Lost disk

[depmco]

I have a highly intermittent problem. Occasionally (maybe weekly on average), I lose contact with part of one of my filesystems. I have a 500GB Western Digital disk, /dev/sdb1, that I mount on /home. At the root of the disk are greg and lost+found (e.g. /home/greg and /home/lost+found). /home/greg is my user home directory. When the failure occurs, any reference to /home/greg simply hangs. For example the sequence "cd ~ ; ls" or "ls /home/greg".

mount shows

/dev/sdb1 on /home type ext3 (rw,noatime,commit=0)

Under these circumstances, df also hangs. When I say hangs anywhere above, I mean that the command doesn't return anything, ctl-z will not suspend the command, and the only way to stop the command that I've found is to open a new shell and kill -9 the offending process.In the failing mode, I can successfully ls /home and get

greg lost+found

but as I mentioned, if I ls /home/greg I get the hang. I can, after terminating all processes using /home, unmount and remount /dev/sdb1, but the symptoms persist.

Rebooting the system fixes the problem. Powering down is not necessary. fsck does not reveal any problems with the file system on reboot. I realize I should capture the output from dmesg but I carelessly have failed to do so. I realize that the information here might not be sufficient to diagnose my problem, but I hope someone can tell me what to look for the next time the failure happens so I can get what people need to see in order to help out.

-- Greg

[VinzC]

I once had a similar issue with a USB disk and I think it was due to the fact the disk was put to suspend although I haven't investigated at all. Just an educated suspicion^Wguess. So the trick I found was to run udevadm trigger before doing a backup -- it was a backup disk only though.

So in your case try to run udevadm trigger and see if the disk and partition nodes reappear. You might have to mount the filesystem again.

As for the unmounted /home, it's very weird. The only way I know of to get to such a situation is to umount -l /home (a.k.a. lazy unmount). You'll have to investigate this I guess.
_________________
Gentoo addict: tomorrow I quit, I promise!... Just one more emerge...
1739!