After removing at least 10GB with the rm command, df -h shows the same:

Code:

Filesystem   Size   Used   Avail   Use%   Mounted on
/dev/hda2    36G    36G    0       100%   /

Every previous thread I read says there can be a problem with the inodes, but after rebooting the problem should be solved. I rebooted about 37 times. Inodes seem to be ok. df -i shows:

Code:

Filesystem   Inodes   IUsed   IFree   IUsed  Mounted on
/dev/hda2   9647680   58668 9589012      1%  /

df -h keeps displaying the same as above.
du --max-depth=1 -h shows:

Code:

16K     ./lost+found
352K    ./dev
652M    ./var
4.8M    ./bin
9.6M    ./etc
17M     ./lib
16K     ./mnt
816K    ./tmp
659M    ./home
1.0K    ./proc
3.7M    ./sbin
44K     ./root
36G     .

What does that last line mean? Does du agree with df and both are saying the disk is full?

I unmounted the partition, ran a fsck but it didn't solve anything. What can I do? I swear I looked for help in previous threads quite extensively before bothering here.

Distro: Slackware 9.1.0
Kernel 2.4.22

Thanks in advance.

Tinkster · 03-17-2012, 10:31 PM

Yes, it does. The disk is full.

What do you get from
du -sm * | sort -n

Cheers,
Tink

lennih · 03-17-2012, 11:09 PM

Quote:

Originally Posted by Tinkster

What do you get from
du -sm * | sort -n

Thanks for the quick reply.
I get

Code:

0   adjtime
0   aliases.tmp
0   dhcpd.conf
0   gshadow
0   ioctl.save
0   issue.net
0   ld.so.cache~
0   localtime
0   mgetty+sendfax
0   modprobe.conf
0   modules.conf
0   random-seed
0   rmt
1   DIR_COLORS
1   HOSTNAME
1   aliases
1   aliases.cdb
1   apache
1   backup.list
1   bootptab
1   checkps.list
1   cron.daily
1   cron.hourly
1   cron.monthly
1   cron.weekly
1   chs.login
1   default
1   devfsd.conf
1   dhclient.conf
1   dhcpc
1   dhcpd.conf.old
1   dhcpd.conf.template
1   dialogrc
1   fb.modes
1   fdprm
1   fonts
1   fstab
1   ftpusers
1   gettydefs
1   gpm-root.conf
1   gpm-syn.conf
1   gpm-twiddler.conf
1   group
1   group.preqmail
1   host.conf
1   hosts
1   hosts.allow
1   hosts.deny
1   hosts.equiv
1   inittab
1   inputrc
1   isapnp.conf.sample
1   isapnp.gone.sample
1   issue
1   kamchatka
1   ld.so.cache
1   ld.so.conf
1   login.access
1   login.defs
1   logrotate.conf
1   logrotate.d
1   magic
1   magic.mime
1   mediaprm
1   mgetty.inittab
1   minicom.users
1   minirc.dfl
1   modules.devfs
1   motd
1   mtab
1   mtools.conf
1   my-huge.cnf
1   my-large.cnf
1   my-medium.cnf
1   my-small.cnf
1   named.conf
1   networks
1   nntpserver
1   nscd.conf
1   nsswitch.conf
1   passwd
1   passwd-
1   passwd.preqmail
1   pear.conf
1   ppp
1   printcap
1   procmailrc
1   profile
1   profile.d
1   proftpd.conf
1   protocols
1   rc.d
1   resolv.conf
1   rndc.key
1   rpc
1   samba
1   securetty
1   serial.conf
1   services
1   shadow
1   shadow-
1   shells
1   skel
1   slackware-version
1   ssh
1   ssh2
1   ssl
1   stunnel
1   sudoers
1   syslog.conf
1   tcp.opensmtp.cdb
1   tcp.smtp.cdb
1   termcap
1   termcap-BSD
1   termcap-Linux
1   updatedb.conf
1   wgetrc
4   X11
4   codepages

Tinkster · 03-18-2012, 11:47 AM

Sorry, let's try that again, from the root dir rather than /etc

lennih · 03-18-2012, 01:00 PM

hehe, sorry.

Code:

1       dev
1       lost+found
1       mnt
1       proc
1       root
1       tmp
2       boot
4       sbin
10      etc
17      lib
652     var
657     home
659     usr
34452   nohup.root

Thanks again. What should I do?

Tinkster · 03-18-2012, 01:59 PM

Considering that nohup.root, a single file, consumes 34 GB you may
want to find out what produces it ;}

Look at its content (maybe head -n 30 /nohup.root) ...
Find out which process owns it (lsof | grep nohup.root).

Cheers,
Tink