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

Re: ML9.0:file system completely crashed ext3,"could not find mime type",kde3.0.3



--- "Stephen C. Tweedie" <sct redhat com> wrote:
> > I tested my memory with memtest86 but no errors were detected, I
> > read that this is not conclusive.
> 
> Indeed.  How long did you test it for, by the way?  I usually
> recommend
> an overnight run at the very least.

Oops only had it run a few times.


> looking very carefully through your log files for any signs that
might point
> to something specific (such as CRC errors being reported on the disk
> controller).

No CRC errors found in /var/log/ or other kind of disk read/write
errors
prior the ones I mentioned

Thank you for your response.

Civileme gave me this answer on expert linux-mandrake com:
"OK

What happened is this:

With each sector of data the system generates a 57 byte CRC (Cyclic
Redundancy 
Check)

WD drives of the generation you are using do not calculate that same
CRC for 
comparison--they store it in their big 600-byte sectors along with the
512 
bytes of data.  They may calculate the first in a series of writes, but
not 
all.

The result is that a channel error is not always recognized from mobo
to disk, 
only from disk to memory  No request to retransmit is generated by the
disk 
because it is not capable of independently verifying the channel error.

This is NOT new.  Chipset makers have been letting WD get away with
this crap 
for years

http://kt.zork.net/kernel-traffic/kt20000214_54.html#2

Yep February 2000

Anyway WD is the only drive I know that can turn a transitory noise
problem 
into a more or less permanent hardware problem.  The drive is safe
enough run 
at udma2 (33Mhz and 32 byte CRCs) or lower but has this very low
probability 
of failure at udma3 or higher.  But in computers a low probability of
failure 
means that you can calculate how often to expect it.

ext3 or XFS or Reiserfs or JFS are not inherently more secure than
ext2...  
They just recover from errors caused by sudden poweroffs or static from
the 
family cat brushing up against the computer case somewhat faster than
the 
older non-journaling filesystems.  On a larger partition this can be
the 
difference between back up in 45 seconds or 45 minutes, so the
journaling 
filesysem is much preferable for servers.

If you cannot put a better drive in that box, a 40-pin cable could make
a big 
difference for the safe operation, to force 33MHz or lower transfer
speed.  
KDE constantly updates things on disk, like the position of windows on
the 
screen, so drive reliability is critical with that application.
"
This helped though I still have some questions, they are asked in 
expert linux-mandrake com

vatbier

__________________________________________________
Do you Yahoo!?
Yahoo! Shopping - Send Flowers for Valentine's Day
http://shopping.yahoo.com





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