I have just restarted the daemon and am waiting for it to go down again. We shall see.

I will let you know if this has an effect.
If I go into RO mode, then I will do what Coronas said about swap files and work down the list - thanks for the suggestions - much appreciated.

Edit - has anybody set up a chroot environment on this box? This might work…
Edit2 - … or not. Probably just a stupid idea, it really is easier to check the drive in a linux box.

In addition to my last post, just rebooting isn't fixing the problem. The corrupt block is still sitting there waiting kick you into ro mode the next time your book tries to read from it.

The reason that the device drops into ro mode is that it encounters a corrupt block in your hard drive. usually you would be prompted to run fsck to fix the problem. There are two problems with this a) fsck should not be run on a mounted partition b) the MBSW seems to run out of memory when running fsck anyway.

What we don't know is what is causing the corruption. I suspect that it is hardware related, I have never had the problem myself despite using the same firmware as everyone else.

I don't know if anyone has already, but it would be nice if someone could run fsck on a corrupt partition while connected to a linux box and see if that solves the error. This would certainly be easier than reformatting the whole drive.

I reformatted the partition and recreated my directory structure etc. which stopped the device becoming Read Only so quickly.

It eventually became Read Only though and now takes less time to go back into that state. ( a few minutes )

I have to do a double reboot now like DreaMr

I get this when I # dmesg | tail

<6>Using fractional divider baud 115200, clock 100000000 dlf 40
<6>Using fractional divider baud 115200, clock 100000000 dlf 40
<1>journal_bmap: journal block not found at offset 2060 on md1
<3>Aborting journal on device md1.
<2>EXT3-fs error (device md1) in ext3_reserve_inode_write: Journal has aborted
<2>EXT3-fs error (device md1) in ext3_dirty_inode: Journal has aborted
<2>EXT3-fs error (device md1) in ext3_new_blocks: Journal has aborted
<2>ext3_abort called.
<2>EXT3-fs error (device md1): ext3_journal_start_sb: Detected aborted journal
<2>Remounting filesystem read-only

I have seen alot of different threads about the device going Read Only, and everyone seems to blame a different problem.

I think that it simply runs out of memory - something goes wrong and that's it Read Only ( but I am not a Unix expert )

I think I am going to kill most of the processes and see how long it lasts.

Has anyone found a solution to this yet, one that doesn't involve reformatting?

I used to had trouble with transfering files at first, and I found out that formatting helps. It did do the job, and all was fine for little while, but now the permissions get lost about every 24 hours or so again, though not because of filetransfers now. I really wouldn't like to reformat again, because it would be pretty troublesome to get all the stuff I have in there back to somewhere else for the duration (hard drive space beign the issue here)… and if the problem comes again like it did now, it's not really worth it anyways. I get the following kernel messages too when the shares fail:

<1>journal_bmap: journal block not found at offset 1036 on md3
<3>Aborting journal on device md3.
<2>ext3_abort called.
<2>EXT3-fs error (device md3): ext3_journal_start_sb: Detected aborted journal
<2>Remounting filesystem read-only

Have to reboot mybook twice to get it working again. The first boot causes the drive to panic with flashing leds, second one gets it up again with some warning in kernel messages:

<4>EXT3-fs warning (device md3): ext3_clear_journal_err: Filesystem error recorded from previous mount: IO failure
<4>EXT3-fs warning (device md3): ext3_clear_journal_err: Marking fs in need of filesystem check.
<4>EXT3-fs warning: mounting fs with errors, running e2fsck is recommended
<6>EXT3 FS on md3, internal journal
<6>EXT3-fs: recovery complete.
<6>EXT3-fs: mounted filesystem with ordered data mode.

It's really a pain in the arse to do the double reboot once in a day, especially because it requires resetting from the power button. Otherwise a cron job could be a temporary fix.

My fstab
/dev/root / ext3 rw,noauto,noatime 0 1
proc /proc proc defaults 0 0
devpts /dev/pts devpts defaults,gid=5,mode=620 0 0
sys /sys sysfs defaults 0 0
/dev/md2 none swap defaults 0 0
/dev/md3 /var ext3 defaults,noatime 0 2
/dev/md4 /shares/internal ext3 defaults,noatime 0 2

/dev/root on / type ext3 (rw,noatime,data=ordered)
proc on /proc type proc (rw)
devpts on /dev/pts type devpts (rw)
sys on /sys type sysfs (rw)
/dev/md3 on /var type ext3 (rw,noatime,data=ordered)
/dev/sda4 on /shares/internal type ext3 (ro,noatime,data=ordered)
usbfs on /proc/bus/usb type usbfs (rw)

I ran these commands to get it back into read write mode gave it a reboot and its working again

umount /dev/sda4

mount /dev/sda4 /shares/internal -t ext3 -o rw data=ordered

mount

/dev/root on / type ext3 (rw,noatime,data=ordered)
proc on /proc type proc (rw)
devpts on /dev/pts type devpts (rw)
sys on /sys type sysfs (rw)
/dev/md3 on /var type ext3 (rw,noatime,data=ordered)
usbfs on /proc/bus/usb type usbfs (rw)
/dev/sda4 on /shares/internal type ext3 (rw,noatime,data=ordered)
[root@MyBookWorld /]#

Will see how long it lasts before going again……

Edit - lasted about 5 mins before going read only again :-( I copied about 1gb onto it…..

/dev/sda4 on /shares/internal type ext3 (ro,noatime,data=ordered)

Anyhow, looks better now.

Andy

Anyhow, no write errors found on the partition to match the no read errors previously.

Andy

Apr 13 18:25:40 MyBookWorld user.alert kernel: journal_bmap: journal block not found at offset 3084 on md4
Apr 13 18:25:40 MyBookWorld user.err kernel: Aborting journal on device md4.
Apr 13 18:25:40 MyBookWorld user.crit kernel: ext3_abort called.
Apr 13 18:25:40 MyBookWorld user.crit kernel: EXT3-fs error (device md4): ext3_journal_start_sb: Detected abo rted journal
Apr 13 18:25:40 MyBookWorld user.crit kernel: Remounting filesystem read-only
Apr 13 18:25:40 MyBookWorld user.warn kernel: __journal_remove_journal_head: freeing b_frozen_data
Apr 13 18:29:18 MyBookWorld user.warn kernel: __journal_remove_journal_head: freeing b_committed_data
Apr 13 18:29:19 MyBookWorld user.notice block.agent[1515]: USB umount event not supported
Apr 13 18:48:28 MyBookWorld user.info kernel: kjournald starting. Commit interval 5 seconds
Apr 13 18:48:28 MyBookWorld user.warn kernel: EXT3-fs warning: mounting fs with errors, running e2fsck is rec ommended

(of course fsck isn't possible, still. System obviously could do with a larger swap partition.)

I'm running a reformat with full write check on the md4 partition - it hasn't yet found any write
errors in the past 2 hours … at present rate it'll be another day before it completes.
A full read check (dd if=/dev/md4 of=/dev/nul) completed with no errors.

Looks like a software problem, possibly a race condition.

Andy

After transferring about 2000-7000 files over the network or untaring over ssh the drive locks up and goes into read only mode. After a reboot /dev/md4 fails to mount, so I run fsck on it. 3 cups of tea later it finally finishes and after another reboot I can begin writing again… until the drive locks up again. I've repeated this about 3 times today and have only managed to get 10000 of my many tiny files onto the drive.

I havent tried copying over a large file yet to see just what happens, I'm not sure if there could be any issues with copying 20000+ files to a single directory or if its simply the drive which is faulty.

Has anyone found a working solution and would they like to share it?

blight

I know that from my OS X machine, it will default to the unprivileged guest account unless I specify alternate auth details. When this happens I lose write access to the share.

On setting up telnetd and investigating I find that the initial mount stage:

[root@MyBookWorld /]# mount
/dev/root on / type ext3 (rw,noatime,data=ordered)
proc on /proc type proc (rw)
devpts on /dev/pts type devpts (rw)
sys on /sys type sysfs (rw)
/dev/md3 on /var type ext3 (rw,noatime,data=ordered)
/dev/md4 on /shares/internal type ext3 (rw,noatime,data=ordered)
usbfs on /proc/bus/usb type usbfs (rw)

changes to /shares/internal being "ro" for no apparent reason.
(I've also observed this for /var, but this (obviously) doesn't lead to quite the same problem)

any idea why this happens … and what I can do about it?

Andy

total 24
drwxr-xr-x 4 root root 4096 Mar 26 2007 .
drwxr-xr-x 21 root root 4096 Apr 9 10:34 ..
drwxr-xr-x 2 root root 4096 Mar 26 2007 external
drwxr-xr-x 6 root root 4096 Apr 8 21:50 internal

total 40
drwxr-xr-x 6 root root 4096 Apr 8 21:50 .
drwxr-xr-x 4 root root 4096 Mar 26 2007 ..
drwx-- 2 root root 4096 Apr 8 21:30 .senvidData
drwx-- 2 www-data root 4096 Apr 9 03:16 MOVIES
drwxrwxr-x 2 root www-data 4096 Apr 8 21:30 PUBLIC
drwx-- 2 root root 16384 Apr 9 02:12 lost+found

you should see that the folders belong to www-data, make sure this is the case, use chown if not

After playing with SSH and a few other hacks I have reverted back to the standard config using the recsue procedure. All seemed well yesterday and I set serveral files copying over however at around midnight I lost permission to write to the share.

The share was created using the web gui and I gave everyone full permissions. I have tried re-applying the permissions but its still not working. Short of hacking in SSH again is there any way of fixing this? Also is there any reason for it breaking in the first place? I have tried mapping drives using the root account and another account (with full perms) with no luck.

Ta B3ast