[SOLVED] Can't startx - no space left on device ( What/How to clean/remove stuff)

sjampoo · 08-02-2012, 03:01 AM

Dear all,

I'm stuck.
A Slackware 13.37 x64 machine, running mainly an installation of VirtualBox has stopped working.

Somehow the root partition got full.

It's 46GB in size, and 100% in use.

The VM disk images are on a different partition

df -h shows:
/dev/root 46G 46G 0 100% /
/dev/sda2 42G 984M 39G 3% /home
/dev/sda3 358G 98G 242G 29% /other
tmpfs 3.9G 1.7M 3.9G 1% /dev/shm

So as you can see: I was confident there was enough space for the / partition to grow a bit.

I suspect there's some log of tmp directory that got full, but as I'm no hero at the commandline, and I can't start XFCE.. I'm stuck.

I can ssh into the machine, and that's a comfort, since the machnie is down in the basement; warm, dark and very uncomfortable.

Where on / should I start looking? and which commands do I use?

* can I remove /var/cache/ or is this an unsafe operation? *

ponce · 08-02-2012, 03:07 AM

try this command (better as root), it will generate a sorted file that you can inspect

Code:

cd /
du -a -x | sort -n > /home/disk_space.tmp

then you can inspect the generated file with tail: for example, to see the last 20 lines

Code:

tail -20 /home/disk_space.tmp

when you had deleted the superfluos stuff with "rm -fR", you can rm the generated file too.

NoStressHQ · 08-02-2012, 03:12 AM

Well first of all, manually clean your '/tmp'...

Then ponce gave your the right path, a more "assisted" path would be using ncdu instead (graphical browsing of your fs), it can be found on sbopkg and takes no time to install.

Cheers

Garry.

Code:

# ncdu -x /

The -x avoids ncdu (/du) to scan other physical device than the original one...

Celyr · 08-02-2012, 03:12 AM

You can safely delete everyting in a tmp directory (/tmp and /var/tmp). Then i suggest you to play a bit with du because a full slackware install uses around 8G

sjampoo · 08-02-2012, 03:17 AM

Oh. Rats!

Norman ( Virusscanner ) has got about 38GB as .tmp files.
(or had,.

)

Thanx a million Ponce!

sycamorex · 08-02-2012, 03:20 AM

Additionally as a more permanent solution, you could add the following command:

Code:

/usr/bin/find /tmp -mindepth 1 -maxdepth 1 -print0 | xargs -0r /bin/rm -rf

to your /etc/rc.d/rc.local_shutdown which will automatically clear your /tmp directory during a shutdown.

mrascii · 08-02-2012, 09:05 AM

My church called me a few months back with the same problem on their Linux file server. Strangely, it was the /media directory which was causing / to overflow. The server was being backed up to an external USB drive daily and then the drive removed and stored in the safe until the next day. One day the drive wasn't attached when the backup script ran and the script not finding the /media/backup directory created the directory (on the internal drive, of course) and started the backup. Adding a line to the script to test for the presence of the correct directory before starting the backup fixed the problem.

DNA
AKA mrascii

perbh · 08-02-2012, 01:41 PM

Just as an apropos to the last post - this is how all _my_ backup-scripts work:

Code:

myusb=".................." # insert appropriate blkid for usb-drive
usbdev=""
mountpoint=""
die() { echo "$*"; unset myusb usbdev mountpoint; exit 1; }

usbdev="`blkid -U $myusb`" || usbdev=""
test x$usbdev = x && die "-- usbdisk is NOT present"

# now check if it's allready mounted
mountpoint="`grep -w "^$usbdev" /etc/mtab | cut -d ' ' -f 2`"
test "x$mountpoint" = x && {    # not mounted, so go do it
   mountpoint=/tmp/dir.$$
   test -d $mountpoint || mkdir -p -m 0777 $mountpoint || \
      die "-- could not create $mountpoint"
   mount $usbdev $mountpoint || \
      die "-- error mounting $usbdev on $mountpoint"
   echo ":: $usbdev is mounted on $mountpoint"
}

# do the neccessary backup ...

# unmount
umount $mountpoint
rmdir $mountpoint
unset myusb usbdev mountpoint
exit 0