filesystem becomming read only

Theodore Tso tytso at mit.edu
Mon Jan 29 01:24:28 UTC 2007


On Sun, Jan 28, 2007 at 06:05:33PM -0600, Tim Rupp wrote:
> Thanks Ted, I'll go through that list and try swapping the original
> parts with spares that I have around home.
> 
> I've run fsck since the problem started occurring and it _has_ found
> problems with the filesystem. I don't have the output on hand, but I can
> definitely make the filesystem go read-only again. When I do, I can send
> another mail with the attached output from the fsck. Maybe it will help
> to find the problem.

Well, the most important thing about the fsck error is to see whether
it looks like a single bit error, or an entire block being corrupted,
or a block getting written to the wrong location on disk.  (The last
two can be hard to differentiate, but you see ASCII text in an inode
table block, or an block/inode bitmap, that's usually a good clue that
it was the latter.)  

But at the end of the day, it looks like a hardware problem, and this
won't necessarily tell you exactly what is to blame, so it's not a
high priority thing to do.

You could try using badblocks -w (warning, this is a distructive
read/write test) or badblocks -n to see if you catch the disk doing
something wrong, but it may be that creating a filesystem and then
running your workload will be the best stress test.  Unfortunately we
don't have a good disk drive exerciser that exercises the disk with a
lot of random access read/write and seek patterns in Linux, at least
not as far as I know, anyway.

Good luck,

						- Ted




More information about the Ext3-users mailing list