[Wolves] Things I have learn't in the last 16 hours

[Ron Wellsted]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

David wrote:
>>
>> * [disk manufacturer who will remain nameless] has
>> developed a disk that can detect if its part of a RAID
>> set and deliberately corrupt your data, forcing you to
>> buy the more expensive "server" version of the disk.
>> This was deemed clever, with a side order of evil, in
>> a  "how the fsck?" sauce.
> 
> Can you provide any more info on this? They clearly need to be avoided !
> 
> Is this when using a controller card, software RAID? raid 0, 1, 5 ?
> 
> (More info please please please)
> 
> David.
> 

Cryptic clue: a cow going to the ocean might walk through this.  (email
me off list and out of public view if you want confirmation)

I was involved in a project where 2 servers using Windows Server 2003
were built each using 4 x 400GB SATA drives in a software RAID 10 (one
file/print, other Exchange for about 500 users). Everything built
perfectly and was installed on the customer site and the data
transferred from the old servers. Within a couple of days of live usage,
the servers were crashing and corrupting data.

After investigation, we were blaming the SATA chipset for incorrectly
implementing command queueing and installed hardware RAID controllers.
This improved the situation but did not cure the problem.  We then
discovered that the drive manufacturer web site noted that these drives
were designed for workstation use and not for use in a RAID (which
everyone swore was a recent addition, but was unable to find any proof
(not even via the wayback machine!)) and the new improved model with an
magic "S" on the end of the part number should be used for servers (this
drive had only just been released).  We cured the problem in the end by
switching to a different manufacturers drives (cryptic clue: not
oriental, not analog).

Over the next few weeks we then replaced ALL of these disks in all our
customers (at our cost!).

The original manufacturer has sent since us 4 of their new 750GB
(workstation) drives and 4 of the "S" drives.  Again, when built into a
RAID (h/w or s/w) 0, 1, 5 or 10 the workstation drives will corrupt the
filesystem under heavy load.  The "S" drives work perfectly, even when
serving files close to 1Gb/s (we have a VERY nice toybox to play with
which almost fills 2 x 45U racks).

- --
Ron Wellsted
ron at wellsted.org.uk http://www.wellsted.org.uk
N 52.567623, W 2.137621 Linux Counter No. 202120
FWD:519961
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.3 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iQEVAwUBRaaXdktP/KMNOfRbAQLD2wgAwEQFVC6JFPyMUMIDskejEoSiTXx19b5t
PKwXbK68aOvb2Jpu4q7xKClNQAVDJfKaHCjtr57+2TT5hblK8Fcn9mgWpz1H4ZtB
q8MOsOgYhcc90Tl2IOkTTWKedRehANSLTedk/GAN1VIfzIRUVCaV+b3slUtQLCi4
0ZQJThpbgNYST7kIz073PVuPxmUCKdFxM6mudRCDpLdDBJvL9QWO+nN8guth2VFc
NqE2q+BnGU8Se/j0Mz4eDfbpnIQt9FTTcI0EDCb3vp8IOblB6CHXY9kF7OF6M3xM
CNvwz7n6M3O7Vv/PH5aohXTt8ZEcfmZFUnl9DJ8y+CJdvETWmmkyBA==
=Wah0
-----END PGP SIGNATURE-----