You are not logged in.
Hi !
Since I started using computers in general, and Arch in 2008, I've always been able to solve every one of my problem (and other's too) by myself, but here I need a little help, I'm stuck…
I have two 250G hard drives, partitioned this way :
/dev/sda1 * 63 64259 32098+ fd RAID Linux autodétecté
/dev/sda2 64260 4064444 2000092+ 82 partition d'échange Linux / Solaris
/dev/sda3 4064445 24065369 10000462+ fd RAID Linux autodétecté
/dev/sda4 24065370 488392064 232163347+ 5 Étendue
/dev/sda5 24065433 64067219 20000893+ fd RAID Linux autodétecté
/dev/sda6 64067283 488392064 212162391 fd RAID Linux autodétecté
(sdb is the same)
The raid array is assembled this way :
mdN is sdaN/sdbN
sda2 is swap, so is sdb3 but it's only there to have the same partition table.
md1 is ext2, md3 is reiserfs, md5 and md6 are ext4
Here is the output of $df :
udev 10240 4 10236 1% /dev
tmpfs 1027524 0 1027524 0% /dev/shm
tmpfs 1027524 0 1027524 0% /sys/fs/cgroup
tmpfs 1027524 0 1027524 0% /media
tmpfs 1027524 456 1027068 1% /tmp
/run 10240 408 9832 4% /run
/run 10240 408 9832 4% /var/run
/run 10240 408 9832 4% /var/lock
rootfs 19686680 8929648 9756996 48% /
/dev/md1 30985 16049 13336 55% /boot
/dev/md3 10000072 7611876 2388196 77% /var
/dev/md5 19686680 8929648 9756996 48% /
/dev/md6 208832800 156985276 41239412 80% /home
As you can see, neither partition or drive is full. md5 is 18.8GiB in size, and contains 8.52GiB of data, and doesn't seem to accept more than about 9.6GiB. Every time I try to install some heavy package from AUR (like Trine or Shank) using yaourt (thus writing to /tmp), at some point it triggers a write-error saying that the drive might be full (but it's obviously not).
If I let the drive get "full", then most programs stop acting correctly (firefox refusing to download a file, etc…), complaining that there is no more space left on the drive, and forcing me to do some cleanup.
I booted on a liveCD, and checked every sda* and sdb* partition for badblocks (read-only mode), and got nothing, no badblock at all.
Then I assembled the array, and forced an fsck on each mdN partition (with $e2fsck /dev/mdN -f), got nothing. Forced fsck with read-only badblocks test ($e2fsck /dev/mdN -f -c) and non-destructive read-write ($e2fsck /dev/mdN -f -c -c) with no more success, everything starts to complain again when trying to write more than 1.1GiB on /.
Now I don't know what to do next… How can I know what causes this strange behavior ? How can I (possibly) fix this ? And if it's not fixable (ie hardware), how can I know which of the two drives is failing to replace it ?
Last edited by SataMaxx (2012-01-11 14:39:46)
Offline
You answered your own question: "(thus writing to /tmp)"
/tmp = tmpfs, lies in RAM, default size is half your RAM.
1000
Offline
FUUUUUUUUUUUUUUUUUUUUUUUUUUUUU…
Look through your window, you should see me hanging from a tree… ^^
Wow, I've never felt so stupid than now… and I even didn't see it while fd'ing when it was failing… shame on me !
Anyway, many thanks for the help ! :-)
edit : can someone please delete this shameful thread ? ;-)
Last edited by SataMaxx (2012-01-11 14:52:28)
Offline