SCSI drive corruption

[Randy Starkey]

Hi,

I have an Atlas II 15K 147 gig SCSI drive on an Adaptec controller.

I keep getting corrupt mft's and volumes when I add more than 50 or so files
to the drive. Under 50 files it seems to run OK.

This drive is on a server 2003 box.

Questions...

1) The cable is 3 feet long, with 4 connector on it. The drive is on the
last connector and terminated there. Too long? I am ordering another cable
shorter to test.

2) Bad termination could cause this? I ordered another terminator also.

3) Does formatting it as a primary or extended partition matter? Both have
exhibited the same behavior anyway. It is not a boot drive, just data only.

4) I am getting ready to run Maxtor's drive test software on it now.

5) It is in a carrier that allows the ID to be set via a switch. I will try
it outside of that also to eliminate that.

Anything else I might be missing?

Also, in WinServer2003 - when I pull the drive out of the carrier, why does
explorer continue to show the drive? I can even open a text file on the
drive with it not in there! Give a delayed write error though. Is something
cached, or what? Seems like the moment I pull the drive it should go away.

Thanks!

--Randy Starkey

 

[Arno Wagner]

Hi,

I have an Atlas II 15K 147 gig SCSI drive on an Adaptec controller.

I keep getting corrupt mft's and volumes when I add more than 50 or so files
to the drive. Under 50 files it seems to run OK.

Do yo get corruption wehn you write one really large file?

This drive is on a server 2003 box.

1) The cable is 3 feet long, with 4 connector on it. The drive is on the
last connector and terminated there. Too long? I am ordering another cable
shorter to test.

What SCSI variant? They can take 1.5 ...12 meters, depending on
type. Seagate has a good chart here:

http://www.seagate.com/support/kb/disc/scsi_interface_cable.html

If in doubt turn on bus parity and see whether you get parity errors.

2) Bad termination could cause this? I ordered another terminator also.

Yes. But if parity is on, this should be detected.

3) Does formatting it as a primary or extended partition matter? Both have
exhibited the same behavior anyway. It is not a boot drive, just data only.

Should not make any difference.

4) I am getting ready to run Maxtor's drive test software on it now.

Do that. It has an onterface test as well, if I remember correctly.

5) It is in a carrier that allows the ID to be set via a switch. I will try
it outside of that also to eliminate that.

The switch is read once during SCSI device reset. After that it
is ignored. So even if the swicth is flapy, it does not matter.

Anything else I might be missing?

Bad RAM in your PC would be a suspect as well. Run memtest86+ for a
night or so.

Also, in WinServer2003 - when I pull the drive out of the carrier, why does
explorer continue to show the drive? I can even open a text file on the
drive with it not in there!

OS buffering.

Give a delayed write error though. Is something
cached, or what? Seems like the moment I pull the drive it should go away.

SCSI does not have an unplug notification. The OS cannot tell. If you
do something as dangerous as pulling a drive without telling the OS
first, you should at least disable any buffering on it. Even with
that you can completely destroy your filesystem.

Arno

 

[Randy Starkey]

Do yo get corruption wehn you write one really large file?

no - it seems to be more little files - lots of them.

What SCSI variant? They can take 1.5 ...12 meters, depending on
type. Seagate has a good chart here:

Ultra320


http://www.seagate.com/support/kb/disc/scsi_interface_cable.html

If in doubt turn on bus parity and see whether you get parity errors.

This is a jumper?

Yes. But if parity is on, this should be detected.


Should not make any difference.


Do that. It has an onterface test as well, if I remember correctly.


The switch is read once during SCSI device reset. After that it
is ignored. So even if the swicth is flapy, it does not matter.


Bad RAM in your PC would be a suspect as well. Run memtest86+ for a
night or so.


OS buffering.


SCSI does not have an unplug notification. The OS cannot tell. If you
do something as dangerous as pulling a drive without telling the OS
first, you should at least disable any buffering on it. Even with
that you can completely destroy your filesystem.

Arno

OK on all else.

 

[Arno Wagner]

no - it seems to be more little files - lots of them.

Hmmm. Then it is likely a software issue....

This is a jumper?

Should be a setting in the controller BIOS.

OK on all else.

No Problem...

Arno

 

[Randy Starkey]

One other issue I'm wondering if could affect this...

The many files I'm copying that seem to corrupt the SCSI drive are coming
from a NT server on another domain (trusted) so the security permissions on
those files include "everyone" (I added that) but they also include some
permissions like john doe/abc domain and the SCSI drive is on def domain.
But - I wouldn't think security permissions could trash a drive, would you?
Thanks!

I'll post back after the other tests. The reason I'm suspicious of the above
is because the drive has worked fine with other files, until I added these
(albeit there are thousands of them)

--Randy

 

[Randy Starkey]

I also just discovered that if I change the SCSI ID that WinServer 2003
won't recognize the drive. Is it perhaps necessary to somehow delete or
remove the drive from disk manager in order to change the SCSI ID and get it
to show up? Thanks.

--Randy Starkey

 

[Folkert Rienstra]

Do yo get corruption wehn you write one really large file?

What SCSI variant? They can take 1.5 ...12 meters, depending on type.

Having trouble reading babblebot: Atlas II 15K

http://www.seagate.com/support/kb/disc/scsi_interface_cable.html

If in doubt turn on bus parity and see whether you get parity errors.

As if that has anything to do with 'SCSI variant'.

Yes.

Very unlikely.

But if parity is on, this should be detected.


Should not make any difference.

Do that. It has an onterface test as well, if I remember correctly.

Your memory has always been your problem.

So it likely has a stub piece of cable before it is connected to the cable.
That is a no-no, unless it is at the end of the cable

The switch is read once during SCSI device reset. After that it
is ignored. So even if the swicth is flapy, it does not matter.


Bad RAM in your PC would be a suspect as well. Run memtest86+ for a
night or so.


OS buffering.

Seems like the moment I pull the drive it should go away.

SCSI does not have an unplug notification.

But enclosure services does.

The OS cannot tell.

So yes it can with help of enclosure services.
Apart from that it could always poll the drive if it wanted to.

 

[Folkert Rienstra]

I also just discovered that if I change the SCSI ID that WinServer 2003
won't recognize the drive. Is it perhaps necessary to somehow delete or
remove the drive from disk manager in order to change the SCSI ID and get it
to show up? Thanks.

--Randy Starkey

Nope.

 

[Randy Starkey]

Having trouble reading babblebot: Atlas II 15K


As if that has anything to do with 'SCSI variant'.


Very unlikely.
OK.

Can't run this as it doesn't support Ultra320 controllers. I also have been
hearing some noises coming from the drive, so I'm beginning to think it's
the drive. It can't seem to hold any kind of partition/volumne for long
without corrupting. Other SATA drives on the same box run smooth.

Your memory has always been your problem.


So it likely has a stub piece of cable before it is connected to the
cable.
That is a no-no, unless it is at the end of the cable

no cable stub. It's a direct connector plugin insode the carrier.

 

[Arno Wagner]

Can't run this as it doesn't support Ultra320 controllers. I also have been
hearing some noises coming from the drive, so I'm beginning to think it's
the drive. It can't seem to hold any kind of partition/volumne for long
without corrupting. Other SATA drives on the same box run smooth.

no cable stub. It's a direct connector plugin insode the carrier.

Don't mind the Folkert. He does not know what he is talking about...

Arno

 

[Folkert Rienstra]

Your newclient is breaking quoted lines.

Can't run this as it doesn't support Ultra320 controllers.

SCSIMAX doesn't support Ultra320 controllers?

That would very much surprise me. That thing is old as methusalem, it doesn't
care about controllers as long as it can see the drive through ASPI.
Did you actually check that or did you just read that from the description
on the Maxtor site?

I also have been hearing some noises coming from the drive, so I'm beginning

to think it's the drive. It can't seem to hold any kind of partition/volumne

for long without corrupting. Other SATA drives on the same box run smooth.

Instead of SCSIMAX you can also run Bart Lagerwey's SCSITool
http://www.nu2.nu/scsitool/

Massive rephrase:
So there likely is a stub (traces between cable connector and dock connector,
traces between tray connector and driveside connector, piece of cable be-
tween driveside connector and drive) before it is connected to the SCSI cable.

The real physical end of the cable, so the terminator goes into
the tray too and the stub just becomes part of the SCSI cable.
If that is not possible, either not use a removable tray or use one with
a minimum of connectors and the connectors mounted back to back using
minimal traces, to minimize the stub, similar as with SCSI backplanes.

no cable stub. It's a direct connector plugin insode the carrier.

So it's an SCA drive then. You sure? Have a link for the product?

 

[Randy Starkey]

I'm not the best with quoted lines and etiquette <g> See also below...

Your newclient is breaking quoted lines.




SCSIMAX doesn't support Ultra320 controllers?

That would very much surprise me. That thing is old as methusalem, it
doesn't
care about controllers as long as it can see the drive through ASPI.
Did you actually check that or did you just read that from the description
on the Maxtor site?

Maxtor said that on in their PDF on it, and I couldn't find an ASPI driver
for my Adaptec 39320. Nothing on Adaptec's site AFAICT. What ASPI can I use
to try it?

to think it's the drive. It can't seem to hold any kind of
partition/volumne

Instead of SCSIMAX you can also run Bart Lagerwey's SCSITool
http://www.nu2.nu/scsitool/

OK. Will ck that out. Probably still need an ASPI driver though I assume...

Massive rephrase:
So there likely is a stub (traces between cable connector and dock
connector,
traces between tray connector and driveside connector, piece of cable be-
tween driveside connector and drive) before it is connected to the SCSI
cable.

No cable, but traces I'm sure. But I assume those things are tested and
supposed to work. I did try the drive outside the carrier though with no
success. It froze MMC and Win2003 wouldn't even boot. It was set to a
different ID is all. So I'm wondering if the drive is bad. I suppose it
could be the controller too, but it seems to be OK. I wish I had another
good SCSI drive to measure it against. How important is termination power to
be jumpered at the Atlas. I don't think it is, and Maxtor seems to recommend
it?

The real physical end of the cable, so the terminator goes into
the tray too and the stub just becomes part of the SCSI cable.
If that is not possible, either not use a removable tray or use one with
a minimum of connectors and the connectors mounted back to back using
minimal traces, to minimize the stub, similar as with SCSI backplanes.


So it's an SCA drive then. You sure? Have a link for the product?

No, it's 68pin LVD/SE.
http://www.maxtor.com/portal/site/M...erprise Storage/Atlas 15K Family/Atlas 15K II

 

[Randy Starkey]

Don't mind the Folkert. He does not know what he is talking about...

Arno

maybe with you both I'll figure it out! <g>

 

[Folkert Rienstra]

I'm not the best with quoted lines

I didn't say you were, I said your newsclient was.
So tell your newsclient not to do that anymore
(Set your line breaks at 132 or so).

and etiquette <g> See also below...



Maxtor said that on in their PDF on it, and I couldn't find an ASPI driver
for my Adaptec 39320.

No? Found it in 10 seconds.

Nothing on Adaptec's site AFAICT. What ASPI can I use to try it?
http://search.adaptec.com/site/sear...&from=altWord&collection=All&query=39320 aspi


OK. Will ck that out. Probably still need an ASPI driver though I assume...


No cable, but traces I'm sure.

But I assume those things are tested and supposed to work.

If it specifically says Ultra320 on it, yeah. Otherwise, don't think so.

I did try the drive outside the carrier though with no success.

It froze MMC and Win2003 wouldn't even boot.

But it does when you put it back?

It was set to a different ID is all.
Why?

So I'm wondering if the drive is bad.

I suppose it could be the controller too, but it seems to be OK.

Yes. Now that you mention it: A badly seated Hostbus adapter
can cause problems that you normally assign with drive problems.
(I once had that happen where the drive would function mostly OK
in DOS but fail as soon as Windows was started. Not something
you would immediately associate with a badly seated controller).

I wish I had another good SCSI drive to measure it against.

How important is termination power to be jumpered at the Atlas.
I don't think it is,

Depends upon the distance that it is away from the Host adapter.

and Maxtor seems to recommend it?

Can't hurt to enable it for comparison reasons.

No, it's 68pin LVD/SE.

While it's possible to devise a backplane connector arrangement that
directly fits the drive, I've never seen one. This would be a first.
I am almost 100% sure that is has a clable connection inside.

 

[Randy Starkey]

Folkert - OK - changed line break from 76 to 132. See how it looks.

more below...

I didn't say you were, I said your newsclient was.
So tell your newsclient not to do that anymore
(Set your line breaks at 132 or so).


No? Found it in 10 seconds.


http://search.adaptec.com/site/sear...&from=altWord&collection=All&query=39320 aspi

Ah! I missed the DOS stuff in the file, duh...

If it specifically says Ultra320 on it, yeah. Otherwise, don't think so.

I finally called Maxtor, when I heard some weird noises in the drive, so I'm swapping it out with a new one. I will be able to at
least compare that. I'll also reseat the controller and a few other things. Maxtor said the firmware revision i had also had some

But it does when you put it back?

Yup! Although right before I called Maxtor it was not appearing in disk manager at all. So I'm thinking it died. I really hope
that's it! It will end my search.

Why?

Just to see if it would make any difference.

Yes. Now that you mention it: A badly seated Hostbus adapter
can cause problems that you normally assign with drive problems.
(I once had that happen where the drive would function mostly OK
in DOS but fail as soon as Windows was started. Not something
you would immediately associate with a badly seated controller).



Depends upon the distance that it is away from the Host adapter.

1 ft.

Can't hurt to enable it for comparison reasons.

I'm looking for the link... It's not a name brand. Maybe I should get a CRU...