[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: advice please - io manager magic bad



On Sat, Sep 28, 2002 at 01:47:05AM +1000, Martial Herbaut wrote:
> Following a raid controller crash I am getting this error when attempting 
> to fsck
> 
> -----------
>  /sbin/fsck.ext3 -f /dev/customer_data_volume_group/customer_data_volume
> e2fsck 1.26 (3-Feb-2002)
> /sbin/fsck.ext3: Attempt to read block from filesystem resulted in short 
> read while checking ext3 journal 
> e2fsck: io manager magic bad!
> ------------

"io manager magic bad" is an internal e2fsck error which should never
happen.  It means that data structure passed to a function didn't have
the correct magic number, and is symptomatic of (a) a programming
error, (b) a corrupted executable or shared library, (c) a bad hard
drive or hard drive cable resulting in a corrupted image of the
executable or libext2fs shared libraryin memory, or (d) a memory/cache
error.

No one else has reported any such problems, so the first possibility
is unlikely.  Combined withthe I/O errors reported in your syslogs:

> I/O error: dev 08:21, sector 3211
> I/O error: dev 08:21, sector 9805371

And it looks like you have some kind of hard drive failure.  What's
unusual/strange is that we're not seeing any messages from the scsi
drive detailing exactly what sort of I/O error we're seeing; just that
the SCSI layer is reporting some kind of problem.

In any case, my advice is to copy the data to a known good set of disk
drives ASAP, and then run the badblocks program on the underlying
disks.  If you run the test in read-write (non-distructive if you want
to try to preserve the data, distructive if you don't care since
you've already copied the data to your spare good disks) mode, the
disk may automatically remap the bad blocks, and you may be able to
reuse your disks for some other purpose.  

The reason why I suggest copying the data off is that sometimes
further disk activity may cause more disk blocks to fail (drive
failures are often progressive, with exponentially increasing number
of bad blocks), and my priority is always to assure that the data is
safe before thinking about trying to see whether or not the disks can
be salvaged.  Given how cheap disks are these days, the data on them
is often an order of magnitude or more more valuable that the disks
that they're stored on.

						- Ted





[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]