Bad blocks on Seagate 400GB SATA HDD in RAID-0

[Rob Nicholson]

We've recently purchased two Seagate 400GB ST34008 SATA hard disks to use in
a RAID-0 secondary server. I was having real trouble getting them working in
the target server so did some tests in another PC - tried full formatting
them. One of the formatted perfectly. The other has 10 bad blocks reported
in the event log.

Question 1: should we simply return the disk with 10 bad blocks or are bad
blocks expected? (I think not)

Question 2: would one drive having bad blocks and other being fine cause
problems with RAID-0, i.e. does it stripe to the same blocks on each disk?

Thanks, Rob.

 

[Mike Tomlinson]

Question 1: should we simply return the disk with 10 bad blocks
Yes.

or are bad
blocks expected?

No. The drive has spare blocks and the firmware replaces bad blocks "on
the fly" with spares. If you see bad blocks at the OS level (and they
are still present after re-writing them) the disk is duff.

Question 2: would one drive having bad blocks and other being fine cause
problems with RAID-0,
Yes.

i.e. does it stripe to the same blocks on each disk?

Depending on whether the controller is hardware/firmware/software RAID,
it's controller/BIOS/driver dependent, but I'd expect any half-decent
RAID to be able to cope with bad blocks. With new drives though, as in
your case, they shouldn't be present at all.

 

[Rob Nicholson]

No. The drive has spare blocks and the firmware replaces bad blocks "on

the fly" with spares. If you see bad blocks at the OS level (and they
are still present after re-writing them) the disk is duff.

Okay thanks for that - I'll arrange for it to be returned. I've formatted
the drives twice now and same number of errors. This is without RAID in the
equation.

Depending on whether the controller is hardware/firmware/software RAID,
it's controller/BIOS/driver dependent, but I'd expect any half-decent
RAID to be able to cope with bad blocks. With new drives though, as in
your case, they shouldn't be present at all.

From your first part about the firmware on the drive replacing bad blocks on
the fly, I infer that actually the RAID controller shouldn't even know about
a few bad blocks as they are replaced with the spare ones.

Cheers, Rob.

 

[Rob Nicholson]

Question 1: should we simply return the disk with 10 bad blocks

Yes.

Hmm, suspicious. Whilst the format of the first disk didn't report any
errors whilst the second disk did, upon reboot there was on bad block
reported on the first disk. I'm starting to suspect the Sil 3112 RAID
controller (yes, here we go again). I've bought a Belkin one as a spare
which I've just put in there and surprise, surprise, it's also based on the
Sil3112 I wasn't expecting that really as I'm bobby moore when I ordered
it, it was listed as a plain non-RAID PCI SATA controller. Ohh well, I'll
reformat them again with this different card and see where we get.

Cheers, Rob.

 

[Rob Nicholson]

it, it was listed as a plain non-RAID PCI SATA controller. Ohh well, I'll

reformat them again with this different card and see where we get.

New SATA RAID controller and they both formatted perfectly. Looks like the
original controller is naff.

Rob.

 

[Rob Nicholson]

New SATA RAID controller and they both formatted perfectly. Looks like the

original controller is naff.

Hmm, this Belkin Sil3112 based RAID controller suffers from exactly the same
problem as the original Sil3112 in that it crashes when you try and create a
RAID array in BIOS. As with the other card, I've had to upgrade it.

More suspicion - the BIOS I upgraded the original card with was superseded
on 9th Jan 2006 - just after I upgraded to the previous version. Hmm, wonder
if there was a problem with the 4.2.76 BIOS...

Rob.

 

[Folkert Rienstra]

The drive has spare blocks and the firmware replaces bad blocks "on
the fly" with spares.

No, it doesn't. It does that only under certain conditions.

If you see bad blocks at the OS level

.... things are different to when you see them on the harware level ...

(and they are still present after re-writing them) the disk is duff.

And where exactly do you gather that he rewrote them?
And where exactly do you gather that he saw them?

If he saw them in the filesystem info then that's where they'll stay nomatter
what you do untill you instruct format specifically to retest the bad blocks.
And if nothing was changed, like a wipe of the drive, then they will just be
found again.

Yes.

Just like it would have with a single drive.

Depending on whether the controller is hardware/firmware/software RAID,
it's controller/BIOS/driver dependent,

No, it's not.

but I'd expect any half-decent RAID to be able to cope with bad blocks.

With new drives though, as in your case, they shouldn't be present at all.

But isn't necessarily the drives fault.

 

[Folkert Rienstra]

Hmm, suspicious. Whilst the format of the first disk didn't report any
errors whilst the second disk did, upon reboot there was on bad block
reported on the first disk. I'm starting to suspect the Sil 3112 RAID
controller (yes, here we go again).

I've bought a Belkin one as a spare which I've just put in there
and surprise, surprise, it's also based on the Sil3112

I wasn't expecting that really as I'm bobby moore when I ordered
it, it was listed as a plain non-RAID PCI SATA controller.

So therefor it obviously can't be a Sil3112.
Every logic defies you, doesn't it Rob.
And it's not that you weren't warned when
you mentioned that Adaptec earlier.

Ohh well, I'll reformat them again with this different card and see
where we get.

Absolutely nowhere.

 

[Rob Nicholson]

I wasn't expecting that really as I'm bobby moore when I ordered

So therefor it obviously can't be a Sil3112.
Every logic defies you, doesn't it Rob.

I think the number of people I have blocked after 20 years of using Usenet
can be counted on one hand. You are the first in a long time going on it.

And as for every logic defies me, this is the said page on Dabs:

http://www.dabs.com/productview.aspx?Quicklinx=3Y8H

Nowhere does it state it's a RAID controller so hence my mild-surprise when
it did turn out to be a RAID controller.

Absolutely nowhere.

Actually, it works perfectly - formatted all 800GB and not an error in
sight. Left it filling up with data from another server as a test and so
far, still not a blip.

What doesn't work is the same card and hard disks in the target server - a
dual AMD Athlon MP set-up with a Tyan Tiger MP board. Windows 2000 Server
locks solid when it tries to format the hard disk. But then again, dual CPU
Athlon MP systems never really worked very well and was a technology that
never really caught on. Want to see a couple of hot CPUs? Check out this
server

And is this because it's poor little SATA RAID and the Sil3112 chipset eh?
I've had my fair share of very similar problems with SCSI sub-systems over
the years. Better now I agree but that's because SCSI is pretty stable (dare
I say ancient) technology and the drivers and firmwares have finally been
sorted out.

And yes, I agree - that is an attraction of SCSI over SATA but as I probably
stated elsewhere, this is a secondary storage server so I don't mind trying
out a few new things. That's how one learns isn't it?

Cheers, Rob.

 

[Odie Ferrous]

I think the number of people I have blocked after 20 years of using Usenet
can be counted on one hand. You are the first in a long time going on it.

Don't block Folkert.

I find if I ever feel depressed, I read his posts and realise I'm not
such a bad person after all, and my spirits are buoyed.

He's most definitely someone you don't wish to be like.


Odie

 

[Folkert Rienstra]

I think the number of people I have blocked after 20 years of using Usenet
can be counted on one hand. You are the first in a long time going on it.

And as for every logic defies me, this is the said page on Dabs:

http://www.dabs.com/productview.aspx?Quicklinx=3Y8H

Nowhere does it state it's a RAID controller

So therefor it obviously can't be a Sil3112, right. That figures.
Where exactly does it say that Sil3112 are all raidcontrollers?

so hence my mild-surprise when it did turn out to be a RAID controller.

And a Sil3112 one to boot. You must have been double unlucky. isn't, Rob.

Actually, it works perfectly - formatted all 800GB and not an error in
sight.

So now you are saying that the other card wrote bad sectors and this one
writes good sectors. Sound logic says that you did something with this card
that you didn't do with the other one, something that corrected the bad
sector situation. I'll bet that the other one will work just as well now.

Left it filling up with data from another server as a test and so far, still
not a blip.

So you did some things differently this time.

What doesn't work is the same card and hard disks in the target server - a
dual AMD Athlon MP set-up with a Tyan Tiger MP board. Windows 2000 Server
locks solid when it tries to format the hard disk. But then again, dual CPU
Athlon MP systems never really worked very well and was a technology that
never really caught on. Want to see a couple of hot CPUs? Check out this
server

And is this because it's poor little SATA RAID and the Sil3112 chipset eh?

You were the one complaining about it.

I've had my fair share of very similar problems with SCSI sub-systems over
the years. Better now I agree but that's because SCSI is pretty stable (dare
I say ancient) technology and the drivers and firmwares have finally been
sorted out.

And yes, I agree - that is an attraction of SCSI over SATA but as I probably
stated elsewhere, this is a secondary storage server so I don't mind trying
out a few new things.

That's how one learns isn't it?

And by implying that one card writes bad sectors and the other does not you have
learned what exactly?

 

[Mike Tomlinson]

Hmm, this Belkin Sil3112 based RAID controller suffers from exactly the same
problem as the original Sil3112 in that it crashes when you try and create a
RAID array in BIOS.

I use a SiI3112-based PCI controller with a Seagate 160Gb drive on
Windows 2000. The controller's BIOS is RAID-capable, but I'm only using
the one disc. No problems, apart from the well-known SiI/Seagate
incompatability, which can be worked around by editing the driver's .inf
file.

 

[Rob Nicholson]

I use a SiI3112-based PCI controller with a Seagate 160Gb drive on

Windows 2000. The controller's BIOS is RAID-capable, but I'm only using
the one disc. No problems, apart from the well-known SiI/Seagate
incompatability, which can be worked around by editing the driver's .inf
file.

Oooh, pray tell about this Sil/Seagate incompatibility. This particular
server already has a Sil3112 RAID controller connected to two 300GB Maxtor
SATA drives. It's worked a treat for months and months and is now just full.

So another Sil3112 has gone in there and these happen to be Seagate SATA
drives this time.

Cheers, Rob.