Need to FSCK regularly
 

Hi,

I'm using a small box, running Fedora Core 5, with a 40gig, 200gig and a 320gig. Its running as a Samba file server for all my windows machines.


Since I started to fill my 320gig i have to ran fsck almost every day.

I copy file on it, and when the problem start, i have that error:

mv: cannot create regular file `/mnt/320gig/misc/xxxxxxxx': Read-only file system (xxx is a filename)

The only thing i can do to fix that, is to unmount the drive, and run:

[root@trinity ~]# fsck -C -V /dev/hdg1 -a


Any Idea?
Thanks

And how much space is free???
From the console or xterm session try running; df

Have you looked into removing the junk, like old files, un-needed files (like the yum cached downloaded rpm files) or temp files??

The concerned drive is "/dev/hdg1". On this drive, nothing related to the OS, only files from users on the network.

It's very likely the drive is on its' last legs.
Do this as root:
yum -y install smartmontools
/sbin/service smartd start
/sbin/chkconfig smartd on
smartctl --test=long /dev/hda (I'm assuming the drive is hda, change appropriately)
it'll tell you how long the test will take (don't worry it can run on a live system)
finally run:
smartctl -a /dev/hda

and see what it has to say about your drive.

Btw, I think it's a good idea to run the diagnostic program designed for the drive... they're able to map out the bad sectors with reserved ones that most drives have these days. Seagate has seatools, maxtor has powermax, etc.. these are free to download from the manufacturers web sites. Note that data on any bad sector will be lost forever, so best to run some type of backup if possible.

Finally, after that, run fsck again. I had problems with a Maxtor drive that was doing the same thing as your drive is now... I think the drive was overheating, which caused the failure to happen sooner than later. Drives today really need to have some type of active cooling.

No error.. everything is ok.. and, its a brand new drive (2 week old)

froussy,
Quote from your original post:
...So this looks like it is a permissions problem.

To help you, we need to know what sort of filesystem you have on the disk, the command used to mount the disk, and the output of ls -l /mnt/320gig

note that ext3fs will remount as read-only when problems occur. This happened to that drive I had mentioned was dieing. Must be a safety thing.

Of course, it could still be a permission thing. :)

Always? Or just when you tell it to? Check /etc/fstab. The only remounting I'd expect on my system would be for / , because of the following specific fstab entry:
I would check your syslog for any problems relating to disk, and also post your /etc/fstab here. Also, try the file move as root and report what happens. Your example fsck command indicates you're running as root (duh!), but you really didn't say about the actual mv command. This all could be a simple mount/permissions issue. But if you're triggering some fstab "errors=remount-ro" then it leans more towards filesystem corruption of hardware problems.

In my case once the system detected a filesystem error, it went directly to read only mode automatically. Nothing to do with fstab. Probably designed in to stop errors from cascading out of control (much like a kernel panic)

Ok, there is few answer:

Filesystem is EXT3

When I had the problem, i was on a samba share (from a Win machine)..
then, i looked on the box, as root, and moving or deleting file was impossible on that drive

Why do you have two entries in fstab for the partition??

And since others will be accessing this partition for read/write access (I assume) it might be better to use something like;

In my experience, when a mounted drive suddenly becomes a "read-only" filesystem, it's of two things in most cases but not all. a) it's a controller issue, not a drive or b) it's an outdated or corrupt driver if you're using a raid controller or external card to connect this drive to, etc.

But stop running fsck on the drive, your probably wasting your time. Try unmounting and remounting, if you can write to it, I would suggest looking into what I just pointed out.

If he ran tests on the drive and they came up clean, you'd have to rule out the controller as well. (This is a assuming he actually ran the diagnostic disc from the drive manufacturer like I had asked him to).

The only thing left would be software, so it's very likely it is a driver problem.

Froussy, can you post the output of smartctl -a /dev/hd? where ? is the drive that continously gets corrupted?

I dont think it will be a driver problem, because it should happen on my others drives, No?

There is the output as requested: